A basic introduction to statistical process control (SPC)

Mostafa El-Abany

Session objectives
To understand the basic principles and concepts of control charts To understand variation and how this links with decision making To understand the concepts of a system process and to recognise and measure for improvement To explore different ways of presenting data To construct and analyse a control chart To understand targets in relation to variation To be able to recognise the need for ongoing data monitoring and management

Why measure for improvement?

Testing small changes on patients how do we know these changes have made a difference? Service Improvements impact on larger number of patients how do we know how much difference the improvement has made and if the improvement has stayed in place sustained So ask yourself:
How do I know if the care being delivered is getting better, staying the same or getting worse each year?

variation is the key to redesign

Variation exists in all processes around us No two people are the same Every patient is different Things always change

Two types of variation

Natural (common cause) variation Is inherent in the design of the process Results in a stable IN CONTROL process because the variation is predictable Is due to random or chance causes of variation Special cause variation Is due to irregular or unnatural causes that are not inherent in a process extrinsic Results in an unstable OUT OF CONTROL process because variation is not predictable Is due to non-random or assignable causes of variation (i.e. a signal that the process has changed)

Sources of variation
Staff
skills

Process
motivation unclear guidelines differ

Patients
age motivation
disease race education

illness holiday shifts training

complications anaesthetics

sex

machines not the same supplies rooms not the same

transcription transport

Kit

Information

If people dont understand variation

They see trends where there are no trends They try to explain natural variation as special events They blame and give credit to people for things over which they have no control They cant understand past performance They cant make predictions or plan for the future Their ability to make improvements is limited

Reducing Variation

Statistical Process Control

SPC
tool for identifying problems and make improvements

Quality Measures
Attribute
a process characteristic that can be evaluated with a discrete response good bad; yes - no

Variable
a process characteristic that is continuous and can be measured weight - length

Applying SPC to Service

Nature of defect is different in services Service defect is a failure to meet customer requirements Monitor times, customer satisfaction

Applying SPC to Service (cont.)

Hospitals timeliness and quickness of care, staff responses to requests, accuracy of lab tests, cleanliness, courtesy, accuracy of paperwork, speed of admittance and checkouts Grocery Stores waiting time to check out, frequency of out-of-stock items, quality of food items, cleanliness, customer complaints, checkout register errors Airlines flight delays, lost luggage and luggage handling, waiting time at ticket counters and check-in, agent and flight attendant courtesy, accurate flight information, passenger cabin cleanliness and maintenance

Definitions

Statistical Formulas

Statistics-Mean
Mean = arithmetic average = X bar (as spoken) =X (as written) =SX n where S X = sum of Xs and n = sample size

Example of a mean value

Find the mean value of: 4,6,7,9,4. X= S X = 4+6+7+9+4 n 5 = 30 5 =6

Statistics Median
Median=middle value Half of the values are greater than the median and half are less than the median, when the values are placed in numerical order.

Example of a median value

Find the median value of: 4,6,7,9,4.
Put the numbers in numerical order: 4, 4, 6, 7, 9.

The middle number is the median:

The median number is 6

Statistics Mode
Mode = most frequently occurring value Find the mode of 4,6,7,9,4
The most popular, or mode is 4

Statistics - Range
Range = spread of the extreme results Example: Find the range of 4,6,7,9,4,2,5,8,7,9,3 The extreme values are 2 and 9. The range R = 9 2 = 7

Standard Deviation
Although the range is an easy way to measure variation, it has a drawback. The range only takes into account the extreme values and disregards all other information. To overcome this, Standard Deviation is used as the preferred measure of spread.

Standard Deviation
Standard deviation is a measure which reflects variability and dispersion of a process. The greater the variability, the greater the standard deviation. Standard Deviation is represented by the symbol sigma s

Standard Deviation
Standard deviation is calculated from:
= S ( X X )/n-1

Scientific calculators have eliminated the need for this complex calculation!

Compare 3 distributions of 9 values

3 2 1 1 2 3 4 5 6 7 8 9 10

R=8 SX = 54 n=9 X bar = 6

R=8

s = 2.74

3 2 1 1
3 2 1 1 2 3 4 5 6 7 8 9 10

SX = 54 n=9 X bar = 6
2 3 4 5 6 7 8 9 10

s = 2.40

R=8 SX = 54 n=9 X bar = 6

s = 2.12

Standard Deviation and Prediction

The distribution most frequently encountered in industry, commerce & nature is the Normal Distribution.
This is important because knowledge of its properties enable us to make predictions about machine and process performance.

