Available as a Powerpoint data file The Elsmar Cove [Link]

html

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 1

Statistics

Statistical Process Control & Control Charting

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 2

Recommended Statistical Course Attendance

Basic Business Office, Staff, & Management Statistical Process Control Production Floor, Production Staff & Management, All of Quality & Engineering Measurement System Analysis Production Staff & Management, All of Quality & Engineering Advanced Business Selected Staff, & Management Design of Experiment Specific Production Staff, Quality Engineers& QA Manager & Product Engineering

*Your APQP Team should attend as many classes as time will allow
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 3

Basic Definitions
Statistics:
the science of collecting, analyzing, interpreting and presenting data.

Statistic is a single characteristic taken from this process.

Statistical Analysis and SPC
Slide 4

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Universe, Populations & Samples

Universe:
the collection of all elements

Population:
the set of objects of interest

Sample:
a subset of objects taken from the population

Randam Sample:
all possible samples of the same size have an equal chance of occuring.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 5

Statistical Methodology
Statistical methods are procedures for drawing conclusions about populations utilizing information provided by random samples.
Regression Analysis

Data Collection

Frequency Distribution

Predictive Statistics
Correlation Analysis Experimental Design

Statistical Methodology

Organization & presentation

Histograms Central Tendency

Statistical Inference
Hypothesis Testing

Descriptive Measures
Statistical Analysis and SPC

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Dispersion
Slide 6

Classification of Statistics:
Descriptive

statistics: the methodology of efficiently collecting, organizing, and describing data.

Inductive

To:
Predictive

Statistics: the process of drawing conclusions about unknown characteristics of a population usually based from a sample taken from the population

Statistics: the Tomorrow we will produce 5000 parts, based off of last weeks production of : process of predicting future 5250; 5500; 4500; 4750; 5000 values based on historical data.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 7

Four Levels of Measurements

Nominal: Objects are classified into simple attributable categories

with no quantitative difference between them, (Yes/No, Good/Bad).

Ordinal: Objects are able to be arranged, ranked, or ordered into a

meaningful attributable arrangement with no real measurement. (Yellow/Blue/Green, Square/Round/Triangle)

Interval: observations are able to be ranked into exact differences

between any two observations, measurements with no natural origin or zero, 80 degrees is not twice as hot as 40 degrees. A one unit scale change corresponds to a one unit change on the object being studied.

Ratio: contains all the properties of interval but has a natural origin.
Having a natural origin allows 25 to behalf of 50. Note that each successive level has all the properties of the previous.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 8

SPC is Concerned With:

Regression Analysis

Data Collection

Frequency Distribution

Predictive Statistics
Correlation Analysis Experimental Design

Statistical Methodology

Organization & presentation

Histograms Central Tendency

Statistical Inference
Hypothesis Testing

Descriptive Measures

Dispersion
Slide 9

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Data Collection

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 10

Venn Diagrams
A
Elements distinct objects A a in a set {a, b}
b

b a

c d f

A
SET (A) Collection of Distinct Objects Having Some Attribute in Common This is an example of a

SUBSET (A)
portion of the set defined in some unambiguous way

MUTUALLY EXCLUSIVE

Subsets A & B have no elements in common

A
A

b a

B
d f

A
A B)

Null or Empty Set

a b c d

B
f

Intersect (A

The Total Set of Elements of Interest

Common Elements of sets {a,c}

Complement ()is the inverse of the specific set called out Complement of set A is A or { 0 } subset A = {c,d,f}
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

UNIVERSAL SET

Union (A U B)
Component Elements of sets {a,b,c,d}

Slide 11

Events
A Sample Space (S) is the set of all possible outcomes. An Event is any subset of a sample space. A Simple Event is any subset of the sample space to include a single outcome. S = {H, T} A Compound Event is any subset of the sample space which consists of two or more simple events.
S = {HHH, HHT, HTH, HTT, THH, THT, TTH, TTT}
a He ds
Mutually Exclusive

Tails

HHH
s ead

HHT HTH HTT THH THT TTH TTT

a He

ds

Tails

Tails

ds Hea Tail s

A set of Events is said to be Collectively Exhaustive if all simple events are included, (like the equation above).
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 12

