TOSHIBA EL-ARABY

Control Charts For Attributes

Mohammed Mokbil
July 2008
To Accompany Russell and Taylor, Operations Management, 4th Edition, 2003 Prentice-Hall, Inc. All rights reserved.

Course Outline
Session 1.1

Basic Principles of Control Charts

Day 1
Session 1.2

The Control Chart for Fraction Nonconforming

Session 2.1

The Control Chart for Nonconformity

Day 2
Session 2.2

Control charts for Attributes with variable sample size

Session 1.1 :
Basic Principles of Control Charts

Session Objectives :
When You complete this session you should be able to :

Identify or Define :
Quality Process Statistical Process Control Quality Improvement Variation Describe or Explain : Causes of Variation the Basic Concept of a Control Chart How To Choose the Control Chart Type

 Definitions of Quality Quality means fitness for use - quality of design - quality of conformance Quality is inversely proportional to variability.

 Quality Improvement Quality improvement is the reduction of variability in processes and products.

Alternatively, quality improvement is also seen as waste reduction.

 Process :

 Statistical Quality control is Activities undertaken to regulate quality of a product . Product Quality control is the Activities to evaluate and regulate quality following production inspect and reject inspect and reject
Statistical process control is a collection of tools that when used together can result in process stability and variance reduction. Considers a subset of SQC

 The Magnificent Seven :

The seven major tools of SPC are :

1) Histogram 2) Pareto Chart 3) Cause and Effect Diagram 4) Defect Concentration Diagram

5) Control Chart
6) Scatter Diagram 7) Check Sheet

 what are Types Of Data ? In God we trust .... all others must bring data.
-- The Statisticians Creed

We may have lots of data, but . Does it represent the process outputs we are interested in ? Is it representative of our current process ? Can we split it into subsets to aid problem solving ? Can it be paired with process inputs ? Is there operational definitions for how measurements are taken and data recorded ?

 what are Types Of Data ?

1- Attribute (discrete) data : is that which can be counted Examples: Broken or
unbroken?

On or Off?

2- Variable (continuous) data : is that which can be physically be measured on a continuous scale. Examples:
Temperature Weight

Attribute Vs. Variable data

Variable Which type of data ? Length in millimeters Attribute

SMC (standard manufacturing cost)

Number of breakdowns per day Average daily temperature Proportion of defective items Number of spars with concession

Lead time (days)

Mean time between failure

Which is best ?
Variable data should be the preferred type as it tells us more about what is happening to a process. Attribute - tells us little about the process

Variable - gives plenty of insight into the process

 Variation Its everywhere. No 2 things are alike. Variation exists - even if variation small and appears same, precision instruments show differences. Ability to measure variation necessary before can control.

 Variation Its everywhere.

Basically

there are 3 categories of variation in piece part production :

1. Within piece e.g. surface roughness 2. Piece to piece eg. dimensions 3. Time to time different outcomes e.g. morning & afternoon, tool wear, workers tired

 Sources of Variation :

Equipment : tool wear, Vibrations etc

Material : tensile strength, moisture content etc

Environment : temperature, light, humidity etc.

Operator : method, motivation level, training etc

Inspection : inspector, inspection equipment, environment etc

Causes Of Variation : Chance & Assignable

Chance or random causes are unavoidable As long as fluctuate in natural/expected/stable pattern of chance causes of variation which are small . This is in state of statistical control When causes of variation large in magnitude; can be identified, classified as assignable causes of variation. If present, process variation is excessive (beyond expected natural variation) state of out of control assignable cause Example : Body temperature - 36.5oC ~ 37.5oC

Common Causes vs. Special Causes Process in control vs. Process out of control
A process in control.
What management likes. Boring predictability. The same today, tomorrow and every day.

A process out of control.

its interesting & exciting. unpredictable and great for fire fighting .

Not so good for planning through.

 Data Distribution :
DATA CAN BE GROUPED TO PROVIDE EASIER ANALYSIS
Average

Grouped Frequency Dispersion

Dispersion

 Distributions can vary in :

1- Location.
2- Shape. 3- Spread.
Location Spread Shape

Size

Size

Size

MEASURES OF CENTRAL TENDENCY

Mode =Median =Mean Median Mean

Mode

Normal Distribution

Skewed Distribution

MEASURES OF DISPERSION

Range:

The difference between the largest and smallest values.

Variance:

Equal to the sum of the squared deviations from the mean, divided by the sample size.

Standard Deviation:
of the variance

The square root

Rational Subgroups Subgroups or samples should be selected so that if assignable causes are present, the chance for differences between subgroups will be maximized, while the chance for differences due to these assignable causes within a subgroup will be minimized.

As the percentage of lots in samples is increased: the sampling and sampling costs increase. the quality of products going to customers increases. Typically, very large samples are too costly. Extremely small samples might suffer from statistical imprecision. Larger samples are ordinarily used when sampling for attributes than for variables.

Constructing Rational Subgroups

Select consecutive units of production. Provides a snapshot of the process. Good at detecting process shifts. Select a random sample over the entire sampling interval. Good at detecting if a mean has shifted out-of-control and then back in-control.