Variability
Random
common causes inherent in a process can be eliminated only through improvements in the system

Non-Random
special causes due to identifiable factors can be modified through operator or management action

Common and Special Causes

Common Causes

Special Causes

Natural (common cause) variation

a stable process in control
80 70 60 50 40 30 20 10 0 F M A M J J A S O N D J F M A M J J A S O N D
Upper process limit Mean

Lower process limit

Special cause variation

- an unstable process out of control
90 80 70 60 50 40 30 20 10 0 F M A M J J A S O N D J F M A M J J A S O N D
Upper process limit Mean

Lower process limit

Control Charts

A graph that establishes control limits of a process Control limits upper and lower bands of a control chart

Control Charts (cont.)

Types of charts
Attributes
p-chart c-chart

Variables
range (R-chart) mean (x bar chart)

The Normal Distribution

The Normal Distribution

The normal distribution is the one of the most popular and useful continuous probability distributions.
The formula for the probability density function is rather complex:

1 f (X ) e s 2

( x )2 2s 2

The normal distribution is specified completely when we know the mean, , and the standard deviation, s .

The Normal Distribution

The normal distribution is symmetrical, with the midpoint representing the mean. Shifting the mean does not change the shape of the distribution. Values on the X axis are measured in the number of standard deviations away from the mean. As the standard deviation becomes larger, the curve flattens. As the standard deviation becomes smaller, the curve becomes steeper.

The Normal Distribution

Normal distribution
| | |

40

= 50

60

40

= 50

60

The Normal Distribution

| | |

40

= 50 Smaller , same s

60

= 40

50

60

Larger , same s
| | |

40

50

= 60

2-35

The Normal Distribution

Same , smaller s

Same , larger s

The Normal Distribution

Variability
The world tends to be bell-shaped

Even very rare outcomes are possible (probability > 0)

Fewer in the tails (lower)

Most outcomes occur in the middle

Fewer in the tails (upper)

Even very rare outcomes are possible (probability > 0)

The Normal Distribution Characteristics

Mean at the center Mean=median=mode Unimodal.

Probability and the Normal Curve

The normal distribution is a continuous probability distribution. This has several implications for probability. The total area under the normal curve is equal to 1. The probability that a normal random variable X equals any particular value is 0. The probability that X is greater than a equals the area under the normal curve bounded by a and plus infinity (as indicated by the non-shaded area in the figure below). The probability that X is less than a equals the area under the normal curve bounded by a and minus infinity (as indicated by the shaded area in the figure below). Additionally, every normal curve (regardless of its mean

Z Score
Definition: In statistics, a z-score is a standardized score that indicates how many standard deviations a data point is from the mean. A zscore is calculated by subtracting the population mean from the individual score. This difference is then divided by the population standard deviation. Z-scores have an average of zero and can range from -4 to +4. Scores above zero are considered above average, while those below zero are considered below average.

Using the Standard Normal Table

Step 1
Convert the normal distribution into a standard normal distribution. A standard normal distribution has a mean of 0 and a standard deviation of 1 The new standard random variable is Z
Z X

Where

X = value of the random variable we want to measure = mean of the distribution s = standard deviation of the distribution Z = number of standard deviations from X to the mean,

Using the Standard Normal Table

For example, = 100, s = 15, and we want to find the probability that X is less than 130.

130 100 15
P(X < 130) = 100 s = 15

30 2 std dev 15

| 55 | 3

| 70 | 2

| 85 | 1

| 100 | 0

| 115 | 1

| 130 | 2

| 145 | 3

X = IQ

s
2-44

Using the Standard Normal Table

Step 2 Look up the probability from a table of normal curve areas. Use Appendix A or Table (portion below). The column on the left has Z values. The row at the top has second decimal places for the Z values.
AREA UNDER THE NORMAL CURVE Z 1.8 1.9 2.0 0.00 0.96407 0.97128 0.97725 0.01 0.96485 0.97193 0.97784 0.02 0.96562 0.97257 0.97831 0.03 0.96638 0.97320 0.97882

2.1
2.2

0.98214
0.98610

0.98257
0.98645

0.98300
0.98679

0.98341
0.98713

P(X < 130) = P(Z < 2.00) = 0.97725

Haynes Construction Company

Haynes builds three- and four-unit apartment buildings (called triplexes and quadraplexes, respectively). Total construction time follows a normal distribution. For triplexes, = 100 days and s = 20 days. Contract calls for completion in 125 days, and late completion will incur a severe penalty fee. What is the probability of completing in 125 days?

Haynes Construction Company

Z X

125 100 20

25 1.25 20

From Appendix A, for Z = 1.25 the area is 0.89435. The probability is about 0.89 that Haynes will not violate the contract.

