Sei sulla pagina 1di 42

Quality

Enterprise
level Manufacturing Manufacturing
Support Support
Systems Systems
Quality
Control
Factory
Systems
level Manufacturing
Systems
Production
System

Material Facilities
Automation and handling
Control Technologies
Technologies

Manufacturing processes and assembly operations

Groover, M. P. “Automation, Production Systems, and Computer-Integrated Manufacturing”


Quality. Definitions

• There are several definitions; we will follow those from ISO 9000
• Modern quality management is oriented to satisfaction of needs and
requirements of clients. Is not limited to inspection of products and
processes
• Regarding to products, two relevant aspects related with quality are:
 Product features (to be efficiently searched and defined to satisfy
clients)
 Freedom from deficiencies (not only defects or malfunctioning)
Product features Freedom from deficiencies
Design, configuration, size, weight Absence of defects
Function and performance Conformance to specifications
Distinguished features of the model Components within tolerance
Aesthetic appeal No missing parts
Ease of use No early failures
Availability of options
Reliability and dependability
Durability and long service life
Serviceavility
Reputation of product producer 
Quality management evolution

Traditional Modern
Quality is focused on client satisfaction. Internal and
Clients are external to the organization. The sales and
external clients are considered: external by company's
marketing department are responsible for relations with
products, internal are departments or individuals inside
customers
the company
The company is organized by functional departments with little
Quality goals are defined at highest levels and are driven
appreciation of interdependence. The loyalty and viewpoint of
by top management. Concurrent and simultaneous
each department tends to be centered on itself rather than on
engineering use to be implemented 
the company
Quality is the responsability of the inspection department. The
quality function emphasizes inspection and conformance to Quality is not just the job of inspection departament. It is
specifications with the objective to eliminate defects. understood that product design has an important influence
Inspection follows production. There is often a conflict on product quality.
between production objective (make products) and quality Inspection of manufactured products is not enough.
control objective (accept only good products). Quality must be built into the product. Production workers
Knowledge of statistical quality control techniques reside only must inspect their own products.
on QC experts of the company. Worker's responsability is Suppliers are involved
limited to following instructions. Managers and technical staff Quality is a process of continuous improvement 
do all the planning. 
Modern quality management
• Combines technology with management
• Total Quality Management (TQM) seeks for client satisfaction through
quality assurance, continuous improvement and the involvement of the
whole organization: individuals, department, managers
Modern inspection and
Management (TQM)
control technologies

Efficient quality system


Technologies:
• Quality Engineering
• Quality Function Deployment (QFD)
• Statistical process control
• Automated inspection of 100 % of production
• Inspection out of line (product and process design) and in line
(manufacturing and clients relation after product delivery)
• Contactless sensors, computer vision, Reverse Engineering (RE),
Coordinate Measuring Machines (CMM)
Process variability and process capability

Process variations may be:


• Random (independent of how well is the process designed, due to human
variability among work cycles, machine vibration, raw materials variability).
Random variations typically form a normal distribution. After long run output tends
to cluster about the mean and process is said to be in statistical control
• Assignable (indicate and exception: defective raw material, human mistakes, tool
failures, equipment malfunctions). Cause the output to deviate from normal
distribution and the process is said to be out of statistical control
Process capability relates to the normal variations in the output when the process is
in statistical control. By definition equals ± 3 standard deviations (PC = µ ± 3σ),
where µ is process mean and σ standard deviation. It is assumed that the output is
normally distributed and steady state has been achieved. With ± 3σ, 99,73 % of the
output is inside limits.
Sampling for statistical process estimation
Population mean (µ) :

;
and population standard deviation (σ) by:

Process capability best estimate for that sample is PC = ̅ ± 3s; where ̅ and s are
the average and standard deviation of the sample. Mean and standard deviation is
rarely known because population must be complete to be measured. Sampling is
used to estimate population behavior.
With design tolerances greater than process capability range, the great majority of
parts should be in accordance with specifications. Is a way to assure quality in
design stage (early before production)
It is possible if tolerances are wide enough (not always possible) or process
boundaries are near enough (narrow dispersion, good manufacturing conditions;
generally expensive process )
Statistics background. Normal distribution
Design dimension and tolerances:10±0,05
µ = 10; σ = 0,012
Normal distribution
35

