Sei sulla pagina 1di 79

Contents

1 Scope and mission 2

2 Statical indeterminacy 3
2.1 Illustration of the two basic methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Force method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Displacement method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Force method 6
3.1 Steps of the force method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Single redundant structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.1 Force load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.2 Thermal effect – temperature gradient . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.3 Support settlement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.4 Continuous beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.3 Normal forces effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4 Thermal effect – homogeneous temperature change . . . . . . . . . . . . . . . . . . . . . 13
3.5 Single redundants in trusses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5.1 Simple truss, force load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.5.2 Simple truss, thermal effect, compatibility check . . . . . . . . . . . . . . . . . . 15
3.5.3 Checks on the force method solutions . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5.4 Sample frame solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.5.5 Sample truss solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.6 Deflections in statically indeterminate structures . . . . . . . . . . . . . . . . . . . . . . . 19
3.6.1 Deflection in the sample frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.6.2 Deflections in the sample truss . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7 Multiple redundants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.7.1 Clamped beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.7.2 Simple frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.8 Symmetry issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.9 Examples on multiple redundants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.9.1 Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.10 Multiple redundants in trusses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.10.1 Symmetric load case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.10.2 Antisymmetric load case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.11 Selection of the primary structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.12 Three moment equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.12.1 Applications to simple instances . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.12.2 Load terms in the three moment equation . . . . . . . . . . . . . . . . . . . . . . 33
3.13 Method of least work and Castigliano’s second theorem . . . . . . . . . . . . . . . . . . . 33
3.13.1 Castigliano’s second theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.13.2 Method of least work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4 Slope deflection method 36


4.1 Basic equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Moment equilibrium equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.3 Beam with a hinge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4 Example frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

1
4.5 Frames with sidesway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5.1 System matrix symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.5.2 General planar frames with joints translations . . . . . . . . . . . . . . . . . . . . 45
4.5.3 Homogeneous temperature changes and support settlements . . . . . . . . . . . . 46
4.5.4 Homogeneous temperature and support settlements in general frames . . . . . . . 46
4.5.5 Identification of primary unknowns . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.6 Closing remarks on the slope deflection method . . . . . . . . . . . . . . . . . . . . . . . 49

5 Displacement method 49
5.1 End forces and moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2 Equilibrium equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

6 Moment distribution (Cross) method 52


6.1 Distribution factors by the slope deflection method . . . . . . . . . . . . . . . . . . . . . 53
6.2 Iteration, demonstration example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

7 Secondary bending moments in trusses 55

8 Gridwork frame structures 57


8.0.1 Beam torsional stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8.1 A simple rectangular gridwork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
8.2 General gridworks, matrix formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

9 Appendix – solved examples 63


9.1 Force method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.1.1 Cantilever with end roller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
9.1.2 Frame with arch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
9.1.3 Continuous beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
9.1.4 Roof truss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
9.1.5 Simple truss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
9.1.6 Symmetric clamped arch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
9.1.7 Symmetric truss with multiple redundants . . . . . . . . . . . . . . . . . . . . . . 69
9.2 Slope deflection method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
9.2.1 Frames without sidesway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
9.2.2 Frames with sidesway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9.3 Displacement method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
9.4 Gridwork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

1 Scope and mission


The notes are focused on traditional elementary methods of structural mechanics and exclusively, ’man-
ual’ solutions. Experience in using these methods is believed to provide a sound base for understanding
computer codes functions and to give the student an engineering feeling at the same time. These methods
include the force method, the slope deflection and displacement method and some their clones. Active
knowledge and skill are promoted in order to foster the engineering feeling and experience mentioned
above. The didactic method features commented examples as the main tool. This is hopefully consistent
with the promotion of active knowledge.
In a design office, engineers hardly will ever have to solve a statically indeterminate structure man-
ually. The availability and easy use of computer programs might mislead one to the conclusion that the

2
knowledge of the methods hidden in them is superfluous. Seasoned structural engineers warn of this ’black
box’ attitude toward computerized analysis. Allusions to computer codes are made in the section on the
displacement method. They hopefully provide bridges between the methods explained herein and more or
less the standard features of the structural analysis codes. Students should be able to use the codes with
balanced confidence, criticism and understanding.
Contemporary structural mechanics has developed concepts that put the force and displacement methods
in a broader context. Generalized principles (Reissner, Hu-Washizu) allow for the force and displacement
quantities to become primary unknowns at the same time. The two classic methods are perceived as special
cases. Applications of the generalized principles are mainly found in the analysis of multidimensional
continua (three dimensional solids, shells and plates). Computer codes for the analysis of lattice structures
(planar and spatial frames and trusses) still rely on the classic displacement method.
Notation, symbols and graphic conventions, units
Throughout the notes, some simplifying conventions are used. The default coordinate system is the ordinary
cartesian x y frame. Arrows indicate the positive sense of a quantity (force or displacement) in most
cases. Default positive sense is used for the cartesian components of forces (positive senses of the default
coordinate frame) and for the reactions in rollers and links. The reaction in a roller is positive when acting
in the direction of the point of the corresponding triangle token. The reaction in a link is positive when
it is tension. An explicit redefinition of the positive sense of a definite quantity by an arrow in a diagram
overrides the default one. This happens for example, in free-body diagrams of planar frames when the x
and y components of the forces may not conform to the standard cartesian ones.
Bending moment diagrams are drawn on the tension side of the beam axis (the Anglo-Saxon convention
is opposite).
Joints in trusses are labeled with numbers, the members with number couples of the joints they connect.
In order to simplify the expressions, members are indexed with a single variable i or j in expressions,
specifically in sums which run over all members. For this variable, the number couples are substituted
when it comes to actual evaluations.
The default units are kN and m. Default units are used extensively. It saves typing and prepares the
reader for the use of computer programs, since rational programs do not use specific units neither in the
input nor in the output (naturally it is absolutely necessary to use consistent units in computer input data).

2 Statical indeterminacy
The distinguishing features of the statically indeterminate structures can be summarized:
1. Number of the external and internal constraints imposed on the structure members is greater than the
number of degrees of freedom (DOF). Consequently, the number of unknown reaction components
in the equilibrium conditions is greater than the number of equations and the reactions cannot be
uniquely determined from these equations alone. This feature is termed redundance of constraints or
simply redundance.

2. Deformation of the members need be taken into account in order to determine reactions and internal
forces.

3. Thermal effects and support settlements induce forces in statically indeterminate structures (recall that
these effects entail just displacements in statically determinate structures). This is clearly a drawback.

4. Redundance has a favorable consequence. If a support fails to bear the reaction or an internal force
is beyond some limit value (a limit bending moment at a cross-section for instance), the structure
need not necessarily fail at the load level since the force can be redistributed to other constraints and
cross-sections.

3
This list includes just properties relevant in the structural analysis. There are also differences in the design,
economy, erection and maintainance of statically indeterminate structures as compared to the determinate
ones. Most of these aspects speak for the statical indeterminacy. For instance, internal hinges are nearly
indispensable in more complex structures if statical determinacy is to be achieved. Design of internal hinges
is difficult in reinforced concrete structures and the details are prone to defects in service. These aspects
dominate in the selection of the bearing systems so that larger structures are nearly always designed as
statically indeterminate.
For illustration of these issues, a demonstration example is presented. Assume the task is to span two
clearings between three supports. The structure may be a part of a roof or bridge. Two statically determinate
simple beams do the job as shown in Sketch 1.
roadway or roof insulation layers roadway or roof insulation layers

static scheme: static scheme:

w
w

Sketch 1 Left: two simple beams spanning two bays of a roof or bridge. Right: continuous beam with the
same function.

The roadway or the continuous layers of the roof insulation suffer great local extensions above the center
support when deflections occur. The fluctuating live loads (climatic or traffic) entail repeated opening and
closing of the gap between the beams which is a notorious source of defects.
A statically indeterminate continuous beam alleviates the problem entirely. Besides, deflections w and
the extremum bending moments are smaller in the continuous beam. On the other hand, the temperature
difference of the upper and bottom faces of the beams induces bending moments in the continuous beam
whereas no internal forces arise in the simple beams.
The above discussion is valid for all kinds of structures. The scope of these notes is limited to planar
frames and trusses. Basic ideas, principles and methods explained herein can be relatively easy extended to
other types of structures.

2.1 Illustration of the two basic methods


An ideal straight vertical rod is loaded and supported in the vertical direction as shown in Sketch 2. The
example is rather academic, exclusively vertical forces and displacements occur.

4
Force method Displacement method

l1,EA1 N1

F F F
N2 v
a N1 a

l2,EA2 N2

Sketch 2 Axially loaded and supported rod. Forces act in the rod
axis, for visibility they are shown aside the rod

Forces and deflections are positive upwards, normal force is positive when it is tension.

2.1.1 Force method


1. The rod is split at section a which renders the structure statically determinate. In other words, the
vertical constraint is released at section a. The particular location of a in Sketch 2 is a matter of con-
venience (a bit simpler evaluation of the deflections) but any other location would make no difference
in principle. Since the constraint makes the difference between statical determinacy and indetermi-
nacy it is called the redundant constraint and the associated force the redundant force. Two forces
N1 and N2 are introduced at the respective ends such that the equilibrium at that section is preserved,
that is
N1 − N2 − F = 0
The normal forces in both parts of the rod also are in equilibrium with the respective end forces. The
stress state is thus in equilibrium. There are apparently a set of such equilibrium stress states which
can be parameterized with a single parameter, either N1 or N2 . Force N1 is selected here, N2 follows
from the above equilibrium condition.

2. Deflections v1 and v2 are computed of the ends of both parts using elasticity equations
l1 l2
v1 = −N1 v1 = N2
EA1 EA2

3. Compatibility of the deflections is requested v1 = v2 . The


compatibility equation reads after substitutions
l1 l2 l2
−N1 = N2 = (N1 − F )
EA1 EA2 EA2
Its solution with respect to N1 gives
1
N1 = F l1 EA2
1+ l2 EA1

The key equation is the deformations compatibility (consistency) equation and the unknown N1 is the
primary unknown of the problem. With N1 known, all other unknown forces (N2 in this simple case) can be

5
computed from the equilibrium equation(s). If necessary, displacements can be obtained from the elasticity
relations. It is worth mentioning that the selection of the constraint to be released is quite free and several
options are available even in this simple example. It could have been a cut at any other section or the top or
bottom reaction at the respective support.

2.1.2 Displacement method


1. Displacement v is imposed to the structure at section a, as shown in the rightmost diagram of Sketch 2.

2. The forces in both parts are expressed in terms of v with the aid of elasticity relations
v v
N1 = εEA1 = − EA1 N2 = EA2
l1 l2

3. Equilibrium is requested at section a

N2 − N1 + F = 0

which yields after substitutions


1
v = −F EA1 EA2
l1
+ l2

The basic equation in this method is the equilibrium equation and the primary unknown is displacement v.
The remaining unknown variables (N1 and N2 in this case) are computed from elasticity equations.
The reader is encouraged to check on his own that the solutions obtained by both methods are identical.
It is sufficient to evaluate the secondary unknowns for this purpose.
The displacement method is algorithmically simpler but offers less flexibility. Section a could have
been selected somewhere else but not entirely arbitrarily (not at the ends). Besides, the expressions for the
end forces N1 and N2 would be complicated in that case.

3 Force method
More descriptive names are also used for this method, the deflection superposition method or the method
of consistent displacements. Basic steps and terminology are introduced first on simple structures with a
single redundant or, in other words, on structures statically indeterminate to the first degree.

3.1 Steps of the force method


An illustrative example is used to introduce the common nomenclature. A simple continuous beam with
two bays is considered. The beam is loaded by two forces.

F1 F2
The structure is statically indeterminate to the first degree. There are
A B C numerous options for the constraint to be removed and the one selected
below is not necessarily the most advantageous.
Sketch 3 The task to be solved

Step 1 - select the primary structure

6
F1 F2
A constraint is selected and removed, here it is the support B. The struc-
A C
ture becomes statically determinate. This modified structure is called the
X
primary structure. The removed constraint is called the redundant con-
straint or redundant. The term redundant also is used for the force acting
Sketch 4 Primary structure and
in the actual structure in the constraint.
redundant

It is yet unknown and symbol X is thus introduced for it. In this actual example the symbol B could be
used as well but in order to have the notation sufficiently general, X is preferred.
Step 2 - deflection ∆o of the primary structure is computed in the direction of the redundant due to
the actual loading.

F1 F2

∆o The principle of virtual forces (PVf) is almost exclusively used for the
A C
purpose for obvious reasons but potentially any another method could
be applied. Note that positive senses of X and ∆o coincide. It is not
Sketch 5 Deflection ∆o , note
an absolute necessity but highly recommendable since it helps to keep
that positive sense agrees with
consistent signs.
the positive sense if the redun-
dant.

Step 3 - deflection δ of the primary structure is computed due to the unit redundant force X = 1 in
the direction of the redundant. The term compliance is also used for δ.
δ
When PVf is used, the virtual and actual load cases to be considered in
A 1 C this computation apparently are the same and the same virtual load case
is used in step 2, either. The total of only two load cases is to be solved
Sketch 6 Deflection δ, note the for the primary structure in this task. It is more apparent in the next
positive sense again. commented example.

Step 4 - superimpose the two deflections of the primary structure


The deflection due to the actual redundant force is X δ so that the total deflection is obtained by superposi-
tion:
∆ = ∆o + X δ (1)
In this example the total deflection must vanish owing to the support B. Condition
∆ = ∆o + X δ = 0
delivers the unknown redundant. The superposition is applicable to any quantity (internal force, reaction,
deflection) of the two load cases of the primary structure that have been solved in steps 2 and 3.
Step 5 - compute the final internal forces and/or deflections
There are two basic alternatives in this step.
1. Another load case, the final load case, is solved in the primary structure. It is loaded by the given
external load plus the redundant (now known) and solved as any other statically determinate structure
2. Superposition is used for the required quantity, particularly for the internal forces. If the value of
quantity Q (say an ordinate of an internal force) computed in the actual load case in step 2 is denoted
by subscript o and the value obtained in the virtual load case by subscript 1 then the final value of Q
is
Q = Qo + X Q1

7
Execution of this step is shown in demonstration examples.

3.2 Single redundant structures


The algorithm specified in the previous section is applied in several demonstration examples. In these
examples, the thermal effects and support settlement are also treated. Recall that in statically indeterminate
structures these effects induce forces. The effect of the shear force on deflections is neglected.
The principle of virtual forces is exclusively utilized for the computation of deflections forthwith. In the
wake of it, a notation needs to be introduced to distinguish the load cases that occur in the PVf applications.
All load cases refer to primary structure.

• Subscript o indicates the load case actual load

• Subscript 1 indicates the load case unit force in place and direction of the redundant. This load case
plays two parts:

– the virtual load case in PVf (it acts twice in this part in steps 2 and 3), in this role it also can be
distinguished by δ instead of the subcript 1 ,
– the real load case, unit redundant force X = 1

3.2.1 Force load


The algorithm is applied to the structure in Sketch 7. The quantities in step 2, load case 0 (actual load on
the primary structure) get index o in order to distinguish them from the values in other load cases and from
the final values in particular. Steps 1, 2 and 3 are self-explaining in Sketch 7. It remains to complete the
steps by evaluations of ∆o and δ:
!
Z l Mo 1 1·F ·l·l·l 1·F ·l·l·l 5 F l3
∆o = − δM dx = − + =−
0 EJ EJ 2 · 2 · 2 · 2· 3·2·2·2 48 EJ
Z l M1 1 1 l3
δ= δM dx = ·l·1·l·1·l =
0 EJ EJ 3 3EJ

8
STEP 1: PRIMARY STRUCTURE:
F

A l/2 C B Step 4:
l/2 X
The redundant follows from the com-
STEP 2: L.C.1 − virtual patibility equation:
L.C.0 − actual load δM:
Fl/2 M: 5
l/2
δ1 ∆o + X δ = 0 X= F =B
16
l Step 5:
L.C. 1 − real L.C. 0 and 1 are superimposed, the sec-
STEP 3: ond one with factor X. For the bending
M1:
The virtual L.C. is the same moment at the clamped end the super-
1 as in STEP 1.
position yields
l l 5 3
MA = −F + F l=− F l
FINAL RESULTS: 2 16 16
M diagram by superposition: The final load case in the primary structure:
and at the center section
3/16 Fl F
5 l 5
Mc = 0 + F = Fl
5/16 F 16 2 32
5/32 Fl

Sketch 7 Loading by forces, demonstration example

Note that the reaction and internal forces do not depend on the material stiffness (Young modulus) and
cross-section dimensions as far as those are constant in the whole beam. The displacements do, of course.
For later reference, reasonable dimensions, material properties and loading force are introduced:
1 l = 10 m
2 the cross-section 0.5 m (depth) × 0.2 m (width), J =0.00808 m4
3 material concrete, E = 30GPa
4 the loading force F = 100kN.
5
Then X = 16 F = 31.2 kN.

3.2.2 Thermal effect – temperature gradient


The same structure is analyzed as in the previous section. The structure is specific in the sense that a
homogeneous temperature change does not induce any reactions and internal forces. Indeed, this thermal
effect implies simple extension of the beam which is not restricted by supports. This condition is rather
exceptional in practice. If the roller had not vertical axis, for instance, the homogeneous temperature would
induce reaction. Another thermal effect is equally frequent in practice – uneven temperature change when
temperature varies across the depth of the beam. This happens with roof girders and bridge decks for
instance, when the upper face is exposed to sunrays whereas the lower is not. It is usually sufficient to
assume a linear variation of the temperature across the depth. For brevity, this thermal effect is called the
temperature gradient forthwith (although it is not quite correct) and is defined by the temperature difference

∆t = tupperf ace − tlowerf ace

9
If the beam is free to deform, in other words, if it is a part of a statically determinate structure, then every
cross-section experiences curvature change κt in the wake of a temperature gradient
∆t α
κt = − (2)
h
where α is the thermal extension factor of the material and h is the cross-section depth. The sign complies
with the commonly adopted conventions, namely, the curvature is positive when the deflection line is convex
downwards. Genuine temperature gradient obviously is ∆t/h.
The steps of the force method remain unaltered in principle. The primary structure, L.C. 1 (the virtual
state of step 2) and compliance δ can be reused from the previous section. In step 2, however, ∆o is the
effect of the temperature gradient:
Z l ∆tα 1 ∆t α l
∆o = κt δM dx = − 1·l =−
0 h 2 2h
The rest of the algorithm remains unaltered. Temperature gradient 20o C is considered and α = 10−5 is
assumed.
∆o ∆t α l2 3EJ 20 · 10−5 · 3 · 30 · 106 · 0.00208
X=− = = = 3.74 kN
δ 2 h l3 2 · 0.5 · 10
M: V: The final internal forces and other quantities sim-
ply are X multiples of the quantities of the L.C.1
in Sketch 7. The comparison of the force and ther-
−3.74
37.4 mal effects is quite instructive from the practical
point of view. With realistic dimensions and ma-
Sketch 8 Loading by temperature gradient, internal
terial properties a hint is provided on the relative
forces
importance of these effects.