= 100 days s = 20 days Figure 2.11

X = 125 days

2-47

T- Distribution
The t distribution (aka, Students tdistribution) is a probability distribution that is used to estimate population parameters when the sample size is small and/or when the population variance is unknown.

Why Use the t Distribution?

According to the central limit theorem, the sampling distribution of a statistic (like a sample mean) will follow a normal distribution, as long as the sample size is sufficiently large. Therefore, when we know the standard deviation of the population, we can compute a z-score, and use the normal distribution to evaluate probabilities with the sample mean.

Why Use the t Distribution?

But sample sizes are sometimes small, and often we do not know the standard deviation of the population. When either of these problems occur, statisticians rely on the distribution of the t statistic (also known as the t score), whose values are given by: t = [ x - ] / [ s / sqrt( n ) ] where x is the sample mean, is the population mean, s is the standard deviation of the sample, and n is the sample size. The distribution of the t statistic is called the t distribution or the Student t distribution.

Degrees of Freedom
There are actually many different t distributions. The particular form of the t distribution is determined by its degrees of freedom. The degrees of freedom refers to the number of independent observations in a set of data. When estimating a mean score or a proportion from a single sample, the number of independent observations is equal to the sample size minus one. Hence, the distribution of the t statistic from samples of size 8 would be described by a t distribution having 8 - 1 or 7 degrees of freedom. Similarly, a t distribution having 15 degrees of freedom would be used with a sample of size 16.

Properties of the t Distribution

The t distribution has the following properties: The mean of the distribution is equal to 0 . The variance is equal to v / ( v - 2 ), where v is the degrees of freedom and v> 2. The variance is always greater than 1, although it is close to 1 when there are many degrees of freedom. With infinite degrees of freedom, the t distribution is the same as the standard normal distribution.

T Distribution
The t distribution is the sampling distribution of the t statistic. The values of the t statistic are given by: t = [ x - ] / [ s / sqrt( n ) ] where x is the sample mean, is the population mean, s is the standard deviation of the sample, and n is the sample size.

T Distribution (cont.)
The t distribution can be used whenever samples are drawn from populations possessing a bellshaped distribution (i.e., approximately normal). The distribution of the t statistic is similar to the distribution of a standard score . Both distributions are symmetrical with a mean of zero. Also, both distributions are bell-shaped, although the t distribution has a bigger variance.

T Distribution (cont.)
To evaluate probabilities associated with the t distribution, we need to recognize that there are actually many different t distributions. The particular form of the t distribution is determined by its degrees of freedom .

Degrees of Freedom
There are actually many different t distributions. The particular form of the t distribution is determined by its degrees of freedom. The degrees of freedom refers to the number of independent observations in a set of data. When estimating a mean score or a proportion from a single sample, the number of independent observations is equal to the sample size minus one. Hence, the distribution of the t statistic from samples of size 8 would be described by a t distribution having 8 - 1 or 7 degrees of freedom. Similarly, a t distribution having 15 degrees of freedom would be used with a sample of size 16.

Properties of the t Distribution

The mean of the distribution is equal to 0 . The variance is equal to v / ( v - 2 ), where v is the degrees of freedom (see last section) and v> 2. The variance is always greater than 1, although it is close to 1 when there are many degrees of freedom. With infinite degrees of freedom, the t distribution is the same as the standard normal distribution.

When to Use the t Distribution

The t distribution can be used with any statistic having a bell-shaped distribution (i.e., approximately normal). The central limit theorem states that the sampling distribution of a statistic will be normal or nearly normal, if any of the following conditions apply. The population distribution is normal. The sampling distribution is symmetric, Unimodal, without outliers, and the sample size is 15 or less. The sampling distribution is moderately skewed, Unimodal, without outliers, and the sample size is between 16 and 40. The sample size is greater than 40, without outliers.

Notation and t Scores

Statisticians use t to represent the t-score that has a cumulative probability of (1 - ). For example, suppose we were interested in the t-score having a cumulative probability of 0.95. In this example, would be equal to (1 0.95) or 0.05. We would refer to the t-score as t0.05 Of course, the value of t0.05 depends on the number of degrees of freedom. For example, with 2 degrees of freedom, that t0.05 is equal to 2.92; but with 20 degrees of freedom, that t0.05 is equal to 1.725. Note: Because the t distribution is symmetric about a mean of zero, the following is true. t = -t1 - alpha And t1 - alpha = -t Thus, if t0.05 = 2.92, then t0.95 = -2.92.

