Sei sulla pagina 1di 3

Numerical Differentiation

 +    +  


   = lim



o Consider  by Taylor series expansion


  +
 +  =  +    +
+ -. 
2!
  +
 +   =    +
+ -. 
2!

  + 2! + -. 


 +  
=    +



Forward difference approximation


Truncation error is 5

Similarly

  +
-. 
2!
  +
   =    +
-. 
2!

  =     +

 

  + 2! -. 


  
+
=

  

  

  

 +   
2

5

Backward approximation
Truncation error is 5
Centered difference approximation
Truncation error is 5+ 

For  or 2nd derivative


  

 + 2 2 +  + 


+

Forward difference approximation


Truncation error is 5

  

 2  +  2


2

Backward difference approximation


Truncation error is 5

 

 +  2 +  

Centered difference approximation


Truncation error is 5+ 

 +  

5

Page

December 16, 2014

4.2 Error Propagation


1. For  suppose = is an approximation of .
Then by linear approximation:
 =   = =

December 16, 2014

= = | =| |=|| =|

= |=|=

Page

BC. = ||


2. For D , + ,  
Suppose =D , =+ , =  are approximations for D , + ,  
Then:
G
G
G
=D , =+ , =  F
F =D + F
F =+ + + F
F =
GD
G+
G
3. A condition number is define as the ratio of the errors
  = =
K=   = =
=
 =L
=
=
a. If the condition number >1 then the relative error is amplified
b. If the condition number <1 then the relative error is attenuated/reduced
c. If the condition number =1 then the relative error remains unchanged
** does not happen on real life
4.3 Total Numerical Error
The total error in numerical approximation consist of the truncation error and round-off error
4.4 Blunders, Formulation Errors, and Data Uncertainty
Possible errors in numerical calculation
1. Human error: It should be 3 +  but I put 3 + U
2. Formulation error
3. Data uncertainty

December 16, 2014

CHAPTER 5: Bracketing Methods


5.1 Graphical Methods
5.2 The Bisection Methods
Assume  is continuous if Z \ < 0
then  has at least one zero in Z, \


a

Steps:
1. Choose lower ^  and upper _  guesses for the root such that the function changes signs over
the interval `^ , _ a.
BC ^  _  < 0
2. An estimate of the root b  is determined by:
_ ^
b =
2
3. Make following evaluations to determine in which sub interval the root lies
a. If ^  b  < 0 the root is in `^ , b a,
Set _ = b , go back to step 2
b. If ^  b  > 0 the root is `_ , b a,
Set ^ = b , go back to step 2
c. If ^  b  = 0, b is the root, terminate the calculation.
Termination criteria and error estimates

kfg
cd = efg fhij
e 100%
g
B cd < cq rstu
cq prescribed error tolerance for Bisection Method
_ ^
also cd =
hij

lmn

_ + ^

Page




Potrebbero piacerti anche