3.2.3 Support settlement


The section title adopts the conventional and most frequent name for the phenomenon which, however, is
more generally described as an imposed displacement.
PRIMARY STRUCTURE:
The beam from the previous example is used once
more. Settlement s of support B is imposed as
s

A B X
shown in Sketch 9. The solution is simple when
the redundant is selected in the direction of the im-
FINAL RESULTS: posed displacement. In that case the imposed dis-
18.7 M: V: 1.87 placement ∆s must equal the elastic displacement
∆ which implies

∆ = ∆o + X δ = ∆s (3)
Sketch 9 Loading by a support settlement, demonstra-
tion example

When there is no other load then ∆o = 0. The same positive sense of both ∆ variables is naturally assumed.
In the present example ∆s = −s. The minus sign occurs here because the settlement by tradition is pos-
itive downwards whereas all other displacements and forces (δ, ∆, ∆o and X) are positive upwards. For
settlement s = 0.01 m, the redundant is
s 3 EJ 0.01 · 3 · 30 · 106 · 0.00208
X=− = − = −1.87 kN = B
l3 103
10
For various reasons some other redundant might be selected, for instance the fixed end moment. Such
primary structure is adopted in section 9.1.1, Sketch 83. The displacement ∆s of the primary structure
need then be computed in the direction of the redundant. This displacement is said to be conjugate to X.
Conjugate quantities generally are such that their product is the work done by the force quantity (here the
end moment) upon the displacement quantity (here the rotation at the support). In this case, however, dis-
placement ∆s must be compensated by the elastic deformation ∆ (compare with equation (3)):
X
−∆ s s ∆ = ∆o + X δ = −∆s (4)
Sketch 10 shows that ∆s = −s/l in this case. The relation ∆s : s is
Sketch 10 Loading by a elementary in this particular example. In more complicated cases the relation
support settlement, other can be resolved with the aid of virtual displacements rules [1] for rigid frames
redundant and trusses.
The rest of the solution is standard, with regard to equation (4):

s s s 3EJ 0.01 · 3 · 30 · 106 · 0.00208


X δ = −∆s = X= = = = 18.7 kNm
l lδ l2 102

3.2.4 Continuous beam


The steps of the force method are travelled once more with the most frequent statically indeterminate
structure – the continuous beam. Note that the stiffness is different in the two bays. Besides acquiring
skill, the value of this example is in the comparison of the labour spent in solutions with different primary
structures. In the first approach, the central support is selected for redundant (most students would do so
intuitively). The evaluation of ∆o in step 2 turns out quite labourious
      
1 15 2 5 1 5 1 1 1 15
∆o = − F · 5 · 10 + F 2.5 + 2.5 + 2.5 + 2.5 + F · 2.5
EJ 3 2 EJ 2 2 4 2 6 3 4

EJ1 EJ2 = EJ1 /2 PRIMARY STRUCTURE:


F 187F
d ∆o = −
A
10
B
5 5
C
X EJ
  
1 2 1 250
δ M: 5 δ= + 5 · 5 10 =
Mo: 2.5 EJ EJ 3 EJ
2.5
∆o
5/2 F X=B=− = 0.75 F
5/4 F δ
15/4 F
The first (equilibrium) approach is taken in step 5
M: 1.25 F V: 0.625 F
of the solution course according to section 3.1. The
remaining reactions are computed first
−0.125 F 0.375 F
1.87 F
whole beam C⊙ :
−A · 20 − B · 10 + F · 5 = 0 A = −0.125 F
Sketch 11 Continuous beam, central support redun- part Ad d⊙ :
dant Md + A · 15 + B · 5 = 0 Md = 1.875 F

part dC d⊙ : − Md + C · 5 = 0 C = 0.375 F Check: whole beam ↑: A + B + C − F = 0


The internal forces diagrams are elementary with all reactions known.
The same continuous beam is solved with another primary structure shown in Sketch 12.

11
EJ1 EJ2 = EJ1 /2 PRIMARY STRUCTURE:
F X    
2 1 1 1
A B C ∆o = 2.5 0.5 + 0.5 + 2.5 · 0.5
10 5 5 EJ 2 6 3
12.5 F
Mo: δ M: ∆o =
EJ
0.5 10
1 δ=
2.5 F EJ
X = −1.25 F
Sketch 12 Continuous beam, bending moment at B re-
dundant

Needles to say, the final diagrams must be the same as in Sketch 11.
Note that the evaluations are simpler with the latter primary structure. The selection of the redundant
does not affect the results but it may affect the computational demand.
Temperature gradient effect could be solved for in analogy to section 3.2.2 and is not repeated here.
In order to complete the example, settlement s of the central support is considered with the latter primary
structure. The displacement conjugate to X is the relative rotation of the left and right parts of the beam in
the primary structure diagram in Sketch 12. Then
s
∆s = −2
10
where s is positive downwards. According to equation (4)
∆s 2s s EJ
X=− = =
δ δ · 10 50

3.3 Normal forces effect


In all previous examples, normal forces were absent. Their effect on deflections can very often be neglected,
see [1], even if they are present and, consequently, they have then a negligible effect on redundants in stat-
ically indeterminate structures. The steps of the force method in section 3.1 remain the same when the
normal forces are taken into account.

12
F PRIMARY STRUCTURE:
E, J, A

0.4
0.3
AS X

3
5
L.C. 1: Difference occurs in the evaluation of ∆o and ∆.
4 4
M1( δM) It is demonstrated on the example of a cantilever
2.4
L.C. 0: with a skew strut in Sketch 13. Realistic dimen-
8F Mo:
4F
sions are adopted in order to provide an illustration
on the impact of normal forces neglect in practice.
N1( δN)
Cross-section characteristics:
No void everywhere 0.8 J = 1.6 · 10−3 A = 0.12
.0
−1 Z Z
Mo No
∆o = δM dx + δN dx =
M: N: EJ EJ
4F  
3.3 F 1 1 1 32 F
2·4·4F + 2·4·4F =
1.87 F .12
F EJ 2 3 EJ
−4

Sketch 13 Cantilever with strut


Z Z
M1 N1 1 1 1 1 7.7 2.56 5
δ= δM dx δN dx = 2 · 42 · 4 + 0.82 · 4 + 12 · 5 = + +
EJ EJ EJ 3 EA EAs EJ EA EAs
For simplicity the same cross-section area is assumed of the strut, As = A, although it hardly would be true
in practice. Then
moment normal f orce contributions
7.7 7.56 1 z }| { z}|{
δ= + = ( 4800 + 63 )
EJ EA E
∆o
X=− = −4.12 F
δ
and the final internal forces are shown in Sketch 13.

The effect of normal forces is nearly negligible and the strut could have been
considered rigid with nearly the same internal forces and displacements. It
would mean the solution of the substitute structure in Sketch 14. In practice,
Sketch 14 The substitute
the effect of normal forces very often is neglected.
structure

3.4 Thermal effect – homogeneous temperature change


The cantilever from section 3.3 can be used to demonstrate the effect of a homogeneous (uniform) temper-
ature change. Recall that such change induced no forces in the beams dealt with in sections 3.2.2 and 3.2.4.
In the strict sense, homogeneous means the same in the whole structure. In a narrower sense, it means the
same just in a part of the structure, for example in a single beam. In order to distinguish this phenomenon
from the temperature gradient ∆t, the homogeneous temperature change assumes symbol t. The tempera-
ture increase t = 10o C takes place in the strut in the present example, see Sketch 15.

13
PRIMARY STRUCTURE:
E, J, A

0.4
0.3
AS o C X
The same primary structure is adopted as in sec-

3
1 0
t= L.C. 1: tion 3.3.
4 5 4 M1( δM) Z
2.4
L.C. 0: ∆o = εt δN dx = 10−4 · 1 · 5 = 5 · 10−4
Mt and κ t void everywhere
| {z }
must not be neglected

εt N1( δN)
Z Z
M: 0.8 M1 N1
.0
δ= δM dx + δN dx
7.4
−1 EJ | EJ{z }
N: can be neglected
2.47

.09
−3 moment normal f orce
7.7 7.56 1 z }| { z}|{
= + = ( 4800 + 63 )
EJ EA E
Sketch 15 The strut is subject to homogeneous tem-
perature increase

5 · 10−4 E 5 · 10−4 · 3 · 107


=− X=− = −3.09
4863 4863
It is important to note that, exclusively, the terms with the actual, not virtual normal forces, can be
neglected. The former terms represent elastic axial deformation whereas the latter terms represent thermal
axial extensions/contraction.

3.5 Single redundants in trusses


The paradigm of the force method remains valid without exceptions. Several specific features are pointed
out.
1. The virtual work of internal forces is simpler since just axial forces enter it. It is further assumed that
the axial forces are constant in each member.
2. Hinges are not shown by specific graphic symbols in diagrams. Each joint or vertex in the diagrams
is a hinge.
3. Temperature gradient in individual members induces no forces.
4. Notation rules:
(a) Basic numbering refers to joints
(b) Members are labeled by the pairs of adjacent joints numbers in numerical examples
(c) Members in formulas get simple subscripts for brevity. These simple subscripts are resolved to
the respective pairs of the joint numbers when the formulas are applied in examples
The formulas for deflections are resumed for convenience:
X X No,i X X N1,i
∆o = εo,i δNi li = δNi li , δ= ε1,i δNi li = δNi li (5)
all mem all mem EAi all mem all mem EAi

Needless to say, δNi = N1,i . The two symbols are retained in equations (5) in order to remind of the origin
of the respective terms.

14
3.5.1 Simple truss, force load
The same material is assumed in all members of the truss in Sketch 16 with E = 200 GPa. The cross-section
areas are A = 0.05 m2 in members (1, 2), (2, 3) and (2, 4) and A = 1.25 · 0.05 = 0.0625 m2 in members
(1, 4) and (3, 4). In order to simplify arithmetics, 106 times less E is used in the table. True deflections ∆o
and δ thus are 106 times less than those in the table.
F=120 PRIMARY STRUCTURE RESULTS:
4
No: F=120 δN:

−1
−1
.66
−2
160

89

11
.66
40

17
2.66
−1
00
0
1 2 3 0 1 −53.5 −53.5
δ1 δ1
30 30

Sketch 16 Simple truss

Evaluations are summarized in a table:


member l AE l/AE No δN ∆o δ N ∆l
1,2 30 10 3 0 1 3 -53.5 -161
2,3 30 10 3 0 1 3 -53.5 -161
1,4 50 12.5 4 0 -1.66 11.1 89.0 356
3,4 50 12.5 4 -200 -1.66 1330 11.1 -111 -445
2,4 40 10 4 160 2.66 1700 28.4 17.0 68
sum 3030 56.6
Factor 10−6 cancels out when X is computed:
3030
X=− = −53.5 kN Ni = No,i + X Ni
56.6
Note that the work done in the split member 1, 2 must be included in δ! Member 1, 2 is not removed but
merely cut when the primary structure is selected. Its deformation must thus be accounted for though it is
zero in the actual load case. It is also worth noticing that the internal forces do not depend on E in this force
load case.

3.5.2 Simple truss, thermal effect, compatibility check


Assume member 1, 4 is warmed up by 20o C with α = 0.00001. Deflection ∆o then is easy to compute:

∆o = εo,1,4 δN1,4 l1,4 = α t1,4 δN1,4 l1,4 = 0.00001 · 20 · (−1.66) · 50 = −0.0166

This ∆o is the true value. When X is evaluated the deflection δ due to the unit force must also be true so
that the value from the table must be multiplied by 10−6 , see the previous section.
∆o 0.0166
X=− · 106 = · 106 = 293 kN
δ 56.6
The remaining member forces equal X-times the virtual forces column δN in the above table.
It is instructive to use the Williot-Mohr diagram for the compatibility check of the solution. The mem-
ber extensions/contractions are computed first. The length change in member 1, 4 consists of the thermal
extension and the elastic contraction:
l1,4
δl1,4 = α t lt,1,4 + N1,4 = 0.01 − 293 · 1.66 · 4 · 10−6 = 0.00804 m
EA1,4

15
4 4
SCALE:
1 mm THERMAL LOAD

3.15
In all other members it is the elastic de-
4 formation alone,
4
1 2 3
∆li = X δNi (l/EA)i

1.9
6

4
summarized in the table below. The

8.0
1
1 2 3
member elongations are given in mm in
Sketch 17 for the thermal load. The
scale of the diagram is irrelevant. The
4
diagram for the force load case is also
−161 68 4
shown in Sketch 17. The unit of dis-
3 2 −161 1
placements here is µm. The length
SCALE: 1 changes are evaluated in the table in sec-
6
FORCE LOAD
35

10 −4m tion 3.5.1, the last column. The con-


−4

4
45

struction of both diagrams follows the


same steps. The verbal description can
be traced in both diagrams.

Sketch 17 Simple truss, check by Williot-Mohr diagrams

The drawing starts at joint 3 which does not move. Displacements of joints 2 and 1 are easy to draw since
their displacements are horizontal. Then the displacement vector of member 4 is constructed from relative
displacements of that joint with respect to joints 2 and 3.
member δNi (l/EA)i ∆l
1,2 1 3 0.00089
2,3 1 3 0.00089
2,4 2.66 4 0.00315
3,4 -1.66 4 -0.00196
Projections of these relative displacements on the respective member axis are drawn with dashed arrows.
These two projections determine the displacement vector of joint 4 as the intersection of lines indicated by

4 . Joint 1 displacement is constructed last as relative to joints 2 and 4. The locus of all admissible positions
of point 1 again is marked 1 . The locus must go through the point 1 which is already known. This is the
graphic check.

3.5.3 Checks on the force method solutions


Potential checks are scarce. All load cases in the primary structure should be carefully checked for equi-
librium. Then there is little point in checking equilibrium in the final result. Unfortunately, there is no
independent check on the redundant X. The final result can be in perfect equilibrium despite the wrong X.
Totally independent checks do not exist except re-solution with another primary structure and some checks
on displacements like the Williot-Mohr diagram. A numerical version of the diagram can be devised, too,
but it is prohibitively labourious and seldom done in practice.

16
3.5.4 Sample frame solution
Two load cases are considered for the frame in Sketch 18, a concentrated force indicated in Sketch (LC1)
and a homogeneous temperature increase ∆t = 20o C (LC2). The normal forces effect (elastic axial exten-
sion/contraction) is neglected. The same primary structure is selected for both actual load cases. In order to
simplify expressions, the ratio J2 /J1 = a = 5 is utilized. The basic moment of inertia is J1 = 6.22 · 10−4 m4
but it is not actually substituted.

2.5
F=1 2.5 2.5
0.5

2.5
J2 0.3
LC 1
0.3

PRIMARY STRUCTURE
5

J1 FINAL M:
J1
0.292 X

5 5

5 5 5 1
5 5

M: δ M: δ N:

Sketch 18 Frame 1 geometry, primary structure, bending moment diagrams for actual load case LC1 and
the virtual stress state.

For load case LC1, the virtual work integrations deliver:


 
1 1 1 1 400 1
∆o = (5 · 5 · 5) + (5 · 5 · 10) =
3 2 a EJ1 6 EJ1
 
1 1 1 1 800 1
δ= (5 · 5 · 5) + (5 · 5 · 10) + 53 =
3 a 3 EJ1 6 EJ1
X = −0.5
The final bending moment diagram obtained by superposition is in the rigthmost top diagram of Sketch 18.
For load case LC 2, the normal forces in the virtual state must also be known. They are shown in the
rightmost bottom diagram in Sketch 18. The virtual work need be integrated just along the horizontal girder

∆o = ∆t α 10 · 1

∆o ∆t α 10EJ1 6 20 · 10−5 · 6.22 · 10−4 · 6


X=− =− =− = −1.5 · 10−6 EJ1
δ 800 800
The bending moment diagram for LC 2 is the diagram of the virtual stress state multiplied by X and is not
shown.
The solution of this example can be simplified when symmetry is utilized.

3.5.5 Sample truss solution


The truss in Sketch 19 is solved for two load cases indicated in the top left and right bottom diagrams.
The primary structure is defined in the top center diagram. The same axial stiffness AE is assumed in all
members.

17
LC 1 − FORCE EFFECT F=1 PRIMARY STRUCTURE 1.41 0.531
4 5 6

0.
67

67
X

0.

−0.469
51
1

0.

.7
75

−0
1
1 2 3
−1.531 −0.469
1 1
FINAL MEMBER FORCES

2 No: 0
F=1
0.707 δN : −0.707 LC 2 − THERMAL EFFECT v=?
1

−1

−0.707
1
−1
1

−1
.4

0
−1

−1 −1 0.707 −0.707 ∆t ∆t

Sketch 19 Truss 1 geometry, primary structure, member forces diagrams for actual load case LC1 and the
virtual stress state.

Calculations are summarized in Table 1 which is self-explaining to some extent. In columns 1-6 the data is
gathered for the evaluation of the virtual work for LC 1 – the force effect. The data is later reused in further
calculations.
Table 1 Virtual work evaluation for truss 1. Common multiplier 1/(EA) is omitted in all work columns.
column 2 3 45 6 7 8 9
2
member l No δN
lδN No lδN N ∆l lNo2
1,2 1 -1 0.707
-0.707 0.5 -1.53 -1.53 1
2,3 1 -1 0.707
-0.707 0.5 -0.47 -0.47 1
1,5 1.41 -1.41 -11 1.41 0.67 0.95 2.82
2,4 1.41 0 -10 1.41 0.751 1.06 0
2,6 1.41 0 10 1.41 -0.751 -1.06 0
3,5 1.41 1.41 12 1.41 0.67 0.95 2.82
4,5 1 2 0.707
1.41 0.5 1.47 1.47 4.0
5,6 1 0 -0.707
0 0.5 0.53 0.53 0
3,6 1 -1 -0.707
0.707 0.5 -0.47 -0.47 1
∆o =6.121 δ= 8.15 12.64
The redundant can be computed (the axial cross-section stiffness EA cancels out)
∆o
X=− = −0.75
δ
and then the final member forces can be evaluated by superposition,

Ni = No,i + XδNi ,

in column 7 of the table. These forces Ni are then inserted in the top right diagram in Sketch 19 for
convenience and easy check up. Columns 8 and 9 of the table are useful further on when deflection are
computed in the statically indeterminiate truss.
In LC 2, the thermal expansion in members 1,2 and 2,3, the actual stress state consists of these expan-
sions exclusively X
∆o = εt δNi li = tα(0.707 − 0.707) · 1 = 0 X=0
i

18
where t and α are temperature increase and thermal expansion factor, respectively. Surprisingly, this specific
thermal effect does not induce any internal forces at all. Mind that it is pure coincidence, however. In order
to demonstrate this, LC 2 is modified, the temperature effect afflicts now just member 1,2. The modified
load case is denoted LC 2b.
0.707
∆o = tα0.707 · 1 X = −EAtα = −0.087EAtα
8.15
Member forces are simple X multiples of column 4 in table 1 (or the forces in the bottom center diagram
in Sketch 19) and are thus not explicitly given for this load case.

3.6 Deflections in statically indeterminate structures


Recall that in the PVf no discrimination appears in regard to statical determinacy. Deflections in statically
indeterminate structures can thus be computed using the same algorithm which is applied to statically deter-
minate structures. The actual, statically indeterminate, state provides the deformations and displacements
on which the internal and external forces of the virtual state do work. The virtual state, however, is not
unique. Recall the conditions put on the virtual state:

• It must be an equilibrium stress state

• The single external (except reactions) force acting in the virtual state is a unit force in place and
direction of the desired displacement.

There is no condition on the compatibility of displacements in particular. Consequently, any state in the
primary structure matching the above conditions will do. The algorithm is best demonstrated in a simple
example.

3.6.1 Deflection in the sample frame


Let us assume the task is to determine the rotation φ indicated in Sketch 20 left.
F δ1 δ1
φ =? 1

δ M:

Sketch 20 Frame 1, required deflection, two eligible virtual stress states

The bending moment diagram of the actual load case is the final bending moment diagram, the rightmost
top diagram in Sketch 18. The PVf leaves great freedom in the selection of the virtual state. The obvious
option is the virtual unit force acting in the original statically indeterminate structure. In the sample frame
it would be the load case shown in Sketch 20 in the center. Such load case still needs to be solved, which
requires additional work. Another virtual stress state is therefore selected. Necessary conditions are met
by any properly solved load case in the statically determinate primary structure, in particular, the load case
indicated in the right diagram in Sketch 20 is an eligible virtual stress state. Its external forces do not do
any work on the displacements of the actual state except the unit moment. It is obviously easier to solve
for the bending moment diagram in this load case than in the center load case in Sketch 20. The result is

19
shown in the Sketch. The integration of the product of this bending moment with the moment in the actual
load case (see Sketch 18, top right diagram) yields
 
1 1 1 5
φδ1 = − · 1 · 5 · 10 + · 1 · 2.5 · 10 =−
3 2 EJ2 6EJ1
The algorithm is next demonstrated on the sample truss from Sketch 19.