Example
Acme Corporation manufactures light bulbs. The CEO claims that an average Acme light bulb lasts 300 days. A researcher randomly selects 15 bulbs for testing. The sampled bulbs last an average of 290 days, with a standard deviation of 50 days. If the CEO's claim were true, what is the probability that 15 randomly selected bulbs would have an average life of no more than 290 days?

Solution
The first thing we need to do is compute the t score, based on the following equation: t = [ x - ] / [ s / sqrt( n ) ] t = ( 290 - 300 ) / [ 50 / sqrt( 15) ] = = -10 / 12.909945 = - 0.7745966 where x is the sample mean, is the population mean, s is the standard deviation of the sample, and n is the sample size.

The Poisson Distribution

The Poisson distribution is a discrete distribution that is often used in queuing models to describe arrival rates over time. Its probability function is given by:
P( X )

x e
X!

where
P(X) = = e= X= probability of exactly X arrivals or occurrences average number of arrivals per unit of time (the mean arrival rate) 2.718, the base of natural logarithms specific value (0, 1, 2, 3, ) of the random variable

The Poisson Distribution

The mean and variance of the distribution are both .

Expected value = Variance =

The Poisson Distribution

We can use Appendix C to find Poisson probabilities. Suppose that = 2. Some probability calculations are:
P( X )

x e

X! 20 e 2 1(0.1353) P ( 0) 0.1353 0! 1 21 e 2 2(0.1353) P (1) 0.2706 1! 1 2 2 e 2 4(0.1353) P (2) 0.2706 2! 2

The Poisson Distribution

Sample Poisson Distributions with = 2 and = 4

Process Control System Model

Voice of the process Man Statistical Methods Machine Material Product/ Process Service Method Measurement Environment Voice of the customer

Customer

Changing needs and Expectation

Process Control System Model

To prevent variance: Change in operation: Operator training. . Change in material. . Change in the process itself: The machine itself. . Process design . Environmental conditions

Process Control System Model

Action on the output: Action on the output is frequently least economical .

Control Charts
Definition: A statistical tool to determine if a process is in control.

History of Control Charts

Developed in 1920s By Dr. Walter A. Shewhart Shewhart worked for Bell Telephone Labs

Two Types of Control Charts

Variable Control Charts Attribute Control Charts

Variable Control Charts

Deal with items that can be measured . Examples 1) Weight 2) Height 3) Speed 4) Volume

Types of Variable Control Charts

X-Bar chart R chart . MA chart.

Variable Control Charts

X chart: deals with a average value in a process R chart: takes into count the range of the values MA chart: take into count the moving average of a process

Attribute Control Charts

Control charts that factor in the quality attributes of a process to determine if the process is performing in or out of control.

Types of Attribute Control Charts

P chart C Chart U Chart

Attribute Control Charts

P Chart: a chart of the percent defective in each sample set. C chart: a chart of the number of defects per unit in each sample set. U chart: a chart of the average number of defects in each sample set.

Reasons for using Control Charts

Improve productivity Make defects visible Determine what process adjustments need to be made Determine if process is in or out of control

Control Charts
1- Collection: Gather data and plot on a chart. 2- Control: Calculate trial control limits from process data. Identify special causes of variation and act upon them 3- Analysis and improvement

Benefits of control charts

Be used by operators for on going control of a process. Help the process perform consistently and predictably. Allow the process to achieve: High quality. . Lower unit cost. . Higher effective capacity. Provide a common language for discussion the performance of the process. Distinguish special from common causes of variation, as a guide to local action or action on the system.`

Attributes Control Charts

p-charts
uses portion defective in a sample

c-charts
uses number of defects in an item

p-Chart
UCL = p + zsp LCL = p - zsp z = number of standard deviations from process average p = sample proportion defective; an estimate of process average sp = standard deviation of sample proportion

sp =

()

p-Chart Example
Sample 1 2 3 .. .. 20 Number of Defectives 6 0 4 .. .. 18 200 Proportion Defective .06 .00 .04 .. .. .18

20 samples of 100 pairs of jeans

p-Chart Example (cont.)

p =Total defectives/total sample observation = 200/20(100)= 0.10 UCL = p + z p(1 - p)/n = 0.10 +3 0.10(1-0.1)/100 =0.190 LCL = p-z p(1-p)/n = .10-3 0.10(1-0.1)/100 = 0.010

0.20 0.18 0.16 UCL = 0.190

p-Chart Example (cont.)

