1 of 36

Control Charts for Attributes

TECH 4462

2 of 36

See also

ASTM E2587

ANSI/ASQ B1-B3

3 of 36

Attribute Data

“observed values or test results that indicate the presence or absence of specific characteristics or counts of occurrences of events in time or space.” ASTM E2587

4 of 36

Why Use Attributes

Where measurement is not possible

  • Missing component or feature
  • Color
  • Scratches, dents, or other damage

Where measurement is possible but is expensive or does not offer additional information.

  • Inspection with Go/No-Go gage
  • Leakage rates (often pass/fail)

5 of 36

Nonconforming Vs. Defective

Nonconforming Vs. Nonconformity

Nonconforming compares an object to a specification

Defective compares an object to an intended purpose

The whole item can be either conforming or nonconforming,

The object may have one or more nonconformities on it.

6 of 36

Limitations of Variables Control Charts

Cannot be applied to attribute data. Attributes cannot be converted to variables,

but

Variables can be converted to attributed data.

by

comparing to specification limits.

But

there is a loss of data that may have been helpful.

7 of 36

Types of Attribute Charts

Binary distribution

  • p, proportion chart
  • Proportion conforming on nonconforming (expressed as a fraction or %)
  • np, number conforming chart
  • Number of conforming or nonconforming units

Poisson distribution

  • Used for nonconformities
  • c, chart - count of nonconformities where subgroup size is one.
  • u, chart - count of nonconformities where subgroup size changes

8 of 36

The p Chart

Used for subgroups consisting of the fraction occurrence of an event (nonconforming).

p = proportion or fraction nonconforming in the sample or subgroup

n = number in the sample or subgroup

np = number nonconforming in the sample or subgroup

9 of 36

Example

During first shift, 450 inspections are made of book shipments and 5 nonconforming units were found. Production during the shift was 15,000 units. What is the fraction nonconforming?

What about the 15,000 units?

10 of 36

Objectives of p Charts

Determine average quality level

Bring attention to changes in average

Improve quality level

Evaluate performance of operating and management personnel

Suggest places for Xbar and R charts

Determine acceptance criteria of a product before shipment to customer

11 of 36

Constructing a p-chart (Constant subgroup size)

  1. Select quality characteristic
    1. single quality characteristic
    2. a group of characteristics
    3. a part
    4. an entire product
    5. a number of products

Note: data can be “rolled up” into summary charts

Also, charts can be used to evaluate performance of:

An operator, work center, department, shift, plant, or whole corporation.

12 of 36

Constructing a p-chart (Constant subgroup size)

2. Determine the subgroup size and method.

Based on proportion nonconforming and confidence limit desired.� A small proportion of nonconforming units requires high inspection.� A minimum of 50 is suggested starting point.� The following formula offers a precise method for finding subgroup size:

Where:

n = Sample size

p = estimate of proportion nonconforming (use .5 if unknown).

Z⍺/2 = Z value for desired confidence limit (see table in book).

(using normal to approximate binomial)

E = Maximum allowable error in estimate of p

precision * p

13 of 36

Constructing a p-chart (Constant subgroup size)

Notice p is hiding here.

p(1-p)

n

14 of 36

Constructing a p-chart (Constant subgroup size)

3. Collect the Data

At least 25 subgroups (sometimes from historical records)

A good idea is to use a check sheet

Data can be plotted as a run chart until limits can be calculated

For each subgroup, p = np/n

15 of 36

Constructing a p-chart (Constant subgroup size)

4. Calculate Trial Central line and Control Limits

Central line is pbar

pbar is the average proportion nonconforming for all of the subgroups.

n is the size of the subgroups� (not the number of subgroups)

16 of 36

Constructing a p-chart (Constant subgroup size)

5. Calculate Revised Central line and Control Limits

If the process is stable, then pbar can be used as the best estimate of p0

If a subgroup is out of control with assignable cause it should be removed from the calculations.

p0 is the standard reference value for the proportion nonconforming.

17 of 36

Constructing a p-chart (Constant subgroup size)

6. Achieve the objective

Use the chart as a tool to improve quality.

Limits should be recalculated periodically and (hopefully) moved inward.

18 of 36

np/n

average p

First a run chart...

19 of 36

Now with trial limits...

20 of 36

Final chart with outlier (with assignable cause) removed

21 of 36

Notes

If the population nonconforming, ϕ, is known, it is not necessary to calculate trial control limits.

ONLY applies if the nonconformities appear randomly. They must be independent. In some processes, if a defect occurs, it is more likely that another will occur. In this case a p chart is not suitable.

22 of 36

Is the p clear?

p is the proportion nonconforming within a specific subgroup.

pbar is the average proportion nonconforming of many subgroups.

p0 is the standard or reference value of the proportion nonconforming based on the best estimate of pbar. It is used to calculate the revised control limits. It can be specified as a desired value.

ϕ is the population proportion nonconforming. If it is known, it can be used to calculate the limits.

23 of 36

p Chart for Variable Subgroup Size

Not the desired case. Use constant subgroups if possible.

But often used with 100% inspection (so subgroups will likely vary).

Limits have to be calculated for each subgroup.

24 of 36

Notice the UCL and LCL changing in response to the subgroup size.

25 of 36

Minimizing the Effect of Variable Subgroup Sizes

Changing control limits can be confusing. There are two methods to try to simplify the control of this data.

First, keep in mind that with a SMALLER subgroup, the limits are WIDER.

  1. Control limits for an average subgroup size.
  2. Control limits for different subgroup sizes

26 of 36

Number Nonconforming Chart (np Chart)

Easier to understand than p chart.

Subgroup size must be constant. Subgroup size should be shown on chart.

27 of 36

Process capability with attributes

Much simpler than for variables.

It is the percent nonconforming (central line of p-chart).

28 of 36

Control Charts for Count of Nonconformities

Based on poisson distribution:

c Chart - a count of nonconformities (subgroup size is ONE)

u Chart - count of nonconformities per unit (Subgroup size varies)

29 of 36

C-Chart

Count of nonconformities.

Two conditions must be met:

  1. The average count of nonconformities must be less than the total possible count.
    1. e.g., rivets on an airplane.
  2. Occurrences must be independent.
    • i.e, a bad rivet will not make another more likely.

30 of 36

Constructing a c-Chart

  1. Select the quality characteristic.
    1. a single characteristic
    2. a group of characteristics
    3. a part
    4. an entire product
    5. a number of product

Or c-charts can be used to evaluate an operator, work center, department, shift, plant or corporation.

31 of 36

Constructing a c-Chart

2. Determine the subgroup size and method.

The subgroup on a c-chart is one inspected unit.

One airplane, one case, one gross, one canoe.

one case

one canoe

32 of 36

Constructing a c-Chart

3. Collect the Data

33 of 36

Constructing a c-Chart

4. Calculate trial central line and control limits.

34 of 36

Constructing a c-Chart

5. Establish the revised central line and control limits.

excludes assignable cause

35 of 36

Constructing a c-Chart

6. Achieve the objective.

36 of 36

u Charts

c = count of nonconformities in a subgroup.

n = number inspected in a subgroup.

u = count of nonconformities/unit in a subgroup.

u = 5/9 = .556

u = 3/3 = 1

u = 3/6 = .5

u = ⅝= .625

u = 2/4 = .5