3.6.2 Deflections in the sample truss


The vertical deflection v at joint 6 is computed, see the bottom right diagram in Sketch 19. The deflection
is computed first for the actual load case LC 1, point force at joint 6, shown in the top left diagram in
the Sketch. There are many potential virtual stress states that could be used for this purpose as already
indicated in the previous section. Symbol δ N̄i is used for the stress state actually selected from this set. A
set of eligible stress states are obtained when various primary structures are loaded by the virtual unit force
in the place and direction of v. Note that the final stress state Ni in the top right diagram also belongs to the
eligible stress states so that δ N̄i = Ni would be an admissible selection. Among the stress states already
solved and collected in Sketch 19, state No,i in the left bottom diagram also is eligible. Calculations are
simpler with the latter selection and therefore δ N̄i = No,i is adopted.
The actual stress state in the PVf is unique, it is the final stress state with member forces Ni . Neverthe-
less, they can be decomposed back in the two stress states they had been initially superimposed from:
X P X
v · δ1 = Ni δ N̄i li = i Ni No,i li = (No,i + XδNi )No,i li =
i i

P 2 P
= i No,i li + X i δNi No,i li = 12.64 − 0.75 · 6.121 = 8.06
The framed expressions indicate two possibilities the work can be evaluated. The latter one is actually used
since it is easier to calculate. The second sum in the frame, namely, is ready in Table 1 in column 5 and the
first sum is a bit simpler than the sum in the first frame. The first sum is performed in column 9 of Table 1.
The same deflection is computed for the thermal effects dealt with in section 3.5.5. First, the thermal
expansion is considered in both members of the lower chord. The actual load case consists exclusively
of the thermal extensions in these two members since there are no member forces whatsoever. The same
virtual stress state δ N̄i = No,i is adopted as above.

v · δ1 = ∆tα · 1(−1 − 1) = −2∆tα

It is instructive to try for once the other option for the virtual stress state δ N̄i = Ni

v · δ1 = ∆tα · 1(−1.531 − 0.469) = −2∆tα

with the same result as expected.


The other thermal effect is the extension in member 1-2 only, the load case LC 2b in section 3.5.5. Using
the same virtual stress state once again the deflection becomes
X X X XδNi
v= εi δ N̄i li = (εelast
i + εti )No,i li = No,i li + ∆tα(−1 · 1) =
i i i EA

X X
= δNi No,i li − ∆tα = −0.087∆tα · 6.121 − 1) = −1.53∆tα
EA i

20
3.7 Multiple redundants
Basic paradigm and steps of the force method remain valid for structures with multiple redundants. An
extension of notation rules is necessary, nevertheless:

n number of redundants

i, j indexes (subscripts) of the redundants

∆i total displacements in the primary structure in the directions of the redundants

∆0,i displacements due to external load, thermal effects and other external agents in the primary structure
in the directions of the redundants

δi,j displacements in the primary structure in the direction of the i-th redundant due to unit force in the
direction of the j-th redundant (compliances).

The compatibility (deflection superposition) equation reads


n
X
∆i = ∆0,i + δi,j Xj (6)
j=1

Deflections ∆0,i and δi,j can be computed by any method but PVf is preferred in most cases and used
exclusively herein. Then Z Z
Mo No
∆0,i = δMi dx + δNi dx (7)
EJ EA
Z Z
Mj Nj
δi,j = δMi dx + δNi dx (8)
EJ EA
Recall that Mi = δMi and Ni = δNi . The two different symbols are used to distinguish different roles of
the load cases in the PVf. The obvious consequence of these equalities is

δi,j = δj,i , (9)

an outcome which remains valid even if PVf is not used to compute the deflections. This rule is called
Maxwell’s law (1864) and a slightly more general variant of it is Betti’s law (1872, valid for more general
loading). Total displacements ∆i = 0 except when imposed displacements (support settlements) exist.

3.7.1 Clamped beam


A clamped beam in Sketch 21 provides the opportunity to demonstrate the steps and notation. Normal
forces are missing except the states including X3 .

21
f
Z l Mo 1 f l2 1
l ∆0,1 = M1 = 1
X1 X2 X3 0 EJ 3 8 EJ
∆0,2 = ∆0,1 ∆0,3 = 0
Mo
11·1
δ1,1 =
M1
1/8 fl 2
3 EJ
l
1 δ1,2 = δ2,2 = δ1,1
M2
6 EJ
All displacements in the direction of X3 vanish except δ3,3
1
δ1,3 = δ2,3 = 0
Sketch 21 Clamped beam by the
force method

By Maxwell’s law, (9), the remaining deflections are known except δ3,3

1·1
δ3,3 = l
EA
Three deflection compatibility equations result with the coefficient matrix (all equations are multiplied with
3EJ/2) and solution:

X1 X2 X3 abs. term
f l2
1 0.5 0 8
f l2
f l2 X1 = X2 = X3 = 0
0.5 1 0 8
12
0 0 3J
A
0

Several observations are worth mentioning:

1. The axial and lateral forces and displacements can be decoupled in this particular load case. This
always is the case with straight beams.

2. Owing to Maxwell’s law, the coefficient matrix of the equation system is symmetric. This feature is
general and valid for all force method solutions.

3. Diagonal elements of the structure matrix must be positive in the wake of formula (8).

4. The symmetric structure matrix δi,j collects compliances and often is called the compliance matrix.

5. The structure and loading exhibit symmetry with respect to the vertical central axis. The symmetry
could have been utilized in the solution. An immediate consequence of the symmetry is X1 = X2 , for
instance. Symmetry is a specific property, of course, but very often encountered in design practice.

22
3.7.2 Simple frame
PRIMARY STRUCTURE: X2
X1
X1
F
X2

4
1
∆o,1 = − 4F · 4 · 4 = −21.3 F
3
8
8
1
∆o,2 = 4 F · 8 · 4 = 64 F
2
1
Mo: M1: M2: δ1,1 = 2 · 42 · 4 = 42.7
4F 4 4 8
3
FINAL INTERNAL FORCES FOR F=1: 1
0.77 δ1,2 = − 9 · 4 · 4 = −64
−0.355 −0.097 0.77
2
1
N: V: M: δ2,2 = 82 · 4 + 82 · 8 = 426
0.097 −0.097 0.645 0.355 1.81 1.42 3

Sketch 22 Simple frame by the force method

Compatibility equations:

X1 X2 r.h.s
42.7 −64 21.3F X1 = 0.355 F X2 = −0.097 F
−64 426 −64 F

3.8 Symmetry issue


Many structures exhibit symmetry with respect to an axis, mostly a vertical one. If the loading also is sym-
metric then all internal forces, deflections and reactions are symmetric, too. Symmetry can be utilized to
simplify solutions. Prior to going down to details note that any loading can be decomposed in symmetric
and antisymmetric parts on a symmetric structure. Sketch 23 illustrates the way it is done.
f f/2 f/2 f/2
f/2
F
F/2 SYMMETRIC F/2 F/2 F/2
PART
ANTISYMMETRIC
PART

Sketch 23 Symmetric and antisymmetric parts of arbitrary loading

Two load cases need be solved when the original given loading itself is neither symmetric nor antisymmet-
ric. Superposition of the symmetric and antisymmetric load cases must give back the original loading.
There are basically two techniques how the symmetry can be utilized.
Whole structure The structure is considered ’as is’ in contrast to the other technique. Symmetry/antisymmetry
of the loading implies that all quantities (forces, displacements) assume equal/opposite values in the
symmetric halves. If this technique is used in the force method, the primary structure must be sym-
metric, too, which might cause difficulties.

23
Split structure The structure is split along the symmetry axis. Just one half is actually considered. Symme-
try/antisymmetry conditions on the symmetry axis are perceived as additional constraints (supports)
on the half under consideration. The constraints are summarized below:

symmetric antisymmetric
M φ φ
M V
v loading loading
v
kinematic u = 0, φ = 0 v=0
N N u u
V static V =0 M = 0, N = 0
(10)
Sketch 24 Internal forces and displace-
ments at the symmetry axis, notation.
token

Both techniques are demonstrated in example in Sketch 25. Elastic axial deformation (normal forces effect)
is neglected in the example. The symmetric load case turns out trivial in that case. All reactions and internal
forces vanish except the normal force in the horizontal beam which equals −F/2.
The antisymmetric load case remains to be solved. The solution is carried out by both techniques for
comparison.
One primary structure option is shown in Sketch 25 for the first technique, whole structure.
PRIMARY STRUCTURE:

X1 X3 X2 Symmetry entails a bit peculiar primary structure


F X3 with released axial constraint on the symmetry
axis. This option exposes the troubles with pri-
h

mary structure selection in this technique. Of


course, other primary structures are possible which
l
look less odd, the simplest one with all three redun-
dants at the symmetry axis.
ANTISYMMETRIC LOAD CASE: Mo: M1:
1 Integrations of the virtual work need be carried out
F/2 F/2 F/2 1 just on one symmetric half. Antisymmetry implies
X1 = −X2 and X3 =0. There remains just one
independent unknown force X1
Fh/2
 
1
FINAL BENDING MOMENT: ∆o,1 =− 1F hh
Fh 3h/(6h+l) 2
l
δ1,1 = h +
6
Fh(3h+l)/(6h+l) 3F h2
X1 =
6h + l
Sketch 25 Symmetric primary structure

For later reference, the final bending moment at the fixed end support is computed:
3h + l
MA = X − F h = −F h
6h + l
The same example is solved next by the split structure technique. Kinematic conditions (10) are applied
at the symmetry axis. A single condition v = 0 appears there for the displacements of the antisymmetric
load case. This constraint is equivalent with a roller with vertical axis. Instead of the original structure, the

24
split half is analyzed as shown in Sketch 26. A primary structure is selected for this half with no further
limitations and the rest of the algorithm goes on in an entirely standard way. Redundant X is the option
selected here. !
1 l 1 l h
∆o,1 = F h 1 + F h 1 h = F h +
3 2 2 6 2
SPLIT HALF: PRIMARY STRUCTURE: Mo :
Fh 1
M 1:
F/2 1
Fh
1 h l
δ1,1 = 1 · 1 h + 1·1 =h+
h

3 2 6
X
3h + l
l/2 X = −F h
6h + l
Sketch 26 Split half of the structure

The redundant is a check on MA from the previous solution via the whole structure technique.
The example testifies that splitting the structure at the symmetry axis is more suitable in the context
of the force method. There is yet another important aspect of the issue. The first technique is hardly
applicable in computer solutions since the constraints like X1 = X2 are difficult to impose. Skill in the
structure splitting technique and in imposing the appropriate constraints on the symmetry axis, on the other
hand, is very well applicable when preparing input to computer programs. This is true even in case of
general purpose finite element codes. Utilization of symmetry is exercised throughout the rest of the notes
in examples which are suitable for it.

3.9 Examples on multiple redundants


3.9.1 Frame
The frame in Sketch 27 is statically indeterminate to the 5th degree. Normal forces effect is neglected.

Material - concrete, E = 30 GPa, α =


F J1
10−5
1

0.5 Cross-sections data: J1 = 1/24 =


J2 J2
5

0.0415, J2 = 1/8J1 . Two load cases


are considered:
5 10 10 5
1. horizontal force F acting along
0.5

the girder, see Sketch 27.


0.5
2. uniform temperature increase
20o C.
Sketch 27 Frame with 5 redundants

25
FOR ANTISYMMETRIC LOAD: FOR SYMMETRIC LOAD:

Vertical displ. Horizontal displ.


and rotation
restrained
restrained
Utilizing symmetry simplifies the task
dramatically. Constraints to be inserted
for the symmetric and antisymmetric
load cases are shown in Sketch 28
Sketch 28 Constraints for symmetric and antisymmetric load
cases drawn by dashed lines

Load case 1 - horizontal force on the girder All displacements (δ-s and ∆-s) are inversely proportional
to E in this load case. When the compatibility equations are all multiplied by E it disappears totally. With
this in mind, the Young modulus is omitted in this paragraph. The location of the force F along the girder
has no effect on the internal forces except the normal force in the girder itself. This is the consequence of
the neglect of the normal forces elastic effect. This load case is further decomposed in the symmetric and
antisymmetric parts. Both include force F/2 acting upon the left half shown in Sketch 28. Force F/2 acting
upon the right half to the right in the antisymmetric load case and in the opposite direction in the symmetric
load is not shown.
Antisymmetric part of the load:
F/2 M1 :
X1 X2 5
1 1 1
∆0,1 = 5 F/2 · 5 · 5 + 500 F/2
5
M2 : J2 2 J1
Mo :
1 1 1
∆0,2 = − 5 F/2 · 10 · 5 = − 1000 F/2
10 J2 2 J1
5/2 F 10
1 1
δ1,2 = − 5 · 10 · 5 = − 2000
Sketch 29 Primary structure and the elementary load J2 J1
cases for the antisymmetric part of the load
 
1 1 1 3 1 1 1
δ1,1 = 5·5·5+ 5 = 125 + 8 = 1040
J1 3 J2 J1 3 J1
 
1 1 3 1 1 1 1
δ2,2 = 10 + 102 · 5 = 1000 +4 = 4330
J1 3 J2 J1 3 J1
The solution of compatibility equations delivers

X1 = −0.338 F/2 X2 = 0.0743 F/2

Symmetric part of the load:


There are no reactions and internal forces in this load case except the normal force in the horizontal girder.
Complete solution:
Complete superimposed solution of load case 1 (horizontal force F is shown in Sketch 31).

26
1.69F/2
M: V:
2.43F/2

0.743F/2 −0.07F/2
−0.84F/2

1.0F/2
2.56F/2

Sketch 30 Complete superimposed solution of load case 1 – horizontal force. diagram.

Load case 2 - homogeneous temperature increase The same primary structure is selected with different
redundants on the symmetry axis, of course. Some intermediate results from the previous section can thus
be reused. Normal forces diagrams are necessary in the roles of the virtual load cases.
F/2 X3
X1 X2
εο = α ∆ t ∆0,1 = ε0 1 · 5

N1 : M1 : ∆0,2 = ε0 1 · 10
5 ∆0,2 = 0
1 N2 : 5 M2 : Compliance δ1,1 can be taken from the previous
paragraph. It must be recalled, however, that
1 Young modulus E was omitted there. In order to
simplify calculations, compatibility equations are
5
N3 : M3 : multiplied by E J1 and common multiplier εo E J1
1
is omitted from the right hand side of the compat-
N =0
ibility equations. In the end, of course, the result
1 must be multiplied back by

Sketch 31 Primary structure and elementary load 1


εo E J1 = 10−5 · 20 · 3 · 107 = 250
cases for load case 2 – homogeneous temperature in- 24
crease.

1 2 1 2
δ1,2 = 5 · 5 = 500 δ2,2 = 8 5 · 5 = 333
2 3
δ1,3 = −8 · 5 · 1 · 5 = −200 δ2,3 = −100 δ3,3 = 8 · 12 · 5 + 12 · 10 = 50
The ’trimmed’ compatibility equations

X1 X2 X3 r.h.s.
1040 500 −200 −5
500 333 −100 −10
−200 −100 50 0
yield solution (after multiplication by 250, see above):
X1 = 3.85 X2 = −21.6 X3 = −27.8
and internal forces diagrams for this load case:

27
Μ: V:
28
45 3.4

17

59 −20.8

Sketch 32 Final internal forces diagrams for load case 2 – homogeneous temperature
increase.

3.10 Multiple redundants in trusses


There is nothing principally different in the treatment of multiple redundants in trusses. A small formal
nuissance occurs when equations (7) and (8) are trimmed to fit for trusses. Integrals are replaced with sums
over members which entails double subscripts of member forces:
X
∆o,i = εo,k δNi,k lk
k=all members

X X Ni,k
δi,j = εi,k δNj,k lk = δNj,k lk
k=all members k=all members EAk

These double subscripts effectively become triple when the member subscript is further resolved to the stan-
dard pair of end joints. The demonstration example is chosen so as to expose the utilization of symmetry at
the same time. At first sight, neither the structure nor the loading exhibit perfect symmetry in Sketch 33. In
spite of it, symmetry can be used to facilitate the solution. Reaction component Ax can be computed from
the separate horizontal equilibrium equation of the whole truss.

F F
This force is perceived as
a part of the external load-
2

ing forthwith and the cor-


F
A
responding horizontal con-
B
2 2 2 straint is removed. The struc-
ture becomes symmetric al-
SYMMETRIC L.C. ANTISYMMETRIC L.C.
F/2 F/2 though instable. For the
4 5 6 given external load, however,
7 equilibrium in the horizon-
tal direction is now a pri-
F/2 F/2
1 2 3 ori guaranteed and instabil-
ity does not prevent solu-
tion, see the top right dia-
Sketch 33 Truss ’symmetrized’ and decomposition of the load in the sym- gram in Sketch 33.
metric and antisymmetric load cases

The trick is applicable to any given external load. The same cross-section is assumed in all members with
axial stiffness AE. Symmetry is exploited via the ’split’ approach – constraints are inserted on the symme-
try axis. This involves subtle issues. First, the intersection of the diagonals on the symmetry axis must be
perceived as joint 7. Second, the chord members must be cut creating new joints 3 and 6. Standard con-
straints are then inserted at the symmetry axis, see Sketch 33. The reader may miss the rotation constraints

28
in case of the symmetric load case (recall Sketch 28). Rotation constraints lose sense at a truss joint (which
effectively is a hinge) and that is why they can be omitted entirely. Note also that the symmetry condition
alone guarantees that the rotations of members 2, 3 and 5, 6 vanish. The analysis of the split halves of the
structure is carried out separately below. Note that the degree of statical indeterminacy is different for the
symmetric and antisymmetric load cases, namely 2 for the former and 1 for the latter.

3.10.1 Symmetric load case


Standard counts of the DOFs (m) and constraints (r) yield 14 in both cases. In spite of it, the structure
apparently is not statically determinate. Recall that m = r is the condition necessary but not sufficient for
statical determinacy. In this case, joints 3 and 6 are geometrically instable and, on the other hand, there is
apparently a redundant member in the left bay of the truss. Statical determinacy must be assessed by the
basic criterion - whether all internal and external forces can be determined from the equilibrium conditions.
A stepwise assembly of the structure is useful in this analysis.
1 ISD 2 ISD 3 ISI 1 4 ISI 1 5 SI 1 6 SI 2

Sketch 34 Stepwise assembly of the symmetric half of a truss. ISD - Internally Statically Deter-
minate, ISI - Internally Statically Indeterminate, SI 1 - Statically Indeterminate to the 1 degree.

Step 5 in Sketch 34 needs a closer look. Equilibrium can be reached only when there is no vertical load at
joint 3. The condition is always met since joint 3 is not taken into account when potential member loads
are assigned to joints in the original sructure. The same applies to joint 6. The assembly ends up with 2
degrees of statical indeterminacy and it suggests suitable redundants, too. A primary structure is shown in
Sketch 35. The algorithm is standard and calculations are displayed in the table. The actual load F/2 is
replaced by 1/2 for brevity. The final forces must be multiplied by F .
PRIMARY STRUCTURE No also final N:
4 l
5 6 1/2 −0.5 −0.5 1/2 joint AE No N1 N2 ∆o,1 ∆o,2
1, 2 1 .5 −.7 0 −.35 0
7
1
X

2, 3 .5 .5 0 −.5 0 −.1
X2
1, 4 1 0 −.7 0 0 0
1/2 0.5 0.5 1/2
1 1, 5 1.4 0 1 0 0 0
2 3
2, 4 1.4 0 1 0 0 0
N1: −0.707 0 N2: 0 −0.5 2, 5 1 0 −.7 −.5 0 0
2, 7 .7 0 0 .7 0 0
0.
0
1

5, 7 .7 0 0 .7 0 0
70
−0.707

−0.707

−0.5

1
7
0

0
1

4, 5 1 −.5 −.7 0 .35 0


07
0

5, 6 .5 −.5 0 −.5 0 .1
1

7
0.

−0.707 0 0 −0.5 sum 0 0

Sketch 35 Symmetric load case, primary structure, el- Numbers in the table are truncated to save space. Read
ementary load cases. 0.707 instead of .7, 1.414 instead of 1.4, 0.353 instead
of 0.35 and 0.125 instead of 0.1.
The redundants vanish, X1 = 0 and X2 = 0, regardless of the δi,j compliances (providing the two equations
are not singular, which is not the case here). Final member forces of this symmetric load are those of the
stress state No .

29
3.10.2 Antisymmetric load case

5 SI 1 6 SI 1 The assembly can be retraced up to step 4 in Sketch 34. In step 5


in Sketch 36, the structure is marked statically inderminate to the first
degree although it is apparently not geometrically stable - it can trans-
late in the horizontal direction. External load resultant has no horizontal
component, however. No horizontal constraint is thus necessary to es-
Sketch 36 Stepwise assembly tablish equilibrium. In step 6, simple beams 2, 3 and 5, 6 are attached
for theantisymmetric load case, by hinges and supported externally by rollers which does not change the
steps 5 and 6. SI 1 - Statically statical determinacy quality. The analysis concludes with a single degree
Indeterminate to the 1 degree. of statical indeterminacy.