Proportion defective

0.14
0.12 0.10 0.08 0.06 0.04 0.02 2 LCL = 0.010 4 6 8 10 12 14 Sample number 16 18 20 p = 0.10

c-Chart
UCL = c + zsc LCL = c - zsc
sc = where c

c = number of defects per sample

c-Chart (cont.)
Number of defects in 15 sample rooms Sample Number of defectives 1 12 c=190/15=12.67 2 8 3 16 UCL= c +z sc .... .. = 12.67+3 12.67 . ... = 23.35 15 15 LCL = c - z sc = 12.67- 3 12.67 190 = 1.99

24 UCL = 23.35 21 18 15 12 9

Number of defects

c = 12.67

c-Chart (cont.)

6
3 LCL = 1.99

10

12

14

16

Sample number

Variable Control Chart

To track the accuracy of the process - Mean control chart or x-bar chart
To track the precision of the process - Range control chart

XYZ Ltd THICKNESS

PART NAME : GLASS RUN PART NO : SPECS MIN : 1.10 TO 1.50 MAX REASON :

MODEL : PROCESS CAPABILITY STUDY AUDIT DATE

Page 25/9/01

1 1 2 3 4 5
X R 1.50 1.50 1.60 1.50 1.50 1.52 0.10

2
1.50 1.50 1.48 1.48 1.50 1.49 0.02

3
1.50 1.50 1.50 1.52 1.60 1.52 0.10

4
1.50 1.53 1.50 1.50 1.50 1.51 0.03

5
1.50 1.50 1.48 1.53 1.50 1.50 0.05

6
1.50 1.50 1.50 1.50 1.50 1.50 0.00

7
1.50 1.50 1.50 1.50 1.55 1.51 0.05

8
1.50 1.50 1.50 1.50 1.55 1.51 0.05

9
1.50 1.50 1.50 1.45 1.45 1.48 0.05

10
1.50 1.55 1.55 1.50 1.55 1.53 0.05

11
1.60 1.60 1.50 1.50 1.55 1.55 0.10

12
1.50 1.55 1.55 1.50 1.50 1.52 0.05

13
1.60 1.55 1.50 1.50 1.50 1.53 0.10

14
1.50 1.60 1.55 1.50 1.50 1.53 0.10

15
1.60 1.55 1.50 1.50 1.50 1.53 0.10

16
1.55 1.45 1.50 1.50 1.50 1.50 0.10
X

17
1.60 1.60 1.50 1.50 1.45 1.53 0.15
UCL

18
1.55 1.50 1.55 1.60 1.50

19
1.50 1.50 1.60 1.60 1.55

20
1.50 1.48 1.55 1.50 1.55 1 2 3 4

d2

A2

D4
3.27 3.27 2.57 2.29

1.123 2.66 1.128 1.88 1.693 1.02 2.059 0.73

1.600 1.580 1.560 1.540 1.520 1.500 1.480 1.460 1.440 1.420 1.400 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

5 2.326 0.58 2.11 SUM X SUM X1+..+Xn 30.37 1.54 1.55 1.52 X SUM X1+..+Xn/n 1.519 0.10 0.10 0.07 R SUM R1+..+Rn/n 0.074 SIGMA R/d2 0.032 LCL CL 3 SIGMA 3 * R/d2 0.095 6 SIGMA 6 * R/d2 0.190

X - CHART

Cp =
Cpk= MIN OF Cpu OR Cpl

2.11
-0.20 4.41 1.500 1.100

Cpk =
18 19 20

USL LSL FOR X UCL = LCL = FOR R UCL = LCL =

R 0.25 0.20 0.15 0.10 0.05 0.00 -0.05 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

UCL

LCL

CL

X + A2.R X - A2.R (D3 = 0) D4.R D3.R

1.561 1.476

R - CHART

0.155 0.000

PROCESS STATAUS CONTROLLED NOT CONTROLLED 17 18 19 20

Establishing Control Chart

Step No.1
Select quality characteristics which needs to be controlled - Weight - Length - Viscosity - Tensile Strength - Capacitance

Establishing Control Chart

Step No.2
Decide the number of units, n to be taken in a sample. The minimum sample size should be 2. As the sample size increases then the sensitivity i.e. the quickness with which the chart gives an indication of shift of the process increases. However, with the increase of the sample size cost of inspection also increases. Generally, n can be 4 or 5.

Establishing Control Chart

Step No. 3
Decide the frequency of picking up of sample If the shift in the process average causes more loss, then take smaller samples more frequently. If the cost of inspection is high then take smaller samples at large interval.

Establishing Control Chart

As and general guidance, for deciding the frequency of taking a sample, we can use the table given in the next slide.
If our lot size in a shift is say 3000, then in a shift we require 50 units. If the sample size n, is say 4 then Number of visits to the process is = 504 = 12

The time of an 8-hour shift, be divided in 12 equal parts. Samples should be taken round about every 45 minutes.