30

25

20

15

10

0
9,94 9,96 9,98 10 10,02 10,04 10,06

± 6σ (9,92…10,08) 99,9999998; 2∙10‐3 defects/1∙106

With 6σ population limits are wider than design tolerances. Practically


100% of parts are eligible. Is it so magic? Is it only a matter of widening the
amount of ± σ?
Process Capabilities Indexes (PCI)
When design tolerances are specified as being equal to process capability, the upper (U) and
lower (L) boundaries of this range define the natural tolerance limits. The ratio of the specified
tolerance range relative to process capability is known as the process capability index (Cp)

6
An useful index to measure the spread of the population about the average is:

min , ; ;
3 3
Another useful index to measure the location of the target value () respect the average is:
max , ; ;

Design engineers tend to assign dimensional tolerances based on function and performance
(e.g.: fits in joints). Ideally, specified tolerance should be greater than process capability and
design engineers should consider the relation between tolerances and process capability
Process capbility 
Defect rate  Defects per 
index: Tolerance  Comments
(%) million
6
0,333 ± 1 σ 3,17E+01 3,17E+05 Sortation required
0,667 ± 2 σ 4,56E+00 4,56E+04 Sortation required
1,000 ± 3 σ 2,70E‐01 2,70E+03 Tolerance = Cp
1,333 ± 4 σ 6,30E‐03 6,30E+01 Defects are infrequent
1,667 ± 5 σ 5,70E‐05 5,70E‐01 Defects are rare
2,000 ± 6 σ 2,00E‐07 2,00E‐03 Virtually no defects
Graphical representations of PCI

Two populations with nominal dimension and tolerances10±0,05 mm

µ = 10; σ = 0,01279; Cp = 1,3028; Cpk = Cp, Cc = 0


µ = 10,0013; σ = 0,01859; Cp = 0,8964; Cpk = 0,8719, Cc = 0,02727

-3σ -3σ µµ +3σ +3σ


Statistical Process Control (SPC)

• Made to asses process in order to improve quality, reduce variability and solve
process problems.
• There are seven principal methods and tools used in SPC: control charts,
histograms, Pareto charts, check sheets, defect concentration diagrams, scatter
diagrams and cause-effect diagrams
Control charts: Plot over time of statistics computed from a measured of a process.
There are two basic types: for variables (from measurements) and for attributes
(fraction of defects or number of defects in a sample). Data are collected and
organized using data collection sheets.
Control charts for variables represent how average of the sample ( ̅ , called x-bar
chart) varies over time or the range (R, called R chart) to represent the variability of
the process.
Data collection sheets
Data collection sheets should be simple and clear. Are used to collect information
about the product or process. Should show a correct interpretation of the studied
phenomenon. May be for:
• Quantity data: Register the quantity of defects. Data may be classified by
machine, worker, shift
• Measurable data: Weight, concentration, dimension
• Defect location: includes a drawing of the product and data: date, product
reference, department, comments. A graphical code helps to describe defects
and where are located

Defect location graphic example

Scratch
Dent
SPC tools: Control charts

To obtain x-bar charts, 20 samples (5 parts each) are measured. Average and range
of each group of 5 parts are calculated. Grand mean ̿ for the 20 values of ̅ is
calculated and is used as the center line for the x-bar chart. Average of the 20 ranges
is calculated and is used as center line of R chart. Upper and lower limits of charts
are calculated as standard deviation s of the sample or using the following table:

Sample Size ̅  Chart R  Chart


n A2 D3 D4
Upper and lower limits for x-bar chart
3 1,023 0 2,574
̿
4 0,729 0 2,282
̿
5 0,577 0 2,114
6 0,483 0 2,004
7 0,419 0,076 1,924 Upper and lower limits for R bar chart
8 0,373 0,136 1,864
9 0,337 0,184 1,816
10 0,308 0,223 1,777
Statistical tools to asses SPC: control charts,
distribution, histogram, Pareto chart

Mean shift

Displaced
Data

Controlled dimension: 10±0,1 mm (20 samples of 5 measures)