The rest is standard. The same abbreviations apply as in the previous section.
PRIMARY STRUCTURE No :
4 5 6 1/2 −0.166 −0
l
.2 joint AE No N1 ∆ δ N
0.166
35
0.33

−0

1, 2 1 .5 −.7 −.3 .5 .35


1
X

7
.4
7

2, 3 .50 0 0 0 0 0
5
23

1/2 1, 4 1 .33 −.7 −.23 .5 .14


0.

1 −0.5
2 3 1, 5 1.4 0 1 0 1.4 .24
2, 4 1.4 −.47 1 −.66 1.4 −.22
N1: −0.707 0 FINAL FORCES:
2, 5 1 .165 −.7 −.12 .5 0
−0.33
2, 7 .7 .23 0 0 0 .23
−0
0
1

.2
−0.707

−0.707

5, 7 .7 −.23 0 0 0 −.23
3
0.14
1

4, 5 1 −.165 −.7 .12 .5 −.33


24

−0

23
0
1

0.

.2

0.

5, 6 .5 0 0 0 0 0
2

−0.707 0 0.33 sum −1.25 4.83

Sketch 37 Antisymmetric load case, primary struc- Read 0.707 instead of .7, 1.414 instead of 1.4, 0.353
ture, elementary load cases and final forces. instead of 0.35 and 0.125 instead of 0.12 in the table.

Redundancy X1 = 1.25/4.83 = 0.245. The final member forces are displayed in a truss scheme for conve-
nience. The combination of the elementary load cases that makes up the final forces is rather intricate and
is therefore resumed here.
FINAL FORCES:
−0.83 −0.5 −0.17 On the left symmetric half:
−0

22

+ No from the symmetric load case (recall that it


.2

−0.14
0.
3

equals the final forces of the symmetric load case)


0.14

+ N from the antisymmetric L.C.


−0
24

−0

23

.2
0.

.2

0.

On the right symmetric half:


4
2

0.83 0.5 0.17 + No from the symmetric load case


− N from the antisymmetric L.C.
Sketch 38 Final forces.

3.11 Selection of the primary structure


There are no clear and general rules for the selection. A rather vague recommendation says that the removed
constraints should possibly divide the structure in parts which affect each other as little as possible.

30
M4 : M5 :

X1 X4
X2 X3 X5
X6 M4 : M5 :
X X5 X4
X2 1 X6
X3

Sketch 39 Two selections of the primary structure. In the second primary structure, internal
constraints are released and the redundants are shown just for one side of each gap for lack of
space.

In Sketch 39, two selections are shown of the primary structure. The integrations of the bending moment
diagrams for δ4,4 , δ4,5 and δ5,5 apparently are simpler for the other selection and the same is true for the
other redundants. The same issue can also be demonstrated on continuous beams, which is the topic of the
next section.

3.12 Three moment equation


A continuous beam is probably the most frequent statically indeterminate structure in practice.

L C R
lL lR A specific selection of the primary structure leads to a simple and
xL xR elegant solution of arbitrary continuous beams. Relative rotation
XL XC XR constraints above each support are released which transforms the
continuous beam in a sequence of simple beams. Redundants are
exclusively bending moments over supports. The point of this
Mo
choice is that compliance factors δi,j vanish when i and j are not
adjacent or the same redundants. This property and values of δi,j
can be derived from a generic section of an arbitrary continuous
δML
beam as shown in Sketch 40. For brevity, the development is con-
fined to beams with per-bay-constant cross-sections. Potential gen-
1
δMC eralization to variable cross-sections is straightforward. General
loading is assumed and represented in Sketch 40 by the variable
1
continuous load. Bending moment Mo diagram thus is just quali-
δMR tative. ! ! !
1 l l
δC,C = 1·1 +1·1
final M 1 3 EJ L EJ R
!
1 l
δL,C = ·1·1
6 EJ L

Sketch 40 Specific primary structure


in a continuous beam.
Z Z lR
lL xL xR
∆o,C = Mo dxL + Mo dxR
0 (EJl)L 0 (EJl)R

31
The last expression simplifies when constant cross-section is assumed in each bay:
!
1 l
δC,R = ·1·1
6 EJ C

Mo,L Mo,R
z }| { z }| {
Z Z
1 lL 1 lR
∆o,C = xL Mo dxL + xR Mo dxR
(EJl)L 0 (EJl)R 0

Integrals Mo,L and Mo,R can be interpreted as static moments of the respective Mo subareas with respect
to vertical axes going through points L and R respectively. The compatibility equation for redundant XC is
multiplied by 6 for convenience and reads then
load terms
! ! ! ! ! z }| !{
l l l l Mo,L Mo,R
XL + 2 XC + + XR +6 + =0 (11)
EJ L
EJ L
EJ R
EJ R
(EJl)L (EJl)R

Three bending moments – redundants appear in the last equation and that is why it is called the three
moment equation. It is valid for any intermediate support of a continuous beam with per bay constant
cross-sections. The load terms are extended to include the effect of support settlement and thermal effect in
section 3.12.2

3.12.1 Applications to simple instances


Application of the three moment equation to the beam in Sketch 41 is straightforward. It is assumed that
the cross-section is constant in the whole beam.
f

A B C The three moment equation is applied with L = A, C = B and R = C.


l l Apparently XL = XR = 0 and a single redundant remains. The load terms
are Mo,L = 0 and
Mo Mo,R f l3
6 =
FINAL M: fl /8 lR 4
fl /16
The three moment equation becomes
9l/16 2l f l3
2 XB + =0
EJ 4 EJ
Sketch 41 The simplest in-
2
stance of a continuous with solution XB = − f16l . The rest of the computations is elementary.
beam.

The next instance beam has a clamped end A. The bending moment at A is redundant but there is no
L bay here. In spite of it the three moment equation can be applied. The terms of the L bay simply are
omitted – recall that individual terms in the three moment equation are contributions of the adjacent simple
beams to the relative rotation at the central support. When there is no adjacent bay, there is no contribution.
The bending moment at B is known so that no equation need be written for it.

32
F Instead, the moment at B is substituted for XR in the three moment equation
at support A where the contribution from the left bay is omitted:
A B
l lc l l
2XA + (−F lc ) =0
EJ EJ
Sketch 42 The simplest
F lc
continuous beam with a XA =
clamped end. 2

3.12.2 Load terms in the three moment equation


For the most frequent loads, the load terms in the three moment equation are summarized.
a F b
I J Mo,I F a b (2a + b) Mo,J F a b (a + 2b)
l 6 = 6 =
l l l l

f Mo,I Mo,J f l3
6 =6 =
I l l 4
J
The actual values of L, C and R need be substituted for I and J in these expres-
Sketch 43 sions.

The following terms replace/supplement the load terms in the basic equation (11) if thermal effect and/or
support settlement is present instead of, or additionally to the force effect.
∆ t= t upper− t lower
h

Temperature gradient:
! ! !
l α∆tl α∆tl
−3 +
h L
h R

sL sC Support settlement:
sR
 
lL lL sL − sC sR − sC
6 +
lL lR
Sketch 44

3.13 Method of least work and Castigliano’s second theorem


Compatibility equations are derived using superposition of displacements in these notes and the deflections
appearing in them are computed with the aid of the PVf. The same expressions for the deflections can be
derived in a more elegant way from Castigliano’s second theorem. It is necessary to stress, however, that
nothing new and more effective can be achieved with respect to practical application. An outline of the ideas
involved is provided here in order to establish context with work energy methods. Skipping this section does
not impair the reader’s ability to solve problems. On the other hand, books on computer structural analysis
and the finite element method , in particular, nearly exclusively develop the basic equations from work
energy principles. A somewhat broader perspective of the methods used here might therefore be useful.

33
3.13.1 Castigliano’s second theorem
The theorem can be derived from the PVf. The development is confined here to frames subject to bending
and the normal forces effect is neglected. Extension to other structures is straightforward, the expressions
for work need to be completed/replaced. A simple frame is considered in Sketch 45 for demonstration
purposes.

wn
P 1
Suppose deflection wn is to be computed, which is
w P
the projection on direction n of displacement vec-
n n tor w
~ of point P . The actual stress state is M , in
M δM equilibrium with external load (bending moment
diagrams are not shown in Sketch 45). The given
external loading is represented by the partial uni-
form load in Sketch 45 but the development is valid
1 1 for any loading. By the PVf
P P Z
M
w= δM dx (12)
n n f rame EJ
δM δM
Virtual stress state δM is in equilibrium with the
unit virtual force acting at the location P and di-
rection n. This is valid for both statically indeter-
minate and determinate structures.
Sketch 45 Deflection by Castigliano’s second theorem

Up to this point, the standard PVf has been resumed for the computation of deflections. The virtual stress
state δM (top right diagram in Sketch 45) is now used in another function, namely it is superimposed on
the actual stress state M to obtain variations of M . When the magnitude of the variation is X, the varied
state M ∗ reads
M ∗ = M + X δM
The strain energy U ∗ associated with stress state M ∗ is
∗ 1Z M ∗2
U = dx (13)
2 f rame EJ
Strain energy is the recoverable elastic energy stored in the material in the course of the deformation. Unlike
potential energy, virtual work and some other energies that occur in structural mechanics, strain energy has
direct physical meaning. If, in an experiment, the real structure was gradually loaded until the state M ∗ was
reached nad the work spent in the loading process was recorded, its value would be U ∗ . It is easy to show
that the partial derivative of U ∗ with respect to X evaluated at X = 0 equals w:
∂U ∗ ∂ 1Z M ∗2 Z
M ∗ δM Z
M δM Z
δM 2
= dx = dx = dx + X dx
X X 2 f rame EJ f rame EJ f rame EJ f rame EJ

For X = 0, expression (12) is obtained. Obviously, the derivative with variable X represents the deflection
due to the combined effect of the given load and a variable force X. The last equation specifies the exact
meaning of Castigliano’s theorem (1873). Its verbal form
Principle 1 Partial derivative of the strain energy with respect to an applied force is equal to the displace-
ment of the point of application along the line of action of the force.
apparently is not quite precise since the meaning of the strain energy is not exactly specified.

34
3.13.2 Method of least work
The Castigliano’s theorem development can be readily extended to a number of deflections wi , the associ-
ated virtual states δMi and their variation factors Xi . The varied stress state then is:
X
M∗ = M + Xi δMi (14)
i

Equation (13) remains valid and


∂U ∗
wi = (15)
Xi (Xi =0)
Once again the deflection wi (X) due to the combined effect of the given load and all variable forces Xi
simply is the derivative
  2   
∗ Z Z
∂U ∂ 1 X X
wi (X) = =  M + Xj δMj  dx
= M + Xj δMj  δMi dx
Xi Xi 2 f rame j f rame j

Z X Z
= M δMi dx + Xj δMj δMi dx (16)
f rame j f rame

Stress state M ∗ is the central concept of the development and deserves some remarks:

• M is the actual stress state in equilibrium with the given external load. The associated deformation
matches the displacement compatibility conditions. These conditions imply that this stress state is
unique.

• δMi may be any appropriate stress state in equilibrium with the respective unit force δ1i . Whereas
it is unique in a statically determinate structure, a set of admissible states exist to each δ1i in an
indeterminate structure. In particular, the stress states δMi can be obtained by solution of any pri-
mary structure in case the actual structure is indeterminate. Two such possibilities are indicated in
Sketch 45.

An important special application of the extended Castigliano’s theorem represents the case when the actual
structure is a primary structure of the solution by the force method and Xi factors are the redundants. Stress
states δMi are then unique. Compatibility is preserved when

wi (X) = ∆i

where ∆i are imposed displacements and X stands for the vector of redundants Xi . These equations
determine Xi . It is easy to see that they coincide with equations (6). In particular, when there are no
imposed displacements, that is when ∆i = 0, then it holds
∂U ∗
=0
∂Xi
Theses equalities mark out an extrem of the strain energy. It can be shown that it is the minimum. The last
equation thus is verbally expressed in the principle of least work:

Principle 2 The magnitudes of the redundants of a statically indeterminate structure are such that the strain
energy stored in the structure is minimum.

35
4 Slope deflection method
Slope deflection method was originally proposed by Mohr for computing secondary stresses in trusses. G.
A. Maney presented it as a method in its own right in 1915. Its relatively simple algorithm has always
attracted engineers although frames with sway spoiled the picture. The computer era removed this draw-
back since the number of primary unknowns ceased to be critical, independent joint translations could be
included and the displacement method was adopted for practice. Popularity and algorithmic simplicity of
the displacement method inspired attempts to extend it to two-dimensional continuum problems which gave
birth to the finite element method . Computer codes have made the slope deflection method obsolete for
engineering practice. In spite of it, it remains an important part of the courses in structural mechanics for
its didactic value. There is no better way to show students what goes on in a computer code than to solve a
few examples with the slope deflection method.
The basic assumption of the slope deflection method is that the elastic axial deformation of straight
beams is negligible. In some frames this assumption implies that joint translations vanish entirely. Such
frames are called frames without sidesway. The algorithm of the slope deflection method is particularly
simple for these frames and the development starts with them.

4.1 Basic equation


A straight prismatic beam with fixed ends is considered with general continuous load f . The bending
moments at both ends are selected for redundants but with regard to later use the sign convention of the
moments is different than standard, see Sketch 46 and a new term end moments is introduced for them.
Besides the continuous load, rotations φ1 and φ2 are imposed at both ends.
f
The structure is nominally indeterminate to the 3rd degree but lateral and
x l
axial effects decouple. The axial displacement vanish and axial forces
remain indeterminate. The lateral effects alone leave two degrees of in-
φ2 determinacy. Standard force method is used for the solution in the lateral
direction. Compliances are the same as in the three moment equation in
φ1 section 3.12 but for two sign changes due to the opposite positive sense
X1 X2 of the redundant at the left support:
Mo,R
z }| {
Mo: Z l Mo 1 Z l
∆o,1 = M1 dx = − Mo (l − x)dx
M1 : 0 EJ EJ l 0

1 Mo,L
z }| {
Z
M2 : 1 l
∆o,2 = Mo xdx
EJ l 0
1
l l
δ1,1 = δ1,2 = − δ2,2 = δ1,1
Sketch 46 Straight prismatic 3EJ 6EJ
beam with clamped ends.

Solution of the compatibility equations yields:


f
M1,2
z }| {
2EJ 4 2
X1 = (2φ1 + φ2 ) + 2 Mo,R − 2 Mo,L (17)
l l l

36
f
M2,1
z }| {
2EJ 4 2
X2 = (2φ2 + φ1 ) − 2 Mo,L + 2 Mo,R
l l l
Redundancies X1 and X2 with positive senses indicated in Sketch 46, are total end moments. When the
f f
end rotations vanish the rests of the end moments are the fixed-end moments M1,2 and M2,1 . Fixed-end
moments have been integrated for the most frequent load types including temperature gradient and usually
are provided in tables in textbooks.
The lateral effects on the beam are completed next with the last one – the lateral translation of the ends.
The end moments due to this effect are easily obtained when the translation is carried out in two steps as
indicated in Sketch 47.
v
Obviously, only relative translation of the joints induces any forces. The
relative lateral translation is best quantified by beam rotation ψ, positive
counter-clockwise. In the first step, the beam is rotated by ψ as rigid.
ψ No elastic deformation and forces are induced in this step. In the second
step, the ends are rotated back to their original horizontal orientation by
M2,1 angle −ψ. The end moments due to these two rotations can be evaluated
M1,2
−ψ with the aid of equations (17). The compound effect of joint rotations
−ψ φ1 and φ2 , beam rotation ψ and external load then is
2EJ f
Sketch 47 The effect of the rela- M1,2 = (2φ1 + φ2 − 3 ψ) + M1,2
l
tive lateral translation of ends.

The expression for the other end moment can be simply obtained when subscript 1 is changed to 2 in the
equation. Universal equation can thus be written
2EJ f
Mi,j = (2φi + φj − 3 ψi,j ) + Mi,j (18)
l i,j
where standard notation is used for joints and members. Joints are denoted by single numbers, members by
pairs of numbers of adjacent joints. For brevity, symbol ki,j is adopted for the beam bending stiffness
2EJ
ki,j = (19)
l i,j
The expression for the end moments assumes the final form
f
Mi,j = ki,j (2φi + φj − 3 ψi,j ) + Mi,j (20)

A table of the fixed-end moments is provided for easy reference:


Table 2 Fixed-end moments
f f f f
load Mi,j Mj,i load Mi,j Mj,i
Mi a b Mj Mi a Mj
F f
F ab2 −F a2 b 2 −8al+3a2
l
l2 l2
l
f a2 6l 12l2
f a3 3a−4l
12l2
a b f
Mi Mj Mi Mj
M
l
M b 2a−b
l2
M a 2b−a
l2
l
f l2 /20 −f l2 /30
Mi ∆ t= tupper− t lower h
Mj
Mi f Mj
l
l
f l2 /12 −f l2 /12 α ,EJ constant − ∆t αh EJ ∆t α EJ
h

37
4.2 Moment equilibrium equations
Slope deflection method features an a priori compatibility of displacements. In order to grasp the basic
idea, imagine that rotations are imposed to each joint. This is associated with elastic deformation of the
connecting members. Compatibility is conserved in this process.

φi k
k The rotated joint i with three neighbours is
shown in Sketch 48. Owing to elastic defor-
M i,j Mi,k
mations, internal forces and end moments are
j φi induced in members. The end moments are
φi i j i
l
Mi,l assigned an extended subscript to distinguish
l
the member. This notation is not standard and
is utilized here just temporarily for the sake of
Sketch 48 Joint i rotated and adjacent members deflections mathematical rigor.

Moment equilibrium of each joint requires that the sum of all end moments at the connecting members must
equal zero: X
Mi,α = 0 (21)
α=j,k,l,...

The sum goes over all connecting members and the equation is applicable to all joints. The equation actu-
ally is not moment equilibrium of the joint but its reverse. In order to have the joint moment equilibrium,
equation (21) need be multiplied by (−1). It would be purely formal and with this reserve in mind, equa-
tion (21) is simply referred to as the joint equilibrium equation. Caution is necessary when there is an
external couple acting at a joint. This couple must be included in the joint equilibrium condition (21) with
proper sign.
By substitutions from equations (20), a system of n linear equations is obtained for n unknown joint
rotations when the frame has n joints. The rotations, once solved for, are substituted back in equations (20)
to deliver end moments in all beams. The rest of the solution is performed exclusively with the aid of suit-
able equilibrium conditions. This remarkably simple algorithm makes up the trick for frames without sway
– frames with no translation of joints. A simple example demonstrates the application.
f

1 2 Note that the neglected axial elastic deformation implies no translation of joint 3. There
is just one joint that can rotate, joint 2. The relevant end moments in the adjacent mem-
bers are
f l2
3 M2,1 = k1,2 2φ2 − 1,2
12
Sketch 49 A
M2,3 = k2,3 2φ2
frame without
sway

according to formula (20). The expressions are simple, since the rotations at clamped ends 1 and 3 vanish.
Moment equilibrium condition
M2,1 + M2,3 = 0
yields
2
f l1,2 1
φ2 =
12 2(k1,2 + k2,3 )

38
With φ2 known, all end moments can be evaluated using the above expressions and the expressions for the
remaining two end moments:
2
f l1,2
M1,2 = k1,2 (2 · 0 + φ2 + 0) +
12
M3,2 = k2,3 φ2
Note that nothing beyond formula (20) is necessary.
In the next step, shear forces are computed on the member by member basis. Free body diagrams of the
members are helpful.
2
V2,3
V1,2 f M2,1 M2,3 beam 1,2 1 ⊙:
1 2
c
l2,3

M1,2 V2,1 2
f l1,2
l1,2
M2,1 + M1,2 − V2,1 l1,2 − =0
M3,2 2
V3,2
3 M2,1 + M1,2 f l1,2
V2,1 = +
l1,2 2
Sketch 50 Free body diagrams, internal
forces of individual members

M2,3 + M3,2
beam 2, 3 3 ⊙ : M2,3 + M3,2 − V2,3 l2,3 = 0 V2,3 =
l2,3

End moments and shear forces are known and the shear force and bending moment diagrams can be drawn.
In the last step, the end normal forces are determined from the equilibrium conditions of joint 2.
joint 2 N2,3
N2,1 N2,1 = −V2,3 N2,3 = V2,1
V2,1
V2,3 This step need not always yield a unique solution. If another member connected
to the joint there would be more unknown normal forces than available conditions.
This ambiguity associates with the basic assumption of the slope deflection method,
Sketch 51 Joint equi- the neglect of the normal forces effect, and cannot be resolved in the frame of the
librium (moments method. Unique normal forces are always obtained with the displacement method.
omitted)

Exercise problems are provided in appendix 9.2 on the identification of the primary unknown rotations and
the assembly of the equilibrium conditions.