Establishing Control Chart

Lot Size 66 - 100 101 - 180 181 - 300 301 - 500 501 - 800 801 - 1300 1301 - 3200 3201 - 8000 Total Number of items 10 15 25 30 35 40 50 60

Establishing Control Chart

Step No. 4 Collect data on a special control chart data collection sheet. ( Minimum 100 observations) The data collection sheet has following main portions: 1. General details for part, department etc. 2. Columns for date and time sample taken 3. Columns for measurements of sample 4. Column for mean of sample 5. Column for range of sample

Establishing Control Chart

Part SN Date Time X1 1 2 3 .. 25 Operation Measurement Mean Range X2 X3 X4 Other Details

Establishing Control Chart

Step No. 5
Fill up the control chart data sheet

1) As per the plan, visit the process and collect a sample of required number of units. 2) Measure the units and record. 3) Take requisite number of samples ( 20-25). 4) Calculate the mean of each of the sample. 5) Calculate the range of each of the sample.

Establishing Control Chart

A supervisor decided to put his process under statistical control. For the purpose of establishing control chart he collected 10 samples (Normally it should be 20 samples) containing 5 units. The samples were measured and the same is shown in the next slide. The desired target of the process, T is 50. Establish control chart for monitoring the process.

Establishing Control Chart

Subgroup No. 1 2 3 4 5 6 7 8 9 10 Subgroup Reading X1 47 48 49 49 51 50 51 50 48 49 X2 45 52 48 50 50 50 48 48 48 50 X3 48 47 52 52 53 49 50 50 49 50 X4 52 50 50 50 50 51 50 50 50 52 X5 51 50 49 49 48 47 54 52 51 51 Mean of subgroup Range of subgroup

Example - Data Collection

Subgroup No. 1 2 3 4 5 6 7 8 9 10 Subgroup Reading X1 47 48 49 49 51 50 51 50 48 49 X2 45 52 48 50 50 50 48 48 48 50 X3 48 47 52 52 53 49 50 50 49 50 X4 52 50 50 50 50 51 50 50 50 52 X5 51 50 49 49 48 47 54 52 51 51 Mean of subgroup Range of subgroup

Example - Calculation of Subgroup No.1

Measurements are 47, 45, 48, 52 & 51 Mean of measurements of subgroup No. 1 = (47 + 45 + 48 + 52 + 51)/5 = 48.6
Range of measurements of subgroup No. 1 = ( largest reading - smallest reading ) = ( 52 - 45 ) =7

Example - Calculation of subgroup Mean & Range

Subgroup No. 1 2 3 4 5 6 7 8 9 10 Subgroup Reading X1 47 48 49 49 51 50 51 50 48 49 X2 45 52 48 50 50 50 48 48 48 50 X3 48 47 52 52 53 49 50 50 49 50 X4 52 50 50 50 50 51 50 50 50 52 X5 51 50 49 49 48 47 54 52 51 51 Mean of subgroup 48.6 49.4 49.6 50.0 50.4 49.4 50.6 50.0 49.2 50.2 Range of subgroup 7 5 4 3 5 4 6 4 3 3

Establishing Control Chart

Calculate Mean Range, R
R= Sum of ranges of subgroups Total number of subgroups

In our case
R= (7 + 5 +4 3 + 5 + 4 + 6 + 4 + 3 + 3 ) Total number of subgroups

Establishing Control Chart

Step No. 7
Using following table of constants find trial control limit for mean and range control chart
Sub Group Size 2 3 4 5 6 7 A2 1.880 1.023 0.729 0.577 0.483 0.419 D4 3.267 2.527 2.282 2.115 2.004 1.924 D3 0 0 0 0 0 0.076

Establishing Control Chart

Step No. 8
Calculate Trial control Limits with target value, T Trial control limits for mean control chart Upper Control Limit, UCLx = T + A2 x R Lower Control Limit, LCLx = T - A2 x R

Trial control limits for range control chart Upper Control Limit, UCLr = D4 x R Lower Control Limit, LCLr = D3 x R

Calculation of Trial Control Limits

Size of Subgroup, n = 5
Factor A2, when n is 5 = 0.577

Factor D4, when n is 5 = 2.115

Factor D3, when n is 5 = 0 Target value, T = 50

Mean Range, R = 4.4

Establishing Control Chart

Step No. 8
Trial control Limits in our case

For mean control chart Upper Control Limit, UCLx = 50 + 0.577 x 4.4=52.5 Lower Control Limit, LCLx = 50 - 0.577 x 4.4=47.5
For range control chart Upper Control Limit, UCLr = 2.115 x 4.4 = 9.3 Lower Control Limit, LCLr = 0 x 4.4 = 0

Establishing Control Chart

Step No. 9
Discard the outliers

Outliers are those observations which do not

belong to normal population. If Outliers are included in the calculation, then the information is distorted.