Consecutive Samples

Random Samples

 What is a Control Chart ?

A control chart is a statistical tool used to distinguish between variation in a process resulting from common causes and variation resulting from special causes.

It presents a graphic display of process stability or instability over time.

Upper control limit Process average Lower control limit
1 2 3 4 5 6 7 8 9 10

Upper Specification Limit

USL
Upper Control Limit UCL

Out of control

Centerline or Average

Lower Control Limit Lower Specification Limit

LCL LSL

Sample number

Histograms do not take into account changes over time.

Control charts can tell us when a process changes

A Process Is In Control If :

No sample points are outside control limits

Most points are near the process average

About an equal points are above & below the centerline Points appear randomly distributed

Typical Out-of-Control Patterns

Point outside control limits Sudden shift in process average Cycles Trends Hugging the center line Hugging the control limits Instability

Zones For Pattern Tests

UCL

Zone A Zone B Zone C

x + 3 sigma x + 2 sigma x + 1 sigma x - 1 sigma

C.L
Zone C
Zone B Zone A
LCL

x - 2 sigma x - 3 sigma

Identifying Potential Shifts

1. 8 consecutive points on one side of the center line 8 consecutive points up or down across zones 14 points alternating up or down 2 out of 3 consecutive points in Zone A but still inside the control limits 4 out of 5 consecutive points in Zone A or B

2.
3. 4. 5.

Identifying Potential Shifts

Shift in Process Average

Cycles

Trend

Hugging the Centerline or Control Limit

UCL

1/3
1/3 1/3 Process Average

LCL

 Control Charts and the normal Distribution :

 Why Use a Control Chart?

To monitor, control, and improve process performance over time by studying variation and its source.

What Does a Control Chart Do?

Focuses attention on detecting and monitoring process variation over time; Distinguishes special from common causes of variation, as a guide to local or management action; Serves as a tool for ongoing control of a process; Helps improve a process to perform consistently and predictably for higher quality, lower cost, and higher effective capacity; Provides a common language for discussing process performance.

Developing Control Charts

1.

Prepare Choose measurement Determine how to collect data, sample size, and frequency of sampling Set up an initial control chart
Collect Data Record data Calculate appropriate statistics Plot statistics on chart

2.

Next Steps 3. Determine trial control limits Center line (process average) Compute UCL, LCL 4. Analyze and interpret results Determine if in control Eliminate out-of-control points Recompute control limits as necessary

Final Steps 5. Use as a problem-solving tool Continue to collect and plot data Take corrective action when necessary
6. Compute process capability

Choice of Control Limits

68.3%

-4

-3

-2

-1

2s

+/- 1 Std Dev = 68.3%

68.3% of data should be within 1 standard deviations of the mean if no special cause variation is present

Choice of Control Limits

95.5%

-4

-3

-2

-1

4s

+/- 2 Std Dev = 95.5%

95.5% of data should be within 2 standard deviations of the mean if no special cause variation is present

Choice of Control Limits

99.74%

-4

-3

-2

-1

6s
0

+/- 3 Std Dev =99.74%

99.74% of data should be within 3 standard deviations of the mean if no special cause variation is present. Control limits are an estimation of 3 standard deviations either side of the mean.

99.7% of the Data :

The use of 3-sigma limits generally gives good results in practice.

If approximately 99.7% of the data lies within 3 of the mean { i.e., 99.7% of the data should lie within the control limits}, then 1 - 0.997 = 0.003 or 0.3% of the data can fall outside 3 {or 0.3% of the data can fall outside the control limits}. Actually, we should use the more exact value 0.0027

The limits are often referred to as action limits.

Control Chart For Attributes Selection

Defect or Nonconformity Data Constant sample size c chart Variable sample size u chart

Defective or Nonconforming Data Constant n > 50 p or np chart Variable n > 50 p chart

Commonly used control charts :

For Variables data

x-bar

(mean) and R- (range) charts x-bar and s- (standard deviation) charts Charts for individuals

(x-charts) (MR-charts) Moving range charts

For Attributes data

For

defectives (p-chart, np-chart) For defects (c-chart, u-chart)

Control Charts for attributes

For defectives

p-chart : Control chart for fraction nonconforming. np-chart : Control Chart for Number of nonconforming.

For defects

c-chart : Control Chart for Nonconformities. u-chart : Control Chart for Average Number of Nonconformities per Unit.

Session 1.2 :

The Control Chart for Fraction Nonconforming

Control Chart for Fraction Nonconforming p Chart

Fraction of Nonconforming is the Ratio of the number of nonconforming items in a population to the total number of items in that population
The Sample Fraction of Nonconforming is the Ratio of the number of nonconforming items in the sample {D} to the sample size {n}

Mean & Variances

Mean Variance

With specified standard p value :

 When p is not known, it must be estimated from collected data Average of these individual sample fractions nonconforming
Fraction Nonconforming control chart: No Standard Given Trial Control Limit

The np Control Chart :

Alternative to p Control Chart Based on the number nonconforming rather than the fraction nonconforming

If standard value p is not known, use the estimator

Development and operation of the control Chart: Example:

Frozen Orange juice is packed in cans formed on a machine by spinning them from a cardboard stock and attaching a metal bottom panel. By inspection of cans it could possibly leak and thus it is nonconforming. We wish to setup a control chart to improve the fraction of nonconforming cans produced by this machine.