4.3 Beam with a hinge


Where there is a hinge (internal or external), rotations of the parts connected by the hinge are independent
. The additional rotation(s) (primary unknown(s)) can always be eliminated from the condition that the
bending moment at the hinge vanishes. There is an external hinge in the frame in Sketch 52 and a single
rotation φ1 can be eliminated.

39
1 2 There are two moment equilibrium equations in case of the frame in Sketch 52:

M1,2 = 0 M2,1 + M2,3 = 0


3
The first equation reads after substitution
Sketch 52 A f
simple frame M1,2 = k1,2 (2φ1 + φ2 − 3ψ1,2 ) + M1,2 =0
with hinge

The beam rotation is kept in the equation deliberately to obtain a sufficiently general result. From this equa-
tion, φ1 can be expressed in terms of φ2 and ψ1,2 and substituted subsequently in the formula for M2,1 :

3 Mf
f
M2,1 = k1,2 (2φ2 + φ1 − 3ψ) + M2,1 f
= k1,2 (φ2 − ψ1,2 ) + M2,1 − 1,2
2 2
This pre-elimination is always possible. The net effect is that for a beam with a hinge at one end, a modified
formula for the end moment at the other (fixed) end can be used:

3 Mf
f
Mi,j = ki,j (φi − ψi,j ) + Mi,j − j,i hinge at j Mj,i = 0 (22)
2 2
The elimination of the rotation at the hinge is relatively simple and can be carried out again each time it
is necessary. Formula (20) remains then the single one necessary in the slope deflection method. On the
other hand, utilization of formula (22) makes solutions of frames with hinges faster and it probably is worth
remembering. For the actual example
3 f l2 1 f l2 3 f l2
M2,1 = k1,2 φ2 − − = k1,2 φ2 −
2 12 2 12 2 8
M2,3 = k2,3 2φ2 M3,2 = k2,3 φ2
The other equilibrium condition reads then
3 f l2
k1,2 φ2 + k2,3 2φ2 − =0
2 8

4.4 Example frame


The bridge-like frame from section 3.9.1 is solved again for a uniform load upon the girder. Merits of the
force and slope deflection methods can thus be compared. The structure scheme is repeated for convenience
in Sketch 53.
f J1
2
1

3 J J2
5 0.5 Material - concrete, E = 30 GPa, α = 10−5
5

2
1 Cross-sections data: J1 = 1/24 = 0.0415, J2 =
5 10 10 5 1/8J1 .
0.5

The force method solution in section 3.9.1 utilized


0.5 kinematic conditions on the symmetry axis. This
approach to symmetry is applicable in the slope
deflection, too.
Sketch 53 Bridge-like frame from section 3.9.1 sub-
ject to uniform loading on the girder.

40
However, joint 4 has to be inserted on the symmetry axis, this joint moves vertically and, consequently, sway
appears (see Sketch 28). In order to avoid the nuisance, the other approach to symmetry is adopted here
– whole structure is considered, equality φ3 = −φ5 is utilized and joint 4 omitted. A single independent
primary unknown φ3 determines the frame displacement when formula (22) is used.
The end moments at joint 3 are
3 f l2
M3,2 = k2,3 φ3 − 2,3 M3,1 = k1,3 2φ3
2 8
2
f l3,5 f l2
M3,5 = k3,5 (2φ3 + φ5 ) + = k3,5 (2φ3 − φ3 ) + 3,5
12 12
The single moment equilibrium equation reads
 2

l2,3 l3,5
f 8
− 12
3 ⊙ : M3,2 + M3,1 + M3,5 = 0 φ3 =
1.5k2,3 + 2k3,1 + k3,5
Prior to number substitutions, beam bending stiffnesses are expressed in terms of a reference k = k1,3 for
brevity:
EJ2 EJ1 EJ1 EJ1
k = k1,3 = 2 =2 = 0.05EJ1 k2,3 = 2 = 8k k3,5 = 2 = 2k
5 40 5 20
Joint rotation φ3 and end moments:
f (−30.2) f (−30.2) f
φ3 = = = −1.89
k(12 + 2 + 2) 16k k
f f 52 f
M3,2 = 12 (−1.89) − = −25.8 f M3,1 = k · 2(−1.89) = −3.78 f
k 8 k
M3,5 = f (−3.78 + 33.3) = 29.5 f M1.3 = k φ3 = −1.89 f
A bending moment diagram can be drawn with end moments known. Shear forces need be computed from
the equilibrium of individual members.
2
l2,3 1
member 2, 3 3 ⊙ : − V2,3 · 5 + f + M3,2 = 0 V2,3 = f (12.5 − 25.8) = −2.66 f
2 5
It is worth mentioning that the reaction at the roller pulls down according to this shear force. The bridge
designer probably would tend to lower the ratio of the central to the side spans to avoid negative reactions
in the rollers.
M1,3 + M3,1 1.79 + 3.58
V1,3 = V3,1 = = −f = −1.06 f
l1,3 5
f l3,5
member 3, 5 ↑ : V3,5 − V5,3 − f l3,5 = 0 V3,5 = = 10 f
2
Symmetry implied equality V3,5 = −V5,3 is utilized in the last equation. In this particular structure, unique
normal forces can be determined.

joint 3 → : V1,3 − N3,5 = 0 N3,5 = −1.06 f

joint 3 ↑ : − V3,2 + V3,5 + N3,1 = 0 V3,2 = V2,3 − f l2,3 N3,1 = f (2.66 + 5 + 10) = 17.66 f
Internal forces diagrams are provided to complete the example except the normal force which is too trivial
to waste space for.

41
10 M: 25.8 29.5
V:
3.78
−2.66 −7.66
20.5

−1.06 1.89

Sketch 54 Internal forces diagrams for the bridge-like frame

4.5 Frames with sidesway


Formulas (20) and (22) allow for member rotations ψi,j to be included in the equilibrium conditions. Each
member rotation becomes an additional unknown and equilibrium conditions must be extended accordingly.
A simple frame in Sketch 55 demonstrates the application.
ψ2,3
Beam rotation ψ2,3 , in addition to joint rotation φ2 , becomes the other
1 2 primary unknown in this case. Note that ψ2,3 is determined by the trans-
lation of joint 2. It is not the slope of the actual deflection line at joint 2.
l Average slope deflection would also be a correct term for ψ2,3 but beam
l

rotation is preferred. With pre-eliminated φ1 , just one moment equilib-


rium equation is available, M2,1 + M2,3 = 0. Two primary unknowns
3 require another equation to solve the problem. The horizontal equilib-
rium condition of the horizontal girder serves the purpose:
Sketch 55 Simple frame with V2,3 = 0 ⇒ M2,3l2,3 +M3,2
=0
sidesway

End moments follow from formula (22):


f f
M2,1 = 1.5 φ2 k1,2 + M2,1 − 0.5 M1,2 , M2,3 = 2 φ2 k2,3 − 3ψ2,3 k2,3
M3,2 = φ2 k2,3 − 3ψ2,3 k2,3
The two equilibrium equations become after substitutions:
f f
2 ⊙ : φ2 (1.5k1,2 + 2k2,3 ) − 3ψ2,3 k2,3 + M2,1 − 0.5 M1,2 =0
girder1, 2 →: 3φ2 − 6ψ2,3 k2,3 = 0 ⇒ ψ2,3 = 0.5φ2
Assuming k1,2 = k2,3 = k for simplicity, the primary unknowns are
f l2 f l2
φ2 = , ψ2,3 =
16k 32k
and the end moments can be evaluated
f l2
M2,1 = M − 3, 2 = −M2,3 = −
32
It is important to check the results by back-substitution in the two basic equilibrium conditions. They are
obviously satisfied in this simple example. The check is relevant with respect to the solution of the system
of equilibrium equations and the back-substitution. It does not guarrantee that the equations have been
properly set up. It is left to the reader to finish the example by developing the internal forces diagrams.
Check is provided:

42
−0.0312
0.469 V: M:
0.0312
0.469 l
−0.531 0.109

fl fl 2
0.0312
Sketch 56 Simple frame with sidesway - internal forces

The concept can be extended to more complicated frames. Each possible beam (member) rotation becomes
an additional unknown and equilibrium conditions must be extended accordingly. In problems with rect-
angular systems of beams, see sketches 63 and 64, the equilibrium conditions of floors in the horizontal
direction or columns in the vertical direction deliver the necessary additional equations. A consistent de-
velopment of the additional conditions is presented later on, based on the principle of virtual work. In
the simple case in Sketch 57 the condition is rather elementary – the horizontal equilibrium of the girder.
Recall that this load case is also solved by the force method in section 3.9.1. The load is decomposed in
the symmetric and anti-symmetric parts. The symmetric solution is trivial (see section 3.9.1) and just the
antisymmetric component is considered here. The split structure technique is suitable for the antisymmetric
loading.

u u F/2 u The displacements determining the deflection and


the primary unknowns at the same time are φ3 and
2 J1 3 J1 4 ψ1,3 when the pre-elimination of φ2 and φ4 is con-
ψ1,3(<0)
sidered.
J2 = J1 /8 M3,2 = 1.5k2,3 φ3
J2

1 M3,1 = 2k1,3 φ3 − 3ψ1,3


5 10 M3,4 = 1.5k3,4 φ3
M3,4 could also be obtained via the whole struc-
Sketch 57 An antisymmetric part of the bridge-like ture technique. In virtue of antisymmetry, φ5 =
frame from section 3.9.1 subject to horizontal (brake) φ3 , where φ5 is the rotation of the joint 5, see
force on the girder. The constraint on the symme- Sketch 53. Then M3,4 = M3,5 = 2φ3 k3,5 +
try axis is drawn by a dashed line and the tenta- φ5 k3,5 = 3φ3 k3,5 = 1.5φ3 k3,4
tive deflected shape is shown. Material - concrete,
E = 30 GPa, α = 10−5 M1,3 = k1,3 − 3k1,3 ψ1,3
Cross-sections data: J1 = 1/24 = 0.0415, J2 =
1/8J1 .

Besides the moment equilibrium of joint 3


joint 3 ⊙ : M3,1 + M3,2 + M3,4 = 0
equilibrium of horizontal forces on the girder is available:
F M1,3 + M3,1 F
girder → : −V3,1 − = − − =0
2 l1,3 2
Substitutions deliver two equations for the primary unknowns φ3 and ψ1,3 :
φ3 (1.5k1,3 + 2k1,3 + 1.5k1,4 ) − ψ1,3 3k1,3 = 0

43
1 1 F
φ3 3k1,3 − ψ1,3 6k1,3 =
l1,3 l1,3 2
The system matrix becomes symmetric with respect to the main diagonal when the second equation is
multiplied by −l1,3 . This can be achieved with any system of two linear algebraic equations, of course.
Nevertheless, it is the consequence of a general property. Namely, it can be deduced from the PVW that

Rule 4.1 any system of equations of the slope deflection method can be brought to a symmetric system
matrix.

The solution of the above system is


F F
φ3 = −0.54 ψ1,3 = −3.6
k k
Back substitution delivers then

M3,2 = −0.812 F M3,1 = 0.121 F

For the internal forces diagrams, look up Sketch 30 in section 3.9.1.

4.5.1 System matrix symmetry


The assembly of the equilibrium equations is exercised further on the frame in Sketch 58. The primary
unknowns are φ3 , φ4 and ψ1,3 = ψ2,4 . Three of them are independent. The equilibrium conditions read:
F 3 4
h


3 ⊙ : M3,1 + M3,4 = 0 


1 2 4 ⊙ : M4,3 + M4,2 = 0 
(23)
l 3 − 4 →: − V3,1 − V4,2 + F = 0 ⇒ 
s 


⇒ −(M3,1 + M1,3 + M2,4 + M4,2 ) + F h = 0
Sketch 58 A frame loaded
by a force

Substitution for the end moments from


M3,1 = (2φ3 − 3ψ1,3 )k1,3 , M1,3 = (φ3 − 3ψ1,3 )k1,3
M4,2 = (2φ4 − 3ψ2,4 )k2,4 , M2,4 = (φ4 − 3ψ2,4 )k2,4
M3,4 = (2φ3 + φ4 )k3,4 , M4,3 = (φ3 + 2φ4 )k3,4
yields three equations. They are exposed in a table pattern:

φ3 φ4 ψ1,3 l.c.1 l.c.2 l.c.3


3⊙: 2k1,3 + 2k3,4 k3,4 −3k1,3 0 0 −3 sl k3,4
(24)
4⊙: k3,4 2k2,4 + 2k3,4 −3k2,4 0 −3 hl tαk2,4 −3 sl k3,4
3 − 4 →: −3k1,3 −3k2,4 6k1,3 + 6k2,4 −F h −6 hl tαk3,4 0

The first right hand side column (l.c.1) is obtained for the present force load. The latter two refer to two
load cases to be considered later on. Note that the system matrix becomes symmetric when the last equation
is multiplied by −1. Rule 4.1 is thus confirmed.

44
4.5.2 General planar frames with joints translations
The procedure is demonstrated in an example frame in Sketch 59. It is a symmetric part of a bridge-
like frame again but the column is inclined now. An antisymmetric load case is considered, a horizontal
force acting upon the girder. Joint translations are possible but apparently not independent. General joints
translations can be treated in the following way:
• Release the constraints associated with end moments of all members, in other words, insert hinges
at these locations. In the demonstration frame in Sketch 59, the hinges are inserted at member ends
(1, 3), (3, 1), (3, 2) and (3, 4). The last three hinges are equivalent to a single hinge with triple
connection to the respective members. A set of joints connected by links emerges – effectively a
truss. The hinges are not shown in Sketch 59 but the displaced truss is shown instead.
• If the truss is geometrically stable (that is either statically determinate or indeterminate) then there
is no joint displacement possible and the frame is without sway. If the truss is unstable, there is
at least one virtual displacement possible which does not violate the constraints of the truss. The
displacement is imposed and the member rotation centers and rotation angles are determined with
the aid of the rules for virtual displacements of rigid bodies, see for instance [1]. This step is shown
in Sketch 59, including the rotation centers o1,3 , o2,3 and o3,4 . These are absolute, not relative centers
of rotation of the respective members since members are identified here with a pair of numbers.
• The virtual work is evaluated of all forces of the actual stress state (including the end moments of the
members, of course) upon the displacements of the imposed virtual displacements. The total of this
work is zero.
Geometric conditions imply
o3,4
l1,3
δψ1,3 l1,3 = δψ2,3 r2,3 δψ2,3 = δψ1,3
r2,3

l1,3
δψ1,3 l1,3 = δψ3,4 r3,4 δψ3,4 = δψ1,3
r3,4
4
3r,

The virtual work done by actual forces and moments on these vir-
tual displacements is
2 3 4 X F
δw = δψi,j (Mi,j + Mj,i ) − h δψ1,3 = 0
h

1 2
,3

all members
l1

o1,3
The generalized equilibrium equation becomes
o2,3
3
2r,

F
δψ1,3 (M1,3 + M3,1 ) + δψ2,3 M2,3 + δψ3,4 M3,4 − h δψ1,3 = 0
2
Sketch 59 Virtual displacement of the
truss which results when the rota- and after substitution
tional constraints at the ends of the l1,3 l1,3 F
members are released. M1,3 + M3,1 + M2,3 + M3,4 − h = 0
r2,3 r3,4 2

There are two primary unknowns φ3 and ψ1,3 in this case and the last equation together with the moment
equilibrium equation of joint 3 constitute the system of two equations to determine them.
The equation for the virtual work can be generalized to work for any frame:
X
δw = δψi,j (Mi,j + Mj,i ) + δwext f orces = 0 (25)
all members

45
where δwext f orces stands for the work of all external forces on the imposed virtual displacements.
Equations (20), (21) and (25) constitute the minimum mathematical basis of the slope deflection method.
Equation (22) can, but need not be, utilized to simplify the solution in the presence of hinges. Tables of
fixed-end moments like table 2 complete the necessary tools. The relevance of the slope deflection method
in the context of the contemporary structural mechanics is outlined in section 4.6.

4.5.3 Homogeneous temperature changes and support settlements


These effects are easy to treat in rectangular frames where the imposed beam rotations or their differences
are simple to tell since the rotations of the horizontal and vertical beams are independent. The number and
identity of the independent primary unknowns remains the same regardless of the load type.
The frame in Sketch 58 is suitable for demonstration of these assertions. Both the thermal load and
support settlements are indicated in the Sketch, nevertheless, the effects are treated separately. Needless to
say, they can be superimposed.
t
3 4 When a homogeneous temperature increase t takes place in the horizontal girder,
see Sketch 60, beam rotations ψ2,4 and ψ1,3 differ (ψ2,4 = ψ1,3 − t l α/h) but
h

still, the same three independent primary unknowns persist. The equilibrium con-
1 2 ditions (23) remain valid, too, except that some end moments must be edited. The
l s afflicted moment is:
M4,2 = (2φ4 − 3ψ2,4 )k2,4 = (2φ4 − 3ψ1,3 − 3t l α/h)k2,4
Sketch 60 Substitutions in (23) yield the same system matrix but another right hand side
Rectangular frame marked l.c.2 in equation (24).

When there is a horizontal support slide s at support 2, then ψ2,4 = ψ1,3 + s/l replaces the above relation
between ψ1,3 and ψ2,4 and the end moments are modified accordingly.
Next, suppose a homogeneous temperature increase t in leg 2 − 4, see Sketch 61.
3 4
There is an imposed beam rotation ψ3,4 = t h α/l. The afflicted end moments are
h
t

1 M3,4 = (2φ3 + φ4 − 3t h α/l)k3,4 , M4,3 = (φ3 + 2φ4 − 3t h α/l)k3,4


2
s
l Column l.c.3 in (24) contains the right hand side for this load case. When there is
a vertical support settlement s at joint 1, then ψ3,4 = −s/l and the end moments
Sketch 61 change accordingly. The matrix of the equilibrium equations in (24) remains valid.
Rectangular frame

4.5.4 Homogeneous temperature and support settlements in general frames


These effects might pose tricky problems in general frames with inclined members in the context of the
slope deflection method. With regard to the declining importance of the method it does not seem reasonable
to exercise the matter in detail. An illustrative example is provided to demonstrate the approach.

46
ψ2,3 The end moments are:
φ2
ψ1,2 2 3 M2,1 = 2k1,2 φ2 − 3k1,2 ψ1,2
s
M1,2 = k1,2 φ2 − 3k1,2 ψ1,2
M2,3 = 1.5k2,3 φ2 − 1.5k2,3 ψ2,3
1 α
Geometric conditions entail

Sketch 62 Frame with settle- ψ1,2 l1,2 cos α + ψ2,3 l2,3 = −s


ment support

The moment equilibrium condition is elementary

joint 2 ⊙ : ψ2 (1.5k2,3 + 2k1,2 ) − 3ψ1,2 k1,2 − 1.5ψ2,3 k2,3 = 0

Virtual displacement is selected so that the reaction in the roller does not do work (that is, δs = 0). This
condition implies
l1,2
δψ2,3 = −δψ1,2 cos α
l2,3
The virtual work associated with the joint translations is

δψ1,2 (M1,2 + M2,1 ) + δψ2,3 M2,3 = 0

After substitution for δψ2,3 the generalized equilibrium equation reads

l1,2
M1,2 + M2,1 − cos α M2,3 = 0
l2,3

In terms of the kinematic quantities the last equation is


!
l1,2 l1,2
φ2 3k1,2 − 1.5 cos α k2,3 − ψ1,2 6k1,2 − ψ2,3 1.5 cos α k2,3 = 0
l2,3 l2,3

It remains to substitute for ψ2,3 from the geometric condition ψ2,3 = −(s + ψ1,2 l1,2 cos α)/l2,3 . The final
system of two equations becomes then
!
l1,2 s
2 ⊙ : φ2 (1.5k2,3 + 2k1,2 ) + ψ1,2 −3k1,2 + 1.5k2,3 cos α + 1.5k2,3 =0
l2,3 l2,3
!  !2 
l1,2 l1,2
generalized : φ2 3k1,2 − 1.5 cos α k2,3 + ψ1,2 −6k1,2 + 1.5 cos α k2,3 
l2,3 l2,3
l1,2 s
+1.5 cos α k2,3 =0
l2,3 l2,3
The task simplifies substantially when the leg of the frame is vertical, cos α = 0. The solution is provided
mainly for illustration, since the slope deflection method is seldom used for similar tasks in contemporary
design practice. Instead, the displacement method and computer codes are preferred.