Checking for Outliers

Checking for mean outliers
Scan column of sample means. If any mean of sample is more than UCLx or less than LCLx then drop that sample.

Checking for range outliers

Scan column of sample range. If any range is more than UCLr then drop that sample.

Checking for Outliers

If any sample(s) is dropped then recalculate the trial control limits using remaining sample(s). Continue this exercise till there is no further droppings. When there is no further dropping trial control limits becomes control limits for control chart. In all we can drop up to 25% of the samples

Checking for Outliers

In our case
- None of the subgroup mean is more than 52.5

- None of the subgroup mean is less than 47.5 - None of the range is more than 9.3 - None of the range is less than 0
Hence there is no revision of trial control limits is required. These limits can be used for maintaining the control charts.

Calculation of Control Limits for Mean Control Chart

Step No. 10
Compute warning limits for mean control chart
2 x A2 x R Upper warning limit, UWLx = T + 3 2 x A2 x R Lower warning limit, LWLx = T -

Calculation of Control Limits for Mean Control Chart

Warning limits for mean control chart in our example

Uwlx = 50 + = 51.7

2 x 0.577 x 4.4

Lwlx = 50 - 2 x 0.577 x 4.4

3
= 48.3

Constants for Range Control chart

Sample size, n 2 3 4 5 6 7 D4 3.27 2.57 2.28 2.11 2.00 1.92 D3 0 0 0 0 0 0.08 DWLR 0.04 0.18 0.29 0.37 0.42 0.46 DWUR 2.81 2.17 1.93 1.81 1.72 1.66

Calculation of Control Limits for Range Control Chart

Step No. 11
Compute warning limits for range control chart

Upper Warning Limit, UWLr = DWUR x R

Lower Warning Limit, LWLr = DWLR x R

Calculation of Warning Limits for Range Control Chart

In our case
Size of sub group, n = 5 Mean range R = 4.4

DWUR when n is 5 = 1.81

DWLR when n is 5 = 0.37

Calculation of Warning Limits for Range Control Chart

In our case warning limits for range control chart
Upper Warning Limit, UWLr = DWUR x R

= 1.81 x 4.4
=8

Lower Warning Limit, LWLr = DWLR x R

= 0.37 x 4.4

= 1.6

Summary of Effect of Process Shift

When there is no shift in the process nearly all the observations fall within -3 s and + 3 s.
When there is small shift in the mean of process some observations fall outside original -3 s and +3 s zone.

Chances of an observation falling outside original -3 s and + 3 s zone increases with the increase in the shift of process mean.

Our Conclusion from Normal Distribution

When an observation falls within original +3 s and -3 s zone of mean of a process, we conclude that there is no shift in the mean of process. This is so because falling of an observation between these limits is a chance. When an observation falls beyond original +3 s and -3 s zone of process mean, we conclude that there is shift in location of the process

Rules for special causes

RULE 1 RULE 2 RULE 3
RULE 4

Any point outside one of the control limits A run of seven points all above or all below the central line, or all increasing / all decreasing Any unusual patterns or trends within the control limits The proportion of points within the middle 1/3 of the region between the control limits differs excessively from 2/3

Interpreting Control Chart

Because the basis for control chart theory follows the normal distribution, the same rules that governs the normal distribution are used to interpret the control charts. These rules include:
- Randomness. - Symmetry about the centre of the distribution. - 99.73% of the population lies between - 3 s of and + 3 s the centre line. - 95.4% population lies between -2 s and + 2 s of the centre line.

Interpreting Control Chart

If the process output follows these rules, the
process is said to be stable or in control with only

common causes of variation present.

If it fails to follow these rules, it may be out of

control with special causes of variation present.

These special causes must be found and

corrected.

Special causes - Rule 1

One point outside control limit

UCL UWL

Statistics

LWL LCL

8
128

Sample Number

Special causes - Rule 2

Seven consecutive points on one side of the centre line
UCL UWL

Statistics

LWL LCL

8
129

Sample Number

Special causes - Rule 3 Seven consecutive points having upward trend UCL UWL

Statistics

LWL LCL

8
130

Sample Number

Special causes - Rule 3

Seven consecutive points having downward trend
UCL

UWL Statistics

LWL
LCL

8
131

Sample Number

Special causes - Rule 4

Two consecutive points between warning limit and corresponding control limit

UCL UWL

Statistics

LWL LCL

3 4 5 Sample Number

8
132

Special causes - Rule 4

Two points out of three consecutive points between warning limit and corresponding control limit
UCL UWL

Statistics

LWL LCL

3 4 5 Sample Number

8
133

How capable is the process can we meet the target?