̿ = 10,0014; σ = 0,019; Cp = 1,68; Cpk = 0,43, Cc = 0,35
Dimensioning and tolerancing.
Tolerance stack up and quality assurance
Two methodologies commonly used to calculate tolerances stack up are:
1. Worst Case (WC): Worst case using tolerance limits. More interchangeability, but
higher cost.
2. Root Sum of Squares (RSS): Root of the sum of tolerances squares. Considering
that produced part’s measures will be near the mean value in SPC, requires
symmetrical tolerancing. Less interchangeability, but cheaper. Useful when several
components are included in the chain of dimensions

Other methodologies can predict assembly quality from design (not only tolerance
stack up):
Six Sigma: Requires SPC, skilled labor, precise machinery and tooling. Helpful in
world class companies where manufacturing and design facilities are not always near
Minimum cost tolerance allocation: Uses optimization techniques to assign
components tolerances that minimize their manufacturing cost.
Computer Aided tolerancing (CAT): Use to combine the previous two methodologies
in a software able to compute tolerances stack up in 3D and for all components in
assembly. Runs experiments to consider randomness and interchangeability.
Calculates PCI and contribution of each part of the assembly to tolerances stack up
Dimensioning and tolerancing.
WC vs RSS
• WC considers the extreme conditions. RSS is the square root
of the sum of the square of tolerances

• Parts must be produced in a process under SPC to apply RSS
• ±0,18 tolerance may be used assuming that most of the parts
are grouped near the mean. For those near the extremes,
sortation may be needed in assembly

Pin shifts the same with respect to hole in all directions?

+ ‐ WC RSS
24 0,1 0,01
24 0,1 0,01
0,113 0,013
0 0,313 0,181
Unsymmetrical tolerancing and Mean shift

Hole diameter is expected to have a nominal value of 15 and tolerances ±0,05 (all in mm).
Dimensions from 14,95 to 15,05 are acceptable.
14,95 , , 15,05 , or 14,97 ,, should also be accepted to dimension and tolerance the hole.
Correct?
A drill of diameter 15 mm could be found easier than a 14,97 mm one, but ISO standard not always
recommends a centered dimensioning and tolerancing system for unions requiring fit, e.g.: Ø15
H7/j6; Hole: 15 , ; Pin: 15 ,,
Should we center nominal values to use CPI with centered mean (Cc=0)? Customized tooling?
What are the consequences of mean shift in PCI?
Use symmetric tolerancing as much as possible. Respect fit dimensions (because of its influence in
tooling), consider mean shift in those cases. PCI work for symmetrical tolerances
GD&T to help quality assurance in design
and cost reduction
Dimensioning and tolerancing. Quality assurance
in early stages of design

+ ‐ WC RSS
24 0
24 0
0,113 0,013
0,155 0,024
0,155 0,024
0 0,423 0,247

• Wider error accepted to decrease cost (if acceptable) because of material


condition (bonus tolerance)
• There is no ambiguity in part definition
• Flat faces and cylindrical surfaces axis are now properly related
• Pin shifts the same with respect to hole in all directions now
Statistical dimensioning and tolerancing
Different levels of definition and usefulness

Accepted statistical tolerancing with dimensional and GD&T. Designer must


be sure that manufacturer will make parts in a SPC process
Quality assurance in early stages of design

Controlled spread of
population respect mean

nominal value near to


More than ± 4σ population mean

Completely defined statistical tolerancing


with GD&T. Less sensibility to lack of
communication between designer and
manufacturer, and also to labor rotation.
Precise definition for SPC
More complex assemblies

• RSS is of special interest when several parts must assemble properly (tolerances are
always added).
• GD&T helps to define parts properly. Consider usual lack of communication between
designer and manufacturer in modern global companies. Outsourcing makes it more
complicated
• RSS and GD&T helps to lower costs if design predicts assembly quality and SPC process is
used. Cost issues should be studied due to labor training needs, more expensive machinery
and tooling, well designed parts and processes requirements, reliable suppliers needs, etc.
Machines & tooling
Influence on Process Capability

€ €€€€
http://www.lagun.com.es/
High Performance Manufacturing (HPM)
Some figures