47
4.5.5 Identification of primary unknowns
The algorithm of the slope deflection method is rather simple. Experience suggests that the critical step is
the identification of the primary unknowns, in other words, the identification of the displacement modes,
in frames with sidesway. The concept of virtual displacements offers a systematic treatment of the issue
as demonstrated in the previous section. An ’ad hoc’ identification is sufficient in most cases, however.
The two-floor frame in Sketch 63 is rather typical in this respect. The joint rotations φ1 − φ4 are apparent
primary unknowns.
φ3
φ4 ψ2
Columns restrain vertical translations of all
ψ3 ψ2 joints and the floor girders prevent rela-
ψ tive horizontal translations of the respective
ψ 2
2 joints. These constraints imply the displace-
φ1 ment modes and primary unknowns shown
φ2 −ψ4 ψ3
ψ1 ψ1 ψ1 in Sketch 63 (4+2=6 unknowns). Note
ψ1
a b that the displacement mode (dashed lines) is
shown without joint rotations. The dashed
line shape effectively is a virtual displace-
Sketch 64 Primary unknowns ment complying with the conditions specified
Sketch 63 Primary identification in frame with in the previous section.
unknowns identifica- horizontal and vertical sways. The sway can happen in other than horizontal
tion. directions, see Sketch 64.

Joint rotations are not shown at all in this figure to simplify the diagram. When all constraints are taken into
account, the indicated member rotations can be identified. Yet another constraint aψ4 = −bψ3 reduces the
number of independent member rotations to three. There are 6 + 3 = 9 primary unknowns in this example.
h3

The next example is a frame with inclined members and in-


ternal hinges. Member rotations of the inclined roof mem-
7 bers can be derived from the indicated independent rotations
φ3 ψ1,3 , ψ3,5 and ψ4,6 . The independent joint rotations shown
h3

in Sketch 65, follow when the pre-elimination of the rota-


tions at the hinges is utilized. Altogether 3+3=6 independent
primary unknowns exist in the frame. The generalized equi-
5 6
ψ4,6 librium conditions are best derived via the PVd. The simplest
ψ3,5 one is obtained when displacement δψ1,3 is imposed:
h2

φ1 4
(M1,3 + M3,1 + M2,4 + M4,2 )δψ1,3
3 φ2
ψ1,3 ψ2,4=ψ1,3 +work of the given external load = 0
h1

l/2 l/2 Virtual displacements δ3,5 and δ4,6 induce rotations of


the roof members, rotation δψ3,5 alone implies δψ5,7 =
1 2
−δψ3,5 h2 /(2h3 ) δψ6,7 = δψ3,5 h2 /(2h3 ).
Sketch 65 Primary unknowns in a frame

The generalized equilibrium condition reads


h2 h2
(M3,5 − M7,5 + M7,6 )δψ3,5 = 0
2h3 2h3

48
A similar equation can be obtained when δψ4,6 is imposed. There are other options, for instance virtual
displacements δψ3,5 = δψ4,6 and the associated symmetric one δψ3,5 = −δψ4,6 could be imposed instead
of separate rotations δψ3,5 and δψ4,6 . Other generalized equilibrium conditions would result which, never-
theless, must be equivalent to those derived above.

4.6 Closing remarks on the slope deflection method


Some aspects of the method deserve attention:

1. Primary unknowns in the system of equations are rotations of joints and members. In the displacement
method the joint displacement replace the members rotations. Anyway, they are always kinematic
quantities.

2. The equations are equilibrium conditions, some of them joint moment equilibrium conditions, others
are generalized conditions.

3. There is no space for subjective decisions like the selection of the primary structure in the force
method. This feature facilitates the computer code implementation.

4. Geometric relations between the member rotations in frames with sway are difficult to come to by a
simple algorithm. The sway phenomenon poses a serious problem with respect to full automation of
the solution.

5. The higher the degree of statical indeterminacy, the more favorable is the slope-deflection method in
comparison to the force method

6. The slope deflection method is seldom used in practice for tasks with inclined members. The dis-
placement method and computer solution are generally preferred in such cases. Nevertheless, there is
a timeless value in the slope deflection solutions at least from the theoretical point of view. Consider
a frame with progressively thinning members, d/l → 0 where d and l stand for member depths
and lengths respectively. The bending stiffnesses of the members decrease with the third power of
the ratio d/l, whereas the axial stiffnesses are just inversely proportional to it. Both stiffnesses are
indispensable in the displacement method, see the next section. As d/l → 0, the bending stiffnesses
effectively disappear from the equations and the system becomes ill-conditioned. The displacement
method and computer codes then face serious troubles, whereas the slope deflection methods works
well. This is an instance of a more general phenomenon called ’locking’ in finite element literature.
The phenomenon is a serious reason not to use computer codes as black boxes.

5 Displacement method
Beam axial deformation is accounted for in the displacement method. Basic equations are obtained as a
simple extension of the equations of the slope deflection method in section 4.1. Notation and sign conven-
tions are introduced in Sketch 66.

49
i j
φi
φj

ui In order to have strictly conjugate displacement and force quantities, end


vi
vj forces Yi and Xi assume opposite positive senses than the correspond-
uj ing internal forces (recall that end moment Mi,j in the slope deflection
method also has opposite positive sense than the corresponding bending
Yi Yj
Mi EJi,j Mj moment at section i). The lateral and axial deformations and forces re-
Xi Xj main decoupled in a straight beam. Note that the lateral end forces Yi
li,j
depend on the difference vj − vi of the lateral end displacements, not
on their absolute values and the same applies to the axial end forces and
Sketch 66 Notation and positive displacements.
sense for end displacements and
forces

5.1 End forces and moments


End moments can be expressed in terms of the end displacements with the aid of equation (20) since
ψi,j = (vj − vi )/li,j :
3
Mi = ki,j (2φi − φj − (vj − vi ) + Mif
li,j
The formula for Mj follows when the subscripts are swapped. End lateral forces are
Yif
! z }| {
Mi + M j 3ki,j 2 1
Yi = Vi = + Yif,0 = φi + φj + (vi − vj ) + (Mif + Mjf ) + Yif,0
li,j li,j li,j li,j

Yi f,0 Yj f,0

Sketch 67 Free-end lateral forces Yif,0 are defined in Sketch 67 as the reactions of a simple beam,
Free-end lat- whereas Yif are the fixed-end lateral forces on the beam with both ends clamped.
eral forces Yi
definition.

Axial stiffness is for prismatic beams


EAi,j
(ui − uj ) Xi =
li,j
Ordering the kinematic and force quantities properly, a matrix equation can be written for the end forces of
a beam. Subscript i, j of the beam is omitted for brevity in the equation and in the rest of the present section
since all quantities refer to a single beam:
{S} [K] {u} {F }
z }| { z }| {z }| { z }| {
 
 Xi  EA
0 0 − EA 0 0  ui  
 0  

 
 l l    f 


 Yi 
  k k
6 l2 3 l 0 k
−6 l2 3 kl  
vi 

 Y 

 
   
 
 i 

    
 
 M 
 f

 M 
i  2k 0 −3 kl k  φ 
i i

= 
EA
 + (26)


 Xj 



 l
0 0 
 uj 




 0 


k
6 l2 −3 kl Yjf
      




Yj 



 


vj 











     f 
Mj 2k φj 
Mj 

50
where the sub-diagonal symmetric part of matrix [K] is omitted. Matrix [K] alone can represent the beam
elastic properties for the rest of the frame. Computer codes for planar frames are almost exclusively based
on this matrix equation. The square matrix [K] is the beam stiffness matrix. Matrices are divided in sub-
matrices in equation (26) in order to facilitate the next step, which is generalization to inclined beams.
Translations and forces components in equation (26) are in the beam local
(y,v,Y) g
v,Y l

coordinate system with x-axis in the beam axis. Beams of various orienta-
)

, N )l
(y,

j (x,
u tions occur in a general frame. Some unique coordinate system needs to be
i α
established in each frame in order to have common primary unknowns and
guarantee correct superposition of the end forces at joints. This common co-
(x,u,N)g ordinate system is the global one. If the beam axis in inclined with respect
to the global x coordinate axis by angle α then standard transformation ma-
Sketch 68 Global and lo- trix [T ] can be utilized to obtain the correct components of the translation and
cal (beam) coordinate force vectors. Neither rotations φ nor the end moments M depend on the
coordinate system orientation.
For joint i     
 ui 
  cos α sin α 0   ui 
 
vi =  − sin α cos α 0  vi or {ui }l = [T1 ]{ui }g (27)
   

φi l 0 0 1  φi g
The same matrix is applicable to the force components

{Si }l = [T1 ]{Si }g

The transformation matrix for all components of a beam can be composed


( ) " #( )
{ui }l [T1 ] 0 {ui }g
= or {u}l = [T ]{u}g (28)
{uj }l 0 [T1 ] {uj }g

and for the end forces


{S}l = [T ]{S}g (29)
Both transformation matrices [T1 ] and [T ] are orthogonal, it holds

[T ]−1 = [T ]T [T1 ]−1 = [T1 ]T

Equation (26) is rewritten, recall that it is derived in the local beam system:

{S}l = [K]l {u}l + {F }l

Substitutions and orthogonality yield finally:


[K]g {F }g
z }| { z }| {
T T
{S}g = [T ] [K]l [T ]{u}g + [T ] {F }l (30)

Equations (26) and (30) deliver the end forces and moments of a prismatic beam in any position in terms of
its end translations and rotations, in local and global cartesian systems, respectively. It is worth mentioning
that the reverse is not true, since stiffness matrices [K]l and [K]g are singular. Indeed, rigid body motion
cannot be recovered from the end forces.

51
5.2 Equilibrium equations
Equilibrium equations are set up quite analogously to the slope deflection method. Contributions of the
connecting members are added for each joint. In most computer codes this process is implemented in a
loop which runs on all members and for this purpose members are assigned unique indices which appear
as superscripts in this section. This technique has become more or less standard in all finite element codes.
There are three standard equilibrium conditions at each joint and contributions of the members can thus be
assembled in matrix form. The beam stiffness matrix is perceived as composed of four submatrices for the
purpose. " #
e [K s ]ei,i [K s ]ei,j
[K]i,j = (31)
[K s ]ei,jT [K s ]ej,j
where e is the member index. From these submatrices, the global stiffness matrix and load (column) matrix
are assembled. The process is demonstrated on a simple structure in Sketch 69
2 3
2 3 4
1

Global stiffness matrix has 4 · 3 = 12 rows and columns and is gradually


1 filled with submatrices of the members. Corresponding displacements are
not shown in the figure. This is how the matrix looks after beam 1 has been
Sketch 69 Simple frame inserted (left) and when the assembly is complete (right):
without supports, member
indices in circles.

joint 1 2 3 4 joint 1 2 3 4
s 1 s 1 s 1 s 1
1 [K ]1,1 0 [K ]1,3 0 1 [K ]1,1 0 [K ]1,3 0
2 0 0 0 0 2 0 [K s ]22,2 [K s ]22,3 0 All sub-
s 1T s 1 s 1T s 2T s 1 s 2 s 3 s 3
3 [K ]1,3 0 [K ]3,3 0 3 [K ]1,3 [K ]2,3 [K ]3,3 + [K ]3,3 + [K ]3,3 [K ]3,4
4 0 0 0 0 4 0 0 [K s ]33,4T [K s ]34,4
matrices must be transformed to the global coordinate system prior to the allocation in the global matrix.
The load column matrix is assembled analogously from the contributions of the member matrices.
Supports present homogeneous kinematic boundary conditions in the language of the finite element
method and mathematics. Such conditions are implemented in the matrix form of the equilibrium condi-
tions simply by leaving out the corresponding columns and rows of the global stiffness matrix and load
column matrix.
2 3
2 3 4 For the supports in Sketch 70, all three displacement components are re-
1

strained at joint 1, both translations are restrained at joint 2 and horizontal


1
translation is restrained at joint 4. Accordingly, rows and columns 1,2,3,4,5
and 10 are left out from the global stiffness matrix and load column matrix.
Sketch 70 Simple frame
The stiffness matrix shrinks to 6 × 6 size and becomes regular in this step.
with supports

The process described above appears in countless modifications in actual structural analysis codes but
its subtleties are beyond the scope of these notes. Unfortunately, it is also impractical to demonstrate it on
numerical examples, since manual, per matrix element, operations are tedious.

6 Moment distribution (Cross) method


This method was proposed by H. Cross in 1924. It is actually the displacement method in disguise. His-
torically, the method was very popular in times of manual computations, whereas its contemporary value is

52
rather didactic. Valuable allusions and connections to modern methods of structural analysis are found in
the Cross method. The principal idea is best explained on frames with axially rigid beams without sidesway
(no joint translations).
At the start, all joints rotations are restrained. In all beams the end moments equal the respective fixed-
end moments at this state. These end moments are not in equilibrium at joints. The unbalanced moment at
joint i is denoted ∆Mi X
∆Mi = Mij f
j=all neighbours

where Mij f is the fixed-end moment at joint i of member i, j. Next, a (theoretically) infinite sequence of
steps is started. One step is simple. Any joint with imbalance is selected. The joint with greatest imbalance
∆Mi is selected in practice which accelerates the process. The rotation of this joint is released leaving all
other joints restrained. This constitutes a simple task for the slope deflection method with a single primary
unknown.

6.1 Distribution factors by the slope deflection method


Equilibrium requires X
Mi,j = 0
j=all neighbours
Each end moment Mi,j can be split into the flexible and fixed-end parts
e
Mi,j = Mi,j + Mij f
e
Formulas for Mi,j follow from (20) and (22) when the fixed-end moments and beam rotations are omitted.
The former appear separately in the last equation and beam rotations vanish in no sidesway conditions.
Equilibrium condition can be rewritten
X X ef f
e
Mi,j + ∆0 Mi = φi ki,j + ∆0 Mi = 0
j=all neighbours j=all neighbours
(
ef f 2ki,j for joint j clamped
ki,j = (32)
1.5ki,j for joint j hinged
With φi known from this equation the end moments at joint i can be computed for all neighbour members:
ef f
ki,j
Mi,j = ∆0 Mi P ef f + Mij f
j=all neighbours ki,j

The first term on the right hand side is the increment of the end moment owing to the rotation release.
Note that it depends exclusively on the initial imbalance ∆0 Mi and the relative stiffnesses of the neighbour
members
ef f
ki,j
DFi,j = P ef f (33)
k=all neighbours ki,k
which are called distribution factors. Now the end moments are simply
Mi,j = ∆0 Mi DFi,j + Mij f = ∆0 Mi,j + Mij f
The increments of the end moments at joints j of the neighbour joints can be expressed
∆0 Mj,i = ∆0 Mi,j CFi,j (34)
with so-called carry-over factors
(
0.5 for joint j clamped
CFi,j =
0 for joint j hinged

53
6.2 Iteration, demonstration example
The release of joint i implies increments ∆0 Mi,j and ∆0 Mj,i of the end moments. The latter ones modify
the unbalanced moments at neighbour joints j while equilibrium is established at joint i. This state is now
taken for the initial state of the next joint release. Sequential releases of the joints with unbalanced mo-
ments constitute an iteration process. It can be shown that the iteration is always convergent to the correct
distribution of the moments. Prior to starting the iteration, the distribution factors DFi,j need be computed
for each joint i, the fixed-end moments for each member and the initial unbalanced moments ∆0 Mi for each
joint. The increments of the end moments are accumulated until they drop below an appropriate accuracy
criterion. This iteration gave rise to a class of approximate iterative methods for the solution of systems
of linear algebraic equations – the relaxation methods. Relaxation methods are now widely used in many
tasks of mathematics, physics and technology particularly for very large systems of equations. Although the
method is based on the slope deflection method, the joint rotations do not appear in the computations at all
and the user deals exclusively with moments. The Cross method is demonstrated in the frame in Sketch 71.
The same cross-section is adopted in all members for simplicity. It is then
5 4 apparent from equations (32) and (33) that its flexural stiffness EJ cancels
5 out. Distribution factors are computed in the table:
ef f
beam ki,j DF
1.5·2 1.2
1, 2 2.5
= 1.2 4
= 0.3
2·2 2
5

2, 3 2
= 2.0 4
= 0.5
f=12.8 2·2 0.8
2, 4 5
= 0.8 4
= 0.2
1 sum 4.0 1.0
2.5 2
2

3 The fixed-end moment at joint 2 of the beam 1,2 is the single one to be added
at joint 2 so that
Sketch 71 Frame without f 2.52
∆ 0 M1 = − = −10
sideway, Cross method 8
There is no initial imbalance at joint 4. The iteration is best exposed in the
structure scheme in Sketch 72.

5 0.5 4 The distribution factors are inserted in the trunks of the re-
−0.5
0.025

−0.25
spective beams. In more complicated frames the carry-over
0.5
−0.5
1.0

factors can also be inserted, particularly in non-prismatic


beams where they can assume other values than 0.5 or 0. The
seed imbalance −10 at joint 2, beam 1,2 starts the iteration
and is written first. It is distributed according to the factors
0.05
−0.25

0.2
2.0

of joint 2 and increments 3, 5 and 2 of the adjacent end mo-


0.3 2 ments are added in the scheme. Each member end has its own
0. .0

1
12
5 .5

−10.0
5

’account’ where the received increments are cumulated. The


0

3.0
0.075 account files start at the beam trunks and extend away from
them in order to secure enough space for them. One never
0. .5
06
2

knows how many iteration steps will be necessary. Relax-


2

3 ation establishes equilibrium at joint 2 which is indicated by


the foot lines in the adjacent accounts in the scheme. The end
Sketch 72 Frame without sideway, Cross moments increments are then carried over to the neighbour
method, moment distribution diagram joints and written on their accounts.

Joint 4 is released next. The clamped ends obviously are passive recivers of the carry-over moments. The

54
value on the top of the respective account past the last foot line is the current unbalance. In Sketch 72, it is
moment 0.025 at joint 4. When it is sufficiently small, the iteration is finished. Increments in each account
are then added to obtain the end moment at each member end. The Cross method can be extended to frames
with sidesway. Much of its elegance goes then lost, however. Unlike the slope deflection and deformation
methods, it cannot provide further insight in what goes on in contemporary structural analysis programs. Its
basic idea hopefully promotes the understanding of the structures elastic behaviour and can be even helpful
in some specific actual problems of practice when a fast judgment is required on the forces distribution.
Finally, its parenthood is recalled to the relaxation methods.

7 Secondary bending moments in trusses


Bending moments generally occur in real trusses in the wake of excentric member connections, direct lat-
eral member loads and imperfect hinges at joints. Attention is given to the last source in this section. As a
matter of fact, most trusses do not have hinges at all, since it is very difficult to construct a hinge in a steel
or concrete truss. Ideal trusses can still be adopted for their mathematical models, since the members are
slender and the contribution of the bending stiffness to the overall stiffness of the structure is negligible as
compared to the stiffness supplied by the axial stiffness of the members.
4 In order to show the difference, a simple truss is solved with
v4
0.2

3 cross−section:
realistic dimesions. The truss is welded with entirely stiff
.83
01

−0
0.
3

joints so that actually it is a frame. Bending stiffness of the


1

1 0.666 3 2 members is relatively small. The relevant cross-section prop-


v3 erties are
4 4
F=1
A = 2π rd = 2π · 10−3 J = π r3 d = π · 10−5
Sketch 73 Truss with welded joints.

Dimensionally consistent measures of the two stiffnesses are the forces N and V due to unit axial and lateral
translations of the end joint indicated in Sketch 74
4 1
N
1

V EA E2π · 10−3
N= = = E · 1.57 · 10−3
l l
Sketch 74 Axial and
2k · 3ψ 2k · 3 1
lateral unit joint V = = = E · 3.77 · 10−5
translations and end l l l
forces. Axial stiffness is about two orders greater than the bending stiffness.