A Process Is in Control If
1. no sample points outside limits 2. most points near process average 3. about equal number of points above and below centerline 4. points appear randomly distributed

Can we meet the target?

There are four different sorts: Smaller the better (ideally 0 in many cases) - e.g. By 2015 no one should be waiting more than 4 hours in A&E from arrival to admission, transfer or discharge. Bigger the better - e.g. By March 2015, 80% of day case patients must have booked admissions.

Can we meet the target?

Within specified limits - e.g. Therapeutic range for medication or drugs such as Warfarin
On a specified limit - e.g. When filling containers with medicine they must be above the minimum fill specified on the label

Example
A waiting time target has been given of not more than 4 hours so the Spec Limit is 4 hours (the maximum allowable waiting time). A better wording for this target would be:
All patients should wait less than 4 hours, the shorter time the better This encourages people to think about how to improve rather than just work towards the poorest performance that patients will tolerate.

What is wrong with targets?

Targets are brought in with the best intentions, but
Targets are often arbitrary, not based on any objective criteria, just what we think people might accept. They are:
what is believed to be good enough when we know it the minimum requirement what an organisation thinks it can get away with the worst allowable really isnt

Process Capability
The ability of a process to meet product design/technical specifications Conducted only when the process is normally distributed.

Process Capability
Tolerances
design specifications reflecting product requirements

Process capability
range of natural variability in a process what we measure with control charts

Process Capability
Design Specifications (a) Natural variation exceeds design specifications; process is not capable of meeting specifications all the time.

Process
Design Specifications (b) Design specifications and natural variation the same; process is capable of meeting specifications most of the time.

Process

Process Capability (cont.)

Design Specifications (c) Design specifications greater than natural variation; process is capable of always conforming to specifications.

Process Design Specifications

(d) Specifications greater than natural variation, but process off center; capable but some output will not meet upper specification. Process

Process Capability Measures

Process Capability Ratio
Cp = tolerance range process range

upper specification limit lower specification limit = 6s

Computing Cp
Net weight specification = 9.0 oz 0.5 oz Process mean = 8.80 oz Process standard deviation = 0.12 oz
upper specification limit lower specification limit Cp = 6s

9.5 - 8.5 = = 1.39 6(0.12)

What value of cp is acceptable

Cp<1.0 Cp=1.0 Cp1.3 -1.5 Cp=2 poor process so ok good Excellent, that is6 sigma

Process Capability Index Example

A manufacturing process produces a certain part with a mean diameter of 2 inches and a standard deviation of 0.03 inches. The lower and upper engineering specification limits are 1.90 inches and 2.05 inches. Cp=USL-LSL/6 =2.05-1.90/6(.03) Process is not a capable process since Cp<1.

Process Capability Measures

WHY Cpk IS NEEDED? IS Cp NOT ENOUGH?

Process Capability Measures

WHY Cpk IS NEEDED? IS Cp NOT ENOUGH? Cp TELLS U ONLY ABOUT THE SMARTNESS OF CURVE Cpk TELLS U ABOUT THE POSITIONING / LOCATION OF THE CURVE

Process Capability Measures

Process Capability Index
= - lower specification limit x 3s

,
=

Cpk = minimum
upper specification limit - x 3s

Computing Cpk
Net weight specification = 9.0 oz 0.5 oz Process mean = 8.80 oz Process standard deviation = 0.12 oz
= x - lower specification limit , 3s = upper specification limit - x 3s 8.80 - 8.50 9.50 - 8.80 , = 0.83 3(0.12) 3(0.12)

Cpk = minimum

= minimum

Appendix: Determining Control Limits for x-bar and R-Charts

SAMPLE SIZE n FACTOR FOR x-CHART A2 FACTORS FOR R-CHART D3 D4

Fact ors

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

1.88 1.02 0.73 0.58 0.48 0.42 0.37 0.44 0.11 0.99 0.77 0.55 0.44 0.22 0.11 0.00 0.99 0.99 0.88
4-152

Return

0.00 0.00 0.00 0.00 0.00 0.08 0.14 0.18 0.22 0.26 0.28 0.31 0.33 0.35 0.36 0.38 0.39 0.40 0.41

3.27 2.57 2.28 2.11 2.00 1.92 1.86 1.82 1.78 1.74 1.72 1.69 1.67 1.65 1.64 1.62 1.61 1.61 1.59