• Very high precision (tolerances between ±0,005 mm and ±0,015 mm


for parts of around 100 kg)
• Cutting speeds and feed 50% greater than other modern CNC
machines (150…400 m/min of cutting speed for steel, up to 60.000 rpm
o spindle spinning, up to 20 m/min of feed)
• Precision ±0,005 mm and repetitiveness ±0,003 mm (60.000 $) to
±0,001 mm and repetitiveness ±0,001 mm (500.000 $)
• Acceleration up to 1,5 g
• Very high frequencies due to rotating tools spinning speed
• Able to machine hardened steel (52…55 HRC)
Tools performance with HPM

V = 20.000 rev/min x 10 mm = 628,3 m/min


f = 20.000 rev/min x 0,8 mm/rev = 16.000 mm/min
M = 1600 N x 5ꞏ10-3 m = 8 Nꞏm
P = 2094,4 rad/s x 8 Nꞏm = 16.8 kW

Images from: (http://www.coromant.sandvik.com/)


Example of real machining devices

• Expensive tooling
• Needed to obtain precision and repetitiveness in batch or GT production
• Increase PCI and productivity (auxiliary time reduced)
• Direct influence in quality

http://www.pdqwh.com/
CAM strategies and knowledge to increase PC

Images from: (http://www.coromant.sandvik.com/)


CAM examples for High performance machining:
https://www.solidcam.com/en-us/videos/cutting-videos/
Shaft manufacturing in a single mounting
Automatic tolerances analysis
with CAT software

Angle with
tolerance

Clearance
0.5 A
B
0.1 A
B

0.1
A
0.3 A B C

0.5 A B
C
3.0 A B C
C

0.25
A
0.25
A

0.3 A B C 0.5 A B
C

0.5 A

0.5 A B
C 0.5 A
0.25 B
0.5 A A
B
Simulation

0.5 A
B

6
0.3 A B C

C
0.5 A B
99.73%
3.0 A B C

0.25

0.000

0.005

0.010

0.015

0.020
A

+0.021
-0.000

H12 (diameter 24) = +0.021 -0.000


Simulation

Ø24.016
Ø24.010
Ø24.005
Ø24.011 6

99.73%

0.000

0.005

0.010

0.015

0.020
+0.021
-0.000

H12 (diameter 24) = +0.021 -0.000


Manufacturing simulation
Manufacturing = 1.2
3..
Assembly simulation

Randomly selected parts


Results: statistics

100
ASSEMBLY1
Date: XX/XX/XXXX Time: XX:XX

Simulations Performed 1000 50


Random Number Seed 1
Analysis Type Pearson
Est. Range Interval 99.73
0
2.8 4.0 5.2
Nominal 4.000
Lower Spec. Limit 2.800
Upper Spec. Limit 5.200

Cp 1.200
Cpk 0.950

Parameter Value 95% Confidence Interval


Mean xxxxxxxx xxxxxxxx xxxxxxxx
Standard Deviation xxxxxxxx xxxxxxxx xxxxxxxx
Variance xxxxxxxx xxxxxxxx xxxxxxxx
Skewnes xxxxxxxx xxxxxxxx xxxxxxxx
Kurtosis xxxxxxxx xxxxxxxx xxxxxxxx

Observed distribution type is NORMAL

Estimated Below Spec. 0 0.000 Pct


Estimated Above Spec. 1 0.200 Pct
Total Out of Spec. 1 0.200 Pct
Results: contributions

0.5 A B
C

0.5 A

B
0.5 A BASE - FACE SPF|3.0|A|B|C
0.25
0.25 A
A
35.6%
0.3 A B C
MOTOR - SHOULDER SPF|0.3|A|B|C

22.1%
0.5 A B
C

0.5 A 0.5 A
ARM - HOLE POS|DIA0.1(M)|A|B|C
B B
13.9%

0.3 A B C BASE - LIP SPF|0.3|A|B|C

0.5 A B 5.4%
C
0.1 A 3.0 A B C
B TABLE - DAT_B PER|0.1|A
0.25 0.1
A A
1.2%

MOTOR - DAT_A FLT|0.25


C
In line inspection

Utillajes típicos de moldeo por


inyección. Geometría compleja
erosionada por medio de varias
secuencias de penetración.
Taguchi method for Quality Engineering
(Genichi Taguchi)
Off-line quality control On-line quality control