The frame need not therefore be solved by the slope deflection method. On the other hand, the truss mode
supplies a good approximate solution. The PVf yields (note that actual and virtual stress states coincide)
1 X 1 1 13.48
v3 = δN N l = (2 · 0.08332 · 5 + 2 · 0.6662 · 4 + 12 · 3) = (6.93 + 3.55 + 3) =
AE AE AE AE
N3,4 l3,4 10.48 l1,3 3.66
v4 = v3 − = u4 = =
AE AE AE AE
The analysis of the truss model of the structure is finished. Member forces and deflections should approxi-
mate the exact ones. A comparison is provided at the end of this section.
The deflections obtained in the truss model can be further utilized to assess the bending moments in the
actual frame. The deflections are imparted back on the original frame as imposed displacements. Member

55
rotations are computed first
v3 v 4 · 4 u4 · 3
ψ1,3 = −ψ2,3 = − , ψ1,4 = −ψ2,4 = − − , ψ3,4 = 0
4 5·5 5·5
It remains to determine the joint rotations. In virtue of symmetry φ3 = φ4 = 0. Moment equilibrium at
joint 1 entails
2φ1 (k1,4 + k1,3 ) − 3(ψ1,3 k1,3 + ψ1,4 k1,4 ) = 0
With k = k1,3 and k1,4 = k 4/5, substitutions yield
 
9 3 13.48 10.48 · 4 · 4 3.66 · 3 · 4
2φ1 + + + =0
5 AE 4 52 · 5 52 · 5
4.21
φ1 = −
AE
and
k k J
M1,3 = (−2 · 4.21 + 0.75 · 13.48) = 1.58 = 0.79
AE AE A
k J k J
M1,4 = 2.16 = 1.08 M3,1 = 5.9 = 2.95
AE A AE A
The effects of axial forces and bending moments can be compared by means of stresses. Uniform stress in
member 1,3 due to the axial force alone is
N1,3 0.666
σN = = = 106 kPa
A 2π · 10−3
while extreme stresses due to the bending moment at joint 3 are
r M3,1 r
σM = ± = ± 2.95 = ±47.0 kPa
J A
These bending moments and stresses result from the secondary treatment of the truss deflections and are thus
termed secondary moments and stresses. In the present realistic example, they are obviously not negligible.
The exact solution of the frame is available by the displacement method by means of a commercial
computer code. Young modulus E = 1000 kPa was input to the program and substituting this value in the
above expressions for deflections, some relevant quantities are compared in table 3

Table 3 Comparison to the exact values of the approximate values obtained from the truss model and sec-
ondary moment analysis.

v3 v4 N1,3 N1,4 M1,3 M3,1


approximate 2.14 1.67 0.666 -0.833 0.0054 0.0155
exact 2.128 1.65 0.661 -0.826 0.0045 0.015

The truss model apparently offers good approximations of the deflections and axial member forces. For
many purposes the secondary bending moments also are satisfactory.

56
8 Gridwork frame structures
Gridwork frames are a direct generalization of statically determinate planar frames loaded exclusively lat-
erally (balcony girders) to statically indeterminate frames. In these notes, just frames consisting of straight
beams are considered although in principle curved beams also can appear in gridworks. Only lateral loads
are considered, see Sketch 75. Vectors of the external load couples, if present, must lie in the gridwork plane.

2 With this limitation, grids can also be perceived as sets of planar beams
loaded in their planes which additionally resist and bear torque. Both the
1 φ2,y
force and the displacement methods can be applied to grid frames. The
y φ1,y number of redundants usually is high (twelve in Sketch 75) and the dis-
z placement method generally is preferred. The treatment is thus confined
x
to the deformation method here. Torsional deformation occurs when dif-
Sketch 75 A sample rectangular ferent rotations are imposed around the beam axis at two neighbouring
gridwork with global coordi- joints. An instance is indicated in Sketch 75 for beam 1,2.
nate system

It is apparent that joint rotation in a grid is a vector with two components φix , φi,y in the grid plane. The
displacement of joint i is determined by the two rotation components and vertical deflection wi . These three
kinematic variables are the standard primary unknowns in the displacement method. Rotation vectors φ ~ at
joints of a beam are decomposed into the components t in the beam axis and n normal to it, see Sketch 76.
The positive t axis goes from joint i to joint j. Positive sense of axis n is selected so that t, n and z build a
right-handed coordinate system, the local coordinate system of the beam. Standard transformation can then
be applied from x, y to t, n components and back. Two important restrictions are adopted:

• Beams are prismatic.

• Principal axes of their cross-sections lie in the t − n and t − z planes.

These restrictions imply that bending occurs in the member’s t−z plane only and axial rotation components
φti induce pure torsion. Equilibrium conditions can be assembled with the beam stiffness characteristics
known from the planar frames analysis, with one exception – the beam torsional stiffness must be derived.

57
8.0.1 Beam torsional stiffness
The relative torsional rotation of the two end joints of beam i,j in Sketch 76 is ∆φ = φtj − φti .
The torque expectedly depends on this relative rotation and on the tor-
sional stiffness of the beam. The theory is limited to prismatic beams
φj here and then
nj
φ
 
GJt
Ti,j = ∆φ (35)
) l i,j
n (<ο
t
φi n t j φj
Here GJt is the cross-section torsional stiffness. It is derived in the
theory of elasticity and is the product of the shear elasticity modulus G
y φi and the cross-section moment of torque stiffness Jt . The latter equals the
i
φi
t cross-section polar moment of inertia Jp for circular sections, otherwise
x Jt ≤ Jp . Another symbol
z
gi,j = (GJt /l)i,j
Sketch 76 Torque of a single
beam. is introduced for brevity. Torque finally is converted to the end axial
t t
moments Mi,j and Mj,i .

t
Mi,j = −Ti,j = −gi,j (φtj − φti ) = gi,j (φti − φtj ) (36)

It is easy to show that indices i, j can be swapped in the formula:


t
Mj,i = Ti,j = gi,j (φtj − φti )

Positive senses of all rotations and moments in equations (36) is from i to j.

There are two components of end moments at the end of a gridwork


t
beam, bending moment Mi,j and torque Mi,j . Notation is shown
in Sketch 77.

Sketch 77 End moment notation


in isometric view

8.1 A simple rectangular gridwork


Manual assembly of the equilibrium conditions is illustrated in a simple example. The grid frame in
Sketch 78 features one free joint, number 2. The primary unknowns are deflection w and two rotation
components φx and φy of the joint. Joint subscript 2 is omitted for brevity. Coordinate transformation can
be spared in rectangular grids when the local beam axes coincide with the global ones. Care should be given
to proper signs when adding the bending and axial end moments.

58
Flexural and torsional stiffnesses of the cross-section are EJ = 1 and

2
1,
GJt = 0.5 for simplicity. The beam flexural stiffnesses then are k1,2 =
am
be
f 3 0.5, k2,3 = 1. The torsional stiffness is g1,2 = 0.125. The next step
ew

φy
vi

1 view beam 2,3 in the slope deflection or displacement methods paradigm is to set up
2
the expressions for the end forces of beams. The formulas derived in
2
φx
4
w
section 4 are utilized together with equations (36). In order to conform
to the sign convention introduced in section 4, beams should be viewed
Sketch 78 A simple grid frame. against the global x or y axes in order to obtain compatible end moments
from the bending.
All three equilibirum conditions refer to joint 2, the subscript is thus omitted. Moment around the global
x-axis consists of the axial moment from beam 1,2 and the bending end moment from beam 2,3.
t
⊙x : M2,1 + M2,3 = 0

The first moment follows from equation (36)


t
M2,1 = g1,2 ϕx

The second moment can be evaluated with the aid of equation (22) when the beam is viewed against the
global x axis as indicated in Sketch 78.
2
M2,3 = 3/2k2,3 φx − 3/2k2,3 ψ2,3 + 1/8f l2,3

ψ=−w/l w
2 3 Keeping the view direction and noting that positive w is upwards, it turns out that
φx
ψ2,3 = −w/l2,3
Sketch 79 View at
beam 2,3 against
global x axis

Similarly in the global y direction


t
⊙y : M2,3 + M1,2 = 0
M2,1 = 2k1,2 φy − 3k1,2 ψ1,2 ψ1,2 = −w/l1,2
t
M2,3 =0
The end moment at joint 1 completes the formulas for end moments

M1,2 = k1,2 φy − 3k1,2 ψ1,2

Vertical equilibrium reads


↑ : V2,1 + V2,3 = 0
when positive sense of the shear force and view directions in Sketch 78 are taken into account. It is devel-
oped further with the knowledge of the slope deflection method
M1,2 + M2,1 M2,3 f l2,3
V1,2 = V2,3 = +
l1,2 l1,2 2
After substitutions the equilibrium conditions become
2
⊙x : g1,2 φx + 3/2k2,3 φx + 1/8f l2,3 + 3/2k2,3 /l2,3 w = 0

59
⊙y : 2k1,2 φy + 3k1,2 /l1,2 w = 0
! !
6k1,2 1 3 3k2,3 1 5
↑ : 3k1,2 φy + w + k2,3 φx + w + f l2,3 = 0
l1,2 l1,2 2 2l2,3 l2,3 8
The final form of the system after the numbers have been substituted:

w φx φy
0.561 0.75 0.375 1.25f
0.75 1.625 0 0.5f
0.375 0 1 0

It is instructive to identify the shares of the flexural and torsional stiffnesses in the equations. A single
f lex. tors.
z}|{ z }| {
contribution of torque appears in element (2, 2) of the matrix, 1.625 = 1.5 + 0.125. It has a minute effect
on the solution and could be neglected in this specific case. This is quite typical in practice. Simplified
models of gridworks are frequent in which the torsional stiffnesses are neglected altogether.
The joint displacements and internal forces are shown in Sketch 80.

9.3 3.33
3 M:
15.3
1 2 0.68
15.3 19.4 0.36
44

Sketch 80 Joint displacement and rotations components and the internal forces dia-
grams for the simple gridwork Torque 5.90 in beam 1, 2 is not shown.

8.2 General gridworks, matrix formulation


The above manual approach is acceptable for very simple grids. For general grids, application can be made
of the displacement method paradigm, in particular, the localization of individual beam contributions in the
global matrix. For that purpose, the standard stiffness matrix of the displacement method (26) is recalled
here, simplified by leaving out the rows and columns pertaining to axial displacements and forces:

i j
w φ w φ
i Y 6k/l2 3k/l −6k/l2 3k/l
(37)
M 2k −3k/l k
2
Y 6k/l −3k/l
j symm.
M 2k

60
z
y In the context of a general gridwork structure, this
stiffness matrix refers to the rotation components
zl yl φyl φyl in the local coordinate system of each beam as
φyl φ xl indicated in Sketch 81 for beam i . Note, that
x
xl
j φx l components φy , My in the local coord. system
i i have opposite positive sense than joint rotations φ
and end moments M in matrix (37) when x axis
Sketch 81 A sample general gridwork with a global runs from i to j!
coordinate system and a local system of

In order to transform the beam stiffness from the local to global x/y coordinates, matrix (37) is extended to
(in the wake of the above note, elements 3k/l get opposite signs than in matrix (37))

i j
w φx φy w φx φy
2
Y 6k/l 0 −3k/l −6k/l2 0 −3k/l
i
Mx 0 0 0 0 0 0
[Kb ] : (38)
My −3k/l 0 2k 3k/l 0 k
Y 6k/l2 0 3k/l
j Mx symm. 0 0 0
My 3k/l 0 2k

and multiplied by the tranformation matrix [T ] quite analogously as the stiffness matrix of the general
planar frame beam was multiplied in equations 26-29. The submatrices [T1 ] of the transformation matrix
apparently are organized differently now since rotations, not displacements, are subject to transformation:
 
1 0 0
[T1 ] =  0 cos α sin α 

 (39)
0 − sin α cos α

The torsional stiffness of the beam is treated in a very similar way. The stiffness matrix in local coordinates
is, in the contracted and the extended forms ,

Joint i j
w φx φy w φx φy
Y 0 0 0 0 0 0
Joint i j i
Mx 0 GJt /l 0 0 −GJt /l 0
[Kt ] : i GJt /l −GJt /l [Kt ] : (40)
My 0 0 0 0 0 0
j GJt /l
Y 0 0 0
j Mx 0 GJt /l 0
My 0 0 0

It is subject to the same transformation as [Kb ]. Needless to say, matrices [Kb ] and [Kt ] can be added in
local coordinate system and their sum rotated.
Most gridworks are rectangular in the ground plane. The tranformation can then be replaced by proper
localization of the elements of the matrices (37) and (40) in the global stiffness matrix. A gridwork in
Sketch 82 belongs to this category.

61
yg
3 The global coordinate system selected is shown in
4
xl xg xl
the figure as well as the local coordinate systems
1 yl of individual beams. The local φxl rotations can be
3

l
y
yl identified with rotations φ in matrix (37) includ-
xl ing the same positive sense whereas the local φyl
1 2 coincide with rotations φ in (40). Beams 1 and
2 3 are then easily located in the global matrix
Sketch 82 A sample gridwork with rectangular mem-
since their local coordinate system concide with
bers the global one and positive φx is.

The beam numbers are distinguished by circles in Sketch 82 and matrix (8.2). Beam 2 is more intricate
to assemble. When positive rotations in matrix (37) are to agree with positive sense of the global φx rota-
tion, the local xl axis is implied as shown in Sketch 82. Right-handed coordinate system entails then the
indicated sense of the local yl axis, and, finally, the assignement i = 2 and j = 1 of the joint numbers
in (37). For instance, note that [Kj,j ] submatrix of matrix (37) is located to [K1,1 ] submatrix of (8.2). The
difference in respect to [Ki,i ] is solely in the sign of the off-diagonal element which is underlined in (8.2)
for easy reference. This ’manual’ assembly is intricate and prone to errors. In computer codes, the matrices
of individual beams simply are all transformed with the aid of [T ] and the algorithm is much simpler.
1 2
w φx φy w φx φy
6k
l2 1 3k − 6k
l2
Y l
− 3kl 2 0 − 3kl
+ 6k
l2 2
1 2 2

1 3k
2k 1 − GJl t
2
Mx l
0 0 0
1 + GJl t
2
GJt 3k
l
+ l
2
My − 3kl 0 1 0 k
2
2 2k
2
6k
l2 3
Y − 3kl 3k
l
+ 6k
l
2 2
3 2

2k 3
+
2 symmetric Mx − 3kl
3 GJt 0
l 2

3k
2k 2
+
My l
2
0 GJt
l 3

References
[1] P. Řeřicha. Structural mechanics 30, statically determinate structures, lecture notes. Czech Technical
University in Prague, 2002.

62
9 Appendix – solved examples
Examples are designed and ordered to assist in the exercise of the force and slope deflection methods
separately. Nevertheless, most of them can also be solved by the other method with benefit for the reader.
The examples on the slope deflection method, with sidesway in particular, can all be solved by the force
method with moderate numerical effort.
Limited precision arithmetics is deliberately used in most examples with 3 valid decimal digits in all
manual solutions. This exposes the effect of the round-off errors. They are omnipresent in all numerical
operations including the computer solutions although they are screened off from the user of the computer
codes.

9.1 Force method


9.1.1 Cantilever with end roller
The structure is solved in section 3.2. It is solved here again with another selection of the primary structure,
see Sketch 83. The force load is dealt with first.
PRIMARY STRUCTURE:
F X

Α l/2 l/2
B
1 1·F l·1·l 1F l·1·l
STEP 1: ∆o = − +
L.C.1 − real L.C.2 − virtual EJ 3·4·2·2 2 · 4 · 2 · 2·
F δ1
!
1F l·1·l 1 F l2
+ =−
Fl/4
6·4·2·2 16 E J
STEP 2: 1·1·1·1l l
L.C. 2 − real δ= =
The virtual L.C. is the same 3 EJ 3 EJ
1 as in STEP 1.
∆o 3
X=− = F l = Ay
δ 16
Final internal forces diagrams Suppose the bending moment diagram in
M: V: Sketch 83 has been obtained by superposition.
11/16 F
3/16 Fl The shear force diagram then is best derived
with the aid of differential equilibrium relation
5/32 Fl −5/16 F V = dM/dx. A check on the the result is
available – the step in the V diagram must equal
Sketch 83 Cantilever with roller, primary structure, F.
bending moment diagrams in various load cases

Thermal effect is considered next. With reference to section 3.2.2 and Sketch 83
Z l 1 ∆t α
∆o = κt δM dx = 1l
0 2 h
∆t α 3EJ 20 · 0.00001 · 30 · 106 · 0.00208 · 3
X=− =− = −37.4 kNm
2h l 0.5 · 2
For internal forces diagrams see section 3.2.2.
The support settlement is solved with the present primary structure in section 3.2.3.

63
9.1.2 Frame with arch
Solve for the effect of the uniform load per unit of horizontal projection of the arch as indicated in Sketch 84.
The effect of the support settlement s (also indicated in Sketch 84) is considered in a separate load case later
on.
f=1

PRIMARY

5
5

5
STRUCTURE:
r=

r=
φ X
Neglect the effect of normal forces.
Cross-sections: EJ = 1000 every-
5
s=0.01

where, however, for the separate load


5 10
case uniform load in Sketch 84 the
cross-section bending stiffness is irrel-
evant and it is therefore set EJ = 1 for
Mo : 12.5 M1 : 5
simplicity.
5 For the primary structure in Sketch 84,
the top right diagram, the intermediate
results are:

Sketch 84 Frame with arch.

Z π 25 1
∆o = f (1 − cos2 φ) · 5 sin φ · 5 dφ + 2 · 52 · 5 = 417
0 2 3
Z π
δ= 25 sin2 φ · 5dφ = 196.1 X = −1.49
0

The internal forces presented below are the output of an independent analysis by a commercial code:

Sketch 85 Bending moment diagram Sketch 86 Shear force diagram

Differences with respect to the exact values are due to the replacement of the actual circular shape with
a piece-wise straight beam (16 stretches in the semicircle) and the displacement method of solution in
the computer code. The effect of normal forces can only be suppressed but never exactly neglected in the
displacement method, see section 5. The differences are particularly pronounced in the shear force diagram.
Generally, at the centers of the elements the FEM values in diagrams are closest to the correct values. The
rightmost trapezoid in the shear force diagram was removed in order to make place for the reaction at the pin
support. The difference is notable between the shear force value in the last element of the arch (-0.998, see
the leftmost element where the shear force is the same as at the rightmost one) and the horizontal component

64
of the reaction (-1.496). The FEM shear force diagram would have to be interpolated in order to obtain a
better approximation to the correct value at the pin support. The discrepancy testifies that computer results
should be accepted with caution.
Next, the separate effect of the support settlement is considered. The central pin support settles down
0.01 m as indicated in Sketch 84. True cross-section and material properties must be considered in this load
case.

Sketch 88 Support settlement, shear force dia-


Sketch 87 Support settlement, bending moment dia-
gram
gram

9.1.3 Continuous beam

15
Solve separately for the effect of the partial uni-
A B C s=0.01D form load and the support settlement indicated
10 10 10 in Sketch 89. The beam is prismatic with EJ =
2 · 104
Sketch 89 Continuous beam.

Uniform load effect:


Using three-moment method leads to XB = 25 and
XC = 100. The bending moment diagram is the
output of a commercial computer code

Support settlement:
sEJ sEJ
XB = −2.4 = −4.8 XC = 3.6 = 7.2
l2 l2
Internal forces diagrams are omitted for being trivial.

65
9.1.4 Roof truss
f=1 2
PRIMARY STRUCTURE
3 3 4

2.83
The same axial cross-section
3 5
stiffness AE = 1 ap-
plies to all members of the

4
2 2
X1 truss in Sketch 90. The
1 2 6 ’whole structure’ approach is
X2 adopted to account for the
4 4 4
symmetry. Since some mem-
bers cross the symmetry axis,
FINAL N:
column × is introduced in

41
41

N1: the table in order to assess

.
0
.

−1
N o:
−1

0.

−7 −7 −7
the occurence of the beams.
5.

Load case N2,i,j due to X2 =

0.322
66

66

66
0
.

55
−5

0.
−4

.
41
1

−5 1 differs from N1,i,j only in


−1

−0
.4
66

2
−1

−0
.

.
41
5.

45
4 −1 −0.107 0.215 member force N2,1,2 = 0 and

5
6 it is thus not shown.

