Transcript What is manufacturing?

36. Quality Insurance, Testing, and
Inspection

Product Quality
 Quality Assurance
 Total Quality management
 Taguchi methods
 The ISO and QS Standards
 Statistical Methods of Quality Control;
 Reliability
 NDT
 Automated Inspection
Quality

Continuous Improvement in quality
– Never-ending improvement (kaizen in
Japan)
– Quality must built into a product
Quality; customer satisfaction
customer amazement
 Dr. Deming(1900-1993) in Japan, 1954
 Total Quality Management, TQC

TQM

Defect prevention rather than detection
– It is too late to detect at the end of process
– 100 % inspection
– Only a few cents part can ruin an expensive
product
– Customer satisfaction  if not lost money

Leadership, team work
– Only managers can make things worse
– Eliminate fear, eliminate slogan, quota

Continuous Improvement
Dr. Demming
Management must commit to quality
 High quality doesn’t increase cost. Bad
quality actually increase costs
 Break down barriers to workers
(eliminate fear)
 Don’t blame system failures to workers.
 Recognize and Increase workers
potential

Dr. Demming
Recognize pride of workmanship. Avoid
slogans (zero defect), posters,
numerical goals (always increases), and
production quota
 Statistical process control, vendor
provides SPC, JIT
 Teach statistics to workers to improve
quality
 Institute training system

Taguchi Methods
Dr. Demming’s disciple
 Poor quality  customer dissatisfaction
 Costs incurred to service and repair
defective parts
 Credibility diminishes in the market
place
 The manufacturer will lose market share

ISO 9000 standard

1987 (1994 revision), ISO 9000 standard
(Quality Management and quality Assurance
Standard) Statistical Process Control
– ISO 9001 Model for quality assurance in
design/development, production, installation, and
servicing
– ISO 9002 Model for quality assurance in
production and installation
– ISO 9003 Model for quality assurance in final
inspection and test
– ISO 9004 Quality management and quality system
elements-Guidelines
Why Statistics?








Cutting tools, dies, and molds are subject to
wear dimensions vary
Machinery perform differently on its age,
condition and maintenance
Metalworking fluid degrades surface finish,
tool life, and forces are affected
Environment (Temperature, humidity, air
quality) may change
Different shipment of raw material
Operator skill and attention varies
Chance variation (random)
Assignable variation (with specific cause)
Statistical Quality Control






Sample size; the number of parts to be
inspected
Random sampling
Population (universe)
Lot size
The method of variables; quantitative
measurements of dimension, tolerances,
surface finish, physical & mechanical
properties
The method of attributes; Qualitative
Statistical Quality Control

Distribution
– Frequency distribution –e.g. bar charts
– Normal distribution curve (Gaussian)
Arithmetic mean
 Dispersion

– Range R = xmax-xmin
– Standard deviation s =sqrt {S(xi-x0)2/(n-1)}





Manufacturing processes can be judged to be
in control by using statistical measures.
The quality of a product can be measured by
observing attribute values or variable values.
Attributes are discrete measures such as
number of cracks on a surface or number of
defective resistors.
Variables are continuous measures of a
characteristic such as length, weight,
hardness, etc.
The statistical quality control techniques differ
for attribute and variable measures.

The discussion that follows concerns
statistical quality control based on variables
 Two basic questions that a statistical quality
control program can answer are:
 1. Has the average value of a product
characteristic remained within acceptable
bounds?
 2. Has the variability of a product
characteristic remained within acceptable
bounds?

Being able to answer yes to one of these
questions does not necessarily affirm the
other.
 To answer the first, an x chart_can be used
and for the second question, a R chart. Both
of these charts utilize the confidence interval
concept that has been presented earlier.
 Both use a sequence of samples that are
taken over a period of time in order to provide
the evidence needed to answer these
questions.

Statisticians have also developed methods of
establishing these confidence intervals that
use simple calculations based on prepared
tables.
 What follows is a presentation of the methods
without providing the statistical arguments to
justify their use.
 Another concept that is common to both of
these charting techniques is that one first has
to establish the confidence intervals that
represent the process when it is operating
satisfactory (in control).