Product Process Customer


Production
Design Design relations

Process Customer
diagnosis and service
adjustment
System System Feedback to
Design Design Process product
prediction design
Parameter Parameter and
Design Design correction

Tolerance Tolerance Process


Design Design measurement
and action
Robust design
Process or product parameters are defined to maintain stability under disturbances
that are difficult to be controlled (noise factors)
Kind of noise factors:
1. Among units: random inherent variations due to raw materials, machinery or human factors
variations
2. Internal: Internal of product or process. May be (a) time dependents: wear, raw materials damage,
materials fatigue or (b) operational errors like wrong cutting parameters, programming errors
3. External: Process external variations like climate conditions

Product with robust design
Aircraft flying in stormy or clear weather
Car that starts correctly in cold or warm weather
Emergencies of a hospital that conitnue working after an electric cutoff

Process with robust design
Turning operation producing good surface finish in a wide range of turning speeds
Plastic injection molding that molds a good part despite room temperature and
humidity variation
Tolerance design
Wider tolerances than process limits are possible with more controlled and expensive process for
“tighter” tolerances or with less controlled and cheaper process for “looser” tolerances. Looser
tolerances increase yield manufacturing making cheaper the process (less skilled labor, cheaper
machines and tooling, easier setup). However: worst fit, probably worst functioning and lasting,
more difficult due to less interchangeability, loss on safe operation conditions
According to Taguchi loss occurs when the functional characteristic of the product differs from its
nominal or target value. He defined quality as the loss a product costs to society from the time the
product is released for shipment. It includes costs to operate, failure to function, maintenance and
repair costs, client dissatisfaction, injures caused for poor design. Defective products detected,
repaired and reworked are considered manufacturing costs instead of a quality loss
The loss increases at an accelerating rate as the deviation grows:

Where:
L(x): Loss function
x: quality characteristic of interest (a dimension, for example)
k: proportionality constant
N: Nominal value
Taguchi loss function
Traditional tolerances approach
Loss €
Taguchi loss function

Scrap or rework
cost

x1 N x2 x
Li Ls

http://tube.geogebra.org/student/m100653
Example
A company manufactures a shaft with diameter 100 ± 0,2 mm. The company has
studied its repair records and has discovered that if tolerance limits are exceeded
there is a 60 % of probability of product return to be repaired at a cost of 100
€/product (for the company if product is in guarantee or to the client beyond the
guarantee period). Estimate Taguchi loss function constant k for that product:
± 0,2 mm tolerance limits are symmetric with respect to N, then (x-N)=0,2 (100,2-
100=-(99,8-100)=0,2). The expected loss for L(x) is E{L(x)} = 0,6 (100 €) + 0,4 (0 €)
= 60 €

60 = k (0,2)2; k = 60 / 0,04 = 1500 €

L(x) = 1500 (x – N)2

It tolerance is reduced to ± 0,1 mm with the same constant;

L(x) = 1500 (x – N)2 = 15 €

With half of the initial tolerance, repair cost drops by 100. However, production cost
raises (more expensive machines and tooling, more parts are refused, better skilled
labor)
Six Sigma
General goals: Improve customer satisfaction, high quality products and services,
reduced defects, improved process capability through reduction in process
variations, continuous improvement and cost reduction

Implementation: Teams trained in the use of statistical and problem solving tools
as well as project management techniques to define, measure, analyze and make
improvements in the operations of the organization by eliminating defects and
variations in the process. Teams are empowered by management.

Problem solving approach: Called DMAIC: Define project goals, Measure the
process and asses current performance, Analyze the process to determine causes
of defects and variations, Improve the process and Control the future process
performance

Measuring the Sigma level: Number of defects per million is determined and then
is converted to sigma level. Some indexes commonly used are : Defects per
million opportunities (DPMO) that is the ratio, multiplied by 1ꞏ106, between the
number of defects and the product of number of units and number of opportunities
of defect per unit. Defective units per million (DUPM) that is the ratio, multiplied
by 1ꞏ106, between number of defective units and number of units

Potrebbero piacerti anche