Answer:
We will first collect data for trial control limits, With sample size n=50 the following 30 samples data were collected.

data for trial control limits

No. of Sample Sample nonconforming cans, No. No. Di
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 12 15 8 10 4 7 16 9 14 10 5 6 17 12 22 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

No. of nonconforming cans, Di

8 10 5 13 11 20 18 24 15 9 12 7 13 9 6

P=0.2313

Initial fraction nonconforming control chart

We note that two points from samples 15 and 23 plot above the UCL, so the process in out of control. These points must be investigated to see whether assignable cause can be determined. Analysis of the data from sample 15 indicates that a new material was put into production during that half-our sample, it caused irregular production performance. Furthermore, during the half-hour period in which sample 23 was obtained, a relatively inexperienced operator had been temporarily assigned to the machine. Consequently, samples 15 and 23 are eliminated and the new centerline and revised control limits are calculated as :

= Points not included in control limit calculations.

control chart with revised control limits

Now the sample 21 exceeds the UCL . But analysis didnt produce any assignable causes. Therefore, we decided to retain the point. And to use the new control limits for future samples. Sometimes examination of data reveals information that affects other point.

for example : the new operator assigned again to the machine at point 24.
Then we should discard both the two points even if the other point is between control limits.

Before we conclude the process is in control, we must examin the remaining 28 samples for runs.

We find that : the largest run is one of length 5 above the center line. Its Ok.
The process is in control at level P=0.2150 and with the revised control limits. Note : The process is in control , Where the Fraction of nonconforming is too high, but in a stable manner.

That is the Top Management and the Engineering Staff to analyze the process and try to improve the Yield. After the Machine adjustments, the data from the next 3 shifts was colleted as shown in the following table. { 24 samples with n=50 }

No. of Sample Sample nonconforming cans, No. No. Di

31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 9 6 12 5 6 4 6 3 7 6 2 4 3 6 5 46 47 48 49 50 51 52 53 54

No. of nonconforming cans, Di

4 8 5 6 7 5 6 3 5

Continuation of fraction nonconforming control chart

From the last control chart, our immediate impression is that the process may be out of control. But with no reasonable causes, the only logical reason is the machine adjustments made by the engineering staff, and possibly the operators themselves. It seems logical to revise the control limits again. Calculations should be with the most recent samples ( No. 31 to 54 ) . This result in the following chart.

New control limits on the fraction nonconforming control chart

Data for the process during the next five shifts are shown in the following table.
Sample No.
55 56 57 58 59 60 61 62 63 64 65 66 67 68
69 70 71 72 73 74

No. of nonconforming Sample cans, Di No.

8 7 5 6 4 5 2 3 4 7 6 5 5 3
7 9 6 10 4 3

No. of nonconforming cans, Di

5 8 11 9 7 3 5 2 1 4 5 3 7 6
4 4 6 8 5 6

75 76 77 78 79 80 81 82 83 84 85 86 87 88
89 90 91 92 93 94

Completed fraction nonconforming control chart

The control chart should be continued, by marking the time scale of the control chart when a process change is made.

The control chart becomes a logbook in which the timing of process interventions and their subsequent effect on process performance are easily seen.

Control Chart for number of Nonconforming np Chart

Alternative to p Control Chart Based on the number nonconforming rather than the fraction nonconforming

If standard value p is not known, use the estimator p

Revisit

the first data table in the past example. You can find that:
p = 0.2313 n = 50

Therefore, the parameters of the np control chart would be : UCL = np + 3 np(1-p) = 50(0.2313) + 3(50)(0.2313)(0.7687) = 20.510 C.L = np = (50)(0.2313) = 11.565

LCL = np - 3 np(1-p)
= 50(0.2313) + 3(50)(0.2313)(0.7687) = 2.620

Initial number of nonconforming (np) control chart

25 1 20 3.0SL=20.51 1

15 NP=11.57 10

5 -3.0SL=2.621 0 0 10 20 30

Sample Number

Some practitioners prefer to use integer values in control limits instead of decimal values.
In the last example use 2 and 21 as LCL and UCL.

The np chart requires that the sample size of each subgroup be the same each time a sample is drawn. When subgroup sizes are equal, either the p or np chart can be used. They are essentially the same chart.

np Chart properties :

Advantages
np

chart is a scaling of the vertical axis by the constant n, provide the same information as p chart np chart needs less calculation ( no need to calculate Di/ni) often used when n is constant and p is small

Limitations
not

easy for interpretation when n is varied (UCL LCL and CL all vary) only plot of defects without considering sample size, hard to take action