Some degree of good judgment, process
knowledge, and historical information is
needed in developing these "in control"
criteria.
 The methods that are presented below are
based on the premise that the process is in
control and that samples from the process
can be used to establish these "in control"
confidence intervals.
 To proceed on this basis, the sample size has
to be pre-established and continuously used
during later process monitoring.
SPC

If a machine is not in good condition,
manager can’t blame workers for bad
products find reason and fix it from SPC
 Control charts
– Sample size from 2-10 (sample size held constant
throughout the inspection)
– Frequency of sampling; case by case

Control limits; average value
– UCL=x0+3s = x0+A2Ř where Ř is the average of R
– LCL=x0 - 3 s = x0 -A2Ř

Let n be the size of each sample. Let m be
the number of samples that are collected
during the "in control" period of time. For
each sample, compute the mean and range,
(maximum value - minimum value).
 The mean is going to be used to evaluate
average performance and the range will be
used to evaluate process variability.
 The range can be statistically correlated to
the standard deviation, and is much easier
and faster to compute
SPC

Control limits; average value
– UCL=D4 Ř
– LCL= D3 Ř
s = Ř/d2
 In good statistical control; inside the boundary
 Real-time SPC; computer system with
electronic measurements
 Process capability; limits within which
individual measurement values resulting from
a particular manufacturing process normally
be expected to fall when only random
variation is present.

Constant for Control Charts
S.S
2
A2
D4
1.880
3.267
3
1.023
2.575
4
0.729
2.282
5
0.577
2.115
6
0.483
2.004
S.S. Sample Size
D3
0
0
0
0
0
d2
1.128
1.693
2.059
2.326
2.534
Example

Measuring the length of machined
workpieces. Sample size 5, sample number
10, so total 50 parts
 X0=44.296/10=4.430 in
 Ř=1.03/10=0.103 in
 A2=0.577, D4=2.115, D3=0 (from sample size
5)
–
–
–
–
–
UCL=x0+A2Ř=4.430+0.577*0.103=4.489
LCL= x0 -A2Ř =4.430-0.577*0.103=4.371
also
UCL=D4 Ř =2.115*0.103=0.218 in
LCL= D3 Ř =0*0.103=0 in
s = Ř/d2 =0.103/2.326=0.044 in
Samplex1
sizex2
1 4.46 4.40
2 4.45 4.43
3 4.38 4.48
4 4.42 4.44
5 4.42 4.45
6 4.44 4.45
7 4.39 4.41
8 4.45 4.41
9 4.44 4.46
10 4.42 4.43
x3
4.44
4.47
4.42
4.53
4.43
4.44
4.42
4.43
4.30
4.37
x4 x5
xave
4.46 4.43 4.438
4.39 4.40 4.428
4.42 4.35 4.410
4.49 4.35 4.446
4.44 4.41 4.430
4.39 4.40 4.424
4.46 4.47 4.430
4.41 4.50 4.440
4.38 4.49 4.414
4.5 4.49 4.436
Average of average
4.430
R
0.06
0.08
0.13
0.18
0.04
0.06
0.08
0.09
0.19
0.12
0.103
Acceptance Sampling and Control

1920s, WW II, MIL STD 105
 If a certain % is exceeded, the whole lot is
rejected
 Probability; relative occurrence of an event
 Acceptance Quality level (AQL)
– 95% probability of acceptance
– Consumer knows that 95% acceptable
(consumer’s risk)
– Producer’s risk; good parts are rejected (5%)

Rejected lots are salvaged; greater cost
Reliability, Testing and Inspection

Reliability; the probability that a product will perform
its intended function in a given environment and for a
specified period of time without failure.
– Series reliability
– Parallel reliability; back-up system, redundant system

Non-destructive testing (NDT)
– Liquid penetrants technique
– Magnetic-particle inspection; apply fine ferromagnetic
particles (sometimes dyed) on the surface, then magnetized.
Flaws can be seen
– Ultrasonic Inspection; put into couplant(water,oil, glycerin,
grease), 1-25 MHz
– Acoustic methods; pick up by piezoelectric ceramics
– Acoustic Impact technique
Reliability, Testing and Inspection

Radiography; X-ray
– Digital radiography
– Computed tomogrphy
Eddy-current Inspection; using
electromagnetic induction
 Thermal inspection; heat sensitive
paints, papers, liquid crystal
 Holography

– Holographic interferometry
– Acoustic holography
End of Ch 36 Quality