Sketch 90 Roof truss with two redundants

member l/(EA) × No N1 N2 ∆o,1 ∆o,2 δ1,1 δ2,2 δ1,2


1, 2 4 2 4 0 −1 −32.0 0 8.0 0 0
1, 3 5.66 2 −5.66 0 0 0 0 0 0 0
2, 3 4 2 −4 1 1 −32.0 −32.0 8.0 8.0 8.0
2, 5 5.66 1 5.66 −1.41 −1.41 −45.2 −45.2 11.3 11.3 11.3
3, 6 5.66 1 5.66 −1.41 −1.41 −45.2 −45.2 11.3 11.3 11.3
3, 5 4 1 −7 1 1 −28.0 −28.0 4.0 4.0 4.0
3, 4 2.83 2 −1.41 0 0 0 0 0 0 0
2, 6 4 1 0 1 0 0 0 0 4.0 0
sums −182.5 −150.5 42.6 38.6 34.6

X1 = 4.1 X2 = 0.242
This particular selection of redundants is not fortunate since the two Xi = 1 stress states are similar to
each other and the compatibility equations therefore are ill-conditioned. This makes the solution sensitive
to numerical round-off errors. Indeed, the correct X2 = 0.215 is about 8% less than the one obtained above.
Rounding-off to three valid digits is not good enough with this primary structure. The final member forces
have been computed with correct redundants.
Next, consider temperature increase t = 20o C in members (1,3) and (3,4).
Solution: No member forces at all.

66
9.1.5 Simple truss

F2 =80 6 No: F2 =80


.7 −6
6.7

1.5
−66
F1 =60 F1 =60 −6.7
5
4

0
75

−85
−40
3 X=0
1 2 3 0 0
2 2

N1: N: 6.7
0 0 −6 −6
6.7
−0.8 22.1
1 −3
6
−18.4

−63.4
−0.6

−0.6

1 39
−0.8 −0.8 28.8 28.8

Sketch 91 Simple truss with multiple member intersections

member l N1 N0 δ ∆o N ∆
1, 2 2 −0.8 0 1.28 0 28.8 −46.0
2, 3 2 −0.8 0 1.28 0 28.8 −46.0
4, 5 4 −0.8 −6.7 2.56 21.3 22.1 −70.7
1, 4 3 −0.6 −40. 1.08 72.0 −18.4 33.2
2, 6 4.5 0 0. 0 0 0 0
3, 5 3 −0.6 −85. 1.08 153.0 −63.4 114.2
1, 5 5 1 75. 5 375.0 39.0 195.2
3, 4 5 1 0 5 0 −35.9 −180.2
4, 6 2.5 0 −66.7 0 0 −66.7 0
5, 6 2.5 0 −66.7 0 0 −66.7 0
sums 17.28 621.3 0
The last column {∆} = {∆o } + X1 {N1 } + X2 {N2 } is used to check the evaluations in the table. Com-
patibility condition requires ∆ = 0. Be aware that this check does not account for the correctness of the
solutions of the elementary load cases and their input in the table. The relevance of the check thus is limited.

9.1.6 Symmetric clamped arch


The elastic axial deformation (effect of normal forces) is neglected in the arch in Sketch 92. Constraints for
the symmetric and antisymmetric load cases are indicated in the two central diagrams.

67
antisymm.:
symmetric:
f f/2
dψ ψ B
f/2
φ primary
X structure
C r C
N A

Sketch 92 Clamped semi-arch and symmetric and antisymmetric load cases

The symmetric load case turns out to be trivial with the axial incompressibily. Normal force N at cross-
section A is obtained from the vertical equilibrium condition:
Z π/2 f fr
↑: N+ r cos ψdψ = 0 N =−
0 2 2
Constant normal force with magnitude N and zero shear force and bending moment constitute the equilib-
rium system with the given external load and the associated deformations (none) meet the support conditions
at the same time. This is thus the solution of the symmetric load case (momentless stress state, see also the
section on arches in [1]).
The antisymmetric load case is statically indeterminate to the first degree, primary structure is shown in
the rightmost diagram in Sketch 92. For the actual load case the reaction Ay can be easily obtained from
the moment equilibrium with respect to C:

C⊙ : Ay r = 0, Ay = 0

Mind that the uniform loading is concentric and does not produce any moment with respect to C. Forces
Ax and B follow from conditions similar to that used above in the symmetric load case,
fr
−Ax = B =
2
Bending moment in the actual load case is:
Z φ f f r2
Mo = Br sin φ − rdψr sin(φ − ψ) = (sin φ + cos φ − 1)
0 2 2
For the virtual load case:

A⊙ : 1 − Br = 0 B = 1/r
Bending moment in the virtual load case:

δM = Br sin φ = sin φ

The virtual work equations yield


Z π/2 f r2 Z π/2
∆o = Mo δM dφ = (sin φ cos φ − 1) sin φdφ =
0 2 0
" #π/2
f r2 φ 1 f r2
 
1 π 1
− sin(2φ) + sin2 φ + cos φ = −
2 2 4 2 0
2 4 2

68
and Z Z
π/2 π/2 π
δ= δM 2 dφ = sin2 φdφ =
0 0 4
Compatibility equation delivers
f r2 π − 2
X=−
2 π

Final forces
 
fr 1 fr 2−π fr
B= +X = 1+ = 0.55
2 r 2 π 2
The final bending moment is

f r2
M = Mo + XδM = (0.55 sin φ + cos φ − 1)
2
Shear force is best computed as the derivative of the bending moment
dM 1 fr
V =− = − (0.55 cos φ − sin φ)
dφ r 2
Normal force:
symm.LC antisymm.LC
z }| { z }| {
Z
fr φ fr fr
N= − + B sin φ − rdφ sin(φ − ψ) = (−2 + 0.55 sin φ + cos φ)
2 0 2 2

M: V: N:
−0.55

−0.45
9
−0

.0

0.

32
−0

32

9
.0

−0

.0
0.
9

.9

−1
1
0.45 0.45 1 1 −1.45 −0.55

Sketch 93 Clamped semi-arch – internal forces diagrams

9.1.7 Symmetric truss with multiple redundants


The truss was solved in section 3.10 for the effect of a horizontal point load. The split symmetric half was
assigned proper constraints at the symmetry axis there and solved separately. Here, the other technique
(’whole structure’ approach) is adopted to utilize symmetry. Symmetric members are not explicitly listed in
l
the table below. Instead, they are acounted for in column AE × by means of doubled lengths of the members
in the left half of the truss.

69
1 f=1 2 2 1
1 No: 2 2 1
3 4 4’ −2 −4

2.
8
X1 X1

−3
1

8
2.

2.
8
1 2 2’ 0
X2
1 1 1

N2: N1:
1 −0.707 −0.7 −1.84

−0
.8
−0.707
−0.707
−1

−0.5

.2
1

−1
−1.7
.4

1
1

1
1

.2
−0
.4
−1

1 −0.707 1.3 2.16

Sketch 94 Symmetric truss, symmetric loading and symmetric primary structure

l
member No N1 N2 AE
× ∆o,1 ∆o,2 δ1,1 δ1,2 δ2,2 N
1, 3 −3 0 −.707 2 0 4.24 0 0 1 −1.7
1, 4 0 0 1 2.83 0 0 0 0 2.83 −1.83
2, 3 2.83 0 1 2.83 0 8 0 0 2.83 1
1, 2 0 0 −0.707 2 0 0 0 0 1 1.3
3, 4 −2 0 −0.707 2 0 2.83 0 0 1 −0.7
2, 4 −4 1 −0.707 2 −8 5.66 2 −1.414 1 −0.54

2, 4 2.83 −1.414 0 1.414 −5.66 0 2.83 0 0 −0.21
4, 2′ 2.83 −1.414 0 1.414 −5.66 0 2.83 0 0 −0.21

4, 4 −4 1 0 1 −4 0 1 0 0 −1.84

2, 2 0 1 0 1 0 0 1 0 0 2.16
sum −23.3 20.7 9.66 −1.414 9.66

X1 X2 load term
X1 = 2.16
9.66 −1.414 23.3
X2 = −1.83
−1.414 9.66 −20.77

9.2 Slope deflection method


Some solutions have been obtained by a commercial computer code using the displacement method. The
exact slope deflection method solutions can merely be approximated in this manner. Nevertheless, at least
three decimal digits of the results are correct. Another consequence of the displacement method is that
normal forces are always determined. Recall that it is not true in the slope deflection method. The computer
results possess their own graphics look that cannot be amended with a reasonable effort. Nevertheless, the
diagrams can well be used to check the reader’s own results.

70
9.2.1 Frames without sidesway
Experience testifies that the crucial issue in the slope deflection method is the identification of the primary
unknowns and corresponding equilibrium equations. Identify the primary unknowns and set up the equi-
librium conditions for the structures in the figure below. The equilibrium conditions should be written in
terms of joint rotations, beam stiffnesses and fixed-end moments.

3 4 F 5 3 4
F
5 3 4 F 5
A B C
1 2 1 2 1 2
3 4 F 5 3 4 F 5 3 4 F 5
D E F
1 2 1 2 1 2
f
A φ4 , φ5 2φ4 (2k1,4 + 2k2,4 + 2k2,5 ) + φ5 k4,5 + M4.5 = 0, 2φ5 (k4,5 + k2,5 ) + φ4 k4,5 = 0
f f
B φ4 , φ5 , φ4 (1.5k1,4 + 1.5k3,4 + 2k4,5 )φ5 ( k4,5 + M4,5 = 0, 2φ5 (k4,5 + k2,5 ) + φ4 k4,5 + M5,4 =0
f f
C φ4 , φ5 φ4 (+1.5k3,4 + 2k4,5 ) + φ5 k4,5 + M4,5 = 0, 2φ5 (k4,5 + k2,5 ) + M5,4 =0
f f
D φ4 , φ4 (2k1,4 + 2k3,4 + 1.5k4,5 ) + +M4.5 − 0.5M5,4 =0
E Structure with sidesway
F Structure cannot be solved by the slope deflection method.
The solution without preliminary elimination of the rotations at hinges, with exclusive use of equation (20).
A See above.
B φ1 , φ3 , φ4 , φ5 2φ1 k1,4 + φ4 k1,4 = 0, 2φ3 k3,4 + φ4 k3,4 = 0
f f
φ4 (2k1,4 + 2k3,4 + 2k4,5 ) + φ1 k1,4 + φ3 k3,4 + M4,5 = 0, 2φ5 (k4,5 + k2,5 ) + φ4 k4,5 + M5,4 =0
f
C φ3 , φ4 , φ5 , φ3 k3,4 + φ4 k3,4 = 0, φ3 k3,4 + φ4 (2k3,4 + 2k4,5 ) + φ5 k4,5 + M4,5 = 0,
f
φ5 (k4,5 + k2,5 ) + M5,4 = 0
f f
D φ4 , φ5 , φ4 (2k1,4 + 2k3,4 + 2k4,5 ) + M4,5 = 0, 2φ5 k4,5 + φ4 k4,5 + M5,4 =0
Frame 1
5 4
5
The frame is solved by the Cross method in section 6. Note that the
inclinations of the members are not relevant in frames without sidesway.
The same cross-section bending stiffness is used for all beams, EJ = 1.
The primary unknowns:
5

f=12.8 φ2 = 1.571 φ4 = −0.434


1
2.5 2
2

71
M:

Frame 2

M:
1
11111111111111111111111111111111
00000000000000000000000000000000
00000000000000000000000000000000
11111111111111111111111111111111
1 2

3
EJ=1

3 5

2
4
2 3 2

V:

V:
Deflections – primary unknowns:
Constant EJ = 1 everywhere,
72

φ2 = −0.293,
N:

N:
φ3 = 0.178
9.2.2 Frames with sidesway
Frame 1

3 4
1 Constant EJ = 1 everywhere,
Deflections – primary unknowns:

5
φ4 = −2.175 u3 = u4 = 4.35 ψ1,3 = ψ2,4 = −0.87
1 2
5 2
M: V: N:

Frame 2, uniform horizontal load.


EJ = 1.5 everywhere, then k2,4 = 1 = k, k1,3 = 1.5k.
0.558 N: 0.4 −0.222
4 −0.335 V:
M:
1

0.221 0.125
3 0.6

1 1
2

1 2 0.05 −0.05 0.4


2 2.6 3.2 1.2

Sketch 95 Simple frame, uniform load per unit of vertical projection


The equilibrium equation with subsequent substitutions:

−V1,3 − V2,4 + 3 · 1 = 0

M1,3 M2,4
− − −1·1+3·1=0
2 3
22
     
1 3 3 1 3 2 1
− − k ψ1,3 − − k ψ1,3 + 21̇ − 1 · =0
2 2 2 3 2 3 2 8
42
ψ1,3 = − = −1.2
35

73
Normal forces in the columns and normal and shear forces in the sloped roof girder need be computed from
suitable equilibrium conditions. The reader is advised to carry out these computations carefully.
Thermal effect in the structure in Sketch 95.
4
M:

1
Combined effect of homogeneous temperature increase t =
3
30o C and temperature gradient ∆ t = 20o C is considered in
40oC
the left column of the structure (the outer surface 40o C, the
20oC inner surface 20o C and the reference stress free temperature
2

0o C).
1 2 18 27
2
Frame 3, force and thermal effects.
EJ = 1 everywhere, α = 0.00001
4 5
t=

10
20

Independent primary unknowns:


oC

1 3
2 φ5 , φ2 , ψ4,5
10
10 10 10 Beam rotations ψ2,3 and ψ1,5 depend on ψ4,5 .

Sketch 96 Frame with sidesway and in-


clined beam
Results for the load point force:

φ5 = −10.9, φ2 = 10.3, ψ4,5 = −9.33

Reactions: Bending moments:

Results for the load case thermal effect:

φ5 = −8.39 · 10−5 , φ2 = −3.18 · 10−5 , ψ4,5 = 2.02 · 10−3

74
Bending moments:
Reactions:

9.3 Displacement method

1 3
Constant EJ = 6.52 everywhere,
Cross-section 0.98 × 0.98, A = 0.96, EJ = 0.5, AE = 6.25
1.5

xg
α3,1 EA
= 2.5,
2EJ
= k = 0.4
1 xl 2 l l

2 2
Stiffness matrix of element 3, 1 in the local coordinate frame (matrix of element 3, 2 is the same):
 
2.5 0 0 −2.5 0 0

 0.3840.48 0 −0.384 0.48 

 
 0.8 0 −0.48 0.4 
[K]e3,1,l = 

 2.5 0 0 

 
 symm. 0.384 −0.48 
0.8

Rotation ϕ1 can be eliminated from the condition M1,3 = 0. It shows in matrix [K]e3,1,l as the elimination of
the last column with the aid of the last row (the process is called condensation in the finite element method
context):  
2.5 0 0 −2.5 0
 


0.097 0.24 0 −0.097  
[K]e3,1,l (condensed) =   0.6 0 −0.24  

 symm. 2.5 0 

0.097
The two last columns and rows of the condensed matrix can be dismissed since the associated displacements
at joint 1 are constrained:
 
2.5
0 0
[K]e3,1,l (condensed and stripped) = 
 0.097 0.24 

symm. 0.6

The two steps (condensation and stripping) can be carried out in reversed order, too. This local stiff-
ness matrix is valid for beam 3, 2 as well since the same supports (boundary conditions) apply (that is,

75
[K]e3,2,l (condensed and stripped) = [K]e3,1,l (condensed and stripped)). The transformation matrix for
beam 3, 1:  
−4 −3 0
1
[T ]3,1 =   3 −4 0 

5
0 0 1
   
−2 −1.5 0 1.635 1.154 0.144
[K]e3,1,g T   
(cond. and strip.) = [T ]3,1  0.0582 −0.077 0.24  =  0.962 −0.192 

0.144 −0.192 0.6 symm. 0.6
This stiffness matrix is valid for the translation components of joint 3 in the global coordinate system and
its rotation. The transformation matrix for beam 3, 2 is
 
4 −3 0
1
[T ]3,2 =  3 4 0 
5
0 0 1

Transformation of [K]e3,2,l (condensed and stripped) would deliver the other stiffness matrix. The assem-
bly simplifies to a simple sum of the two matrices in this particular example. Instead, symmetry of the
structure is utilized. Note that the actual load case is antisymmetric. The split technique replaces the actual
structure with its symmetric half.
0.5 The stiffness matrix of element 3, 1 is sufficient for the solution. Owing to the
constraint at joint 3, the respective column and row of the matrix can be
3 vertical
dismissed. The final matrix and the right hand side of the equilibrium equations
read:
1.635 0.144 0.5
1 0.144 0.6 0

with solution u3,x = 0.312, ϕ3 = −0.075. It turns out that bending moments and shear forces vanish.
It can be deduced immediately from the diagram of the antisymmetric load case since the symmetric half of
the structure is statically definite. The bending moments and shear forces obviously vanish in this structure.

9.4 Gridwork
The two degrees of freedom constrained at (support) joint 3 are indicated with an arrow and double arrow.
The global coordinates axes directions are shown, too.

10
1 5
4
2 2 3 3 V :
4 5
z
x y

76
T : M:

77
Index
beam of freedom, 3, 75
axis, 3 of indeterminacy, 35
continuous, 30–32 displacement, 3
rotation, 36, 39, 45, 52 distribution factors, 52
stiffness matrix, 50
beam, continuous, 6, 11 end moments, 35–37, 40, 41, 43–45, 52, 54, 57
Betti’s law, 21 fixed, 36, 45, 52, 53, 58, 70

carry-over factors, 53, 54 finite element method, 24, 32, 35, 48, 51, 74
Castigliano’s theorem, 33 finite element solution, 63, 69
compatibility
homogeneous temperature change, 9, 13, 26, 27, 45,
equation, 5, 9, 21, 25, 26, 34, 65
73
graphic check, 16
compatibility, check, 15 imposed displacement, 10, 34
compatibility, equation, 20
compliance, 7, 10, 20, 21, 28, 30, 35 kinematic
comptability boundary conditions, 51
equation, 35 conditions, 39
computer, 3, 24, 32, 35, 46, 48, 51, 56, 61–64, 69 load, 23
conjugate quantities, 11, 12, 49 quantities, 46, 48, 49
convention variables, 56
graphic, 3
labels of joints and members, 3, 14
in moment diagrams, 3
lattice structures, 3
sign, 10, 35, 48, 57
live load, 4
coordinate system
cartesian, 3 Maxwell’s law, 21
default, 3 Mohr, 35
local, 50
local-global, 50, 51, 59, 60, 74 normal force
right-handed, 56, 61 positive, 5
couple, 37, 56 unique/indeterminate, 38, 40, 69
Cross method, 52, 70 virtual, 13, 17
cross-section normal forces
depth, 10 effect on redundants, 12, 16, 23, 25, 33, 38, 63,
dimensions, 9 66
principal axes of, 56 notation, 7, 8, 20, 21, 23, 36, 37
properties, 54, 63, 65, 70
potential energy, 33
stiffness, 63
primary unknowns, 59
torque stiffness, 57
common, 50
torsional stiffness, 57
identification of, 47
variable, 30
independent, 3, 40, 45, 47, 73
cross-section stiffnes
principle, 3, 7, 8, 32, 42
axial, 18
of least work, 34
curvature, 10
of virtual forces (PVf), 7, 8, 18, 20, 32, 33, 55
positive, 10
degrees redundance, 4

78
relaxation methods, 53
rigid body, 11, 13, 36, 44, 50, 52
roller, 3, 9, 24, 29, 40, 46, 62

secondary moments, 35, 54, 56


shear force
effect on deflections, 8
sidesway, 35, 41, 47, 52, 54, 62, 70, 72
split structure technique, 23, 24, 42, 75
stiffness matrix
beam, 51, 59, 60, 75
global, 51, 60
strain energy, 33
superposition, 6–8, 17, 20, 32, 50, 62
support settlement, 4, 10, 21, 31, 32, 45, 46, 62–64
system matrix, 43, 45
ill-conditioned, 48, 65

temperature gradient, 9, 10, 13, 36, 73


thermal effects, 9, 10, 20, 73
torque, 56, 57, 59

units, 3

variation, 33
virtual displacement, 11, 44, 45, 47, 48
virtual forces, 7, 8, 15, 33
virtual load case, 7, 8, 26, 67
virtual stress state
eligible, 19
virtual work, 14, 17, 23, 33, 42, 44, 46, 67

whole structure technique, 23, 24, 40, 42, 65, 68


Williot-Mohr diagram, 15, 16

79

Potrebbero piacerti anche