Putting it all Together

When gathering data for a process, we must realize that

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 13

Organizing Data

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 14

Presenting Data

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 15

Descriptive Measurement

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 16

SPC - Variation
The greatest obstacle to the use of modern quality control methods is the mistaken belief that the math involved is too difficult for the ordinary person to understand. While the Theories on which many statistical methods are based involve some high-powered mathematics, you dont have to be a mathematician to use many of these modern methods nor do you have to fully understand them. Lets start by taking a pie and cutting it into pieces making each piece the same size as best we can. There is an inherent variability in even a very good product. You can see this if you have a measurement device which is sensitive enough to detect that variation. In life this is typically the case. No two flowers, snow flakes, pieces of pie or thumb prints are the same. They may look alike, but when seen through a microscope we can see they are each different. We call these differences Variation.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 17

Causes of Variation
Special Causes of Variation Common Causes of Variation

Special causes are problems that arise in a periodic fashion. They are somewhat unpredictable and can be dealt with at the machine or operator level. Examples of special causes are operator error, broken tools, and machine setting drift. This type of variation is not critical and only represents a small fraction of the variation found in a process.

Common causes are problems inherent in the system itself. They are always present and effect the output of the process. Examples of common causes of variation are poor training, inappropriate production methods, and poor workstation design.

As we can see, common causes of variation are more critical on the manufacturing process than special causes. In fact Dr. Deming suggests that about 80 to 85% of all the problems encountered in production processes are caused by common causes, while only 15 to 20% are caused by special causes

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 18

Causes of Variation
Special Causes of Variation
Accounts for 5-15% of quality problems. Is due to a factor that has "slipped" into the process causing unstable or unpredictable variation. Are unpredictable variations that are abnormal to the process including human error, equipment failure, defective/changed raw materials, acid spills, power failures, etc.; failure to remove them can result is corrosion, scale, metal fatigue, lower equipment efficiency, increased maintenance costs, unsafe working conditions, wasted chemicals, increased down-time (plant shut-down...), etc. Removal of all special causes of variation yields a process that is in statistical control. Correctable by local personnel.

Common Causes of Variation

Account for 85-95% of quality problems Are due to the many small sources of variation "engineered" into the process of the "system. Are naturally caused and are always present in the process because they are linked to the system's base ability to perform; it is the predictable and stable inherent variability resulting from the process operating as it was designed. Standard deviation, s, is used as a measure of the inherent process variability; it helps describe the well-known normal distribution curve. Correctable only by management; typically requires the repair/replacement of a system's component, or the system itself.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 19

Distribution
Sometimes you can look at two slices of pie and tell which is bigger. Sometimes you cannot. Slice a pie up into what you think are equal sized pieces and line them up according to size. Many look the same. If we want to arrange the pieces according to size, we need another way to tell how big each piece is. A weight scale will do quite well. Now lets look at what we would find if we weighed each piece. There are big and little pieces, but you can see that the number of pieces in each step of the graph (weight group) varies from the largest piece to the smallest piece in a fairly regular and symmetrical pattern. This is the Distribution of the weights. The curve is what we would expect if the Distribution was a Normal distribution. Imagine doing this with 100 pies!
Histogram 7 6 5
Count

4 3 2 1 0 4 6 8 10 12 Slice Wt 14 16 18

Frequency Distribution for Slice Wt From () 5.000 6.200 7.400 8.600 9.800 11.000 12.200 13.400 14.600 15.800 To (<) 6.200 7.400 8.600 9.800 11.000 12.200 13.400 14.600 15.800 17.000 Total Count 2 2 3 3 4 6 4 3 2 2 31 Normal Count .994 1.856 2.968 4.065 4.766 4.786 4.116 3.031 1.911 1.032 29.527
Slide 20

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

A Normal Distribution

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 21

Normal Distribution (Bell Curve)

This is a pattern which repeats itself endlessly not only with pieces of pie but in manufactured products and in nature. There is always an inherent Variability. Sometimes its a matter of finding a measurement device sensitive enough to measure it. Measurements may be in volts, millimeters, amperes, hours, minutes, inches or one of many other units of measure. It you take a sample of a population (such as height) and you chart their distribution, you will end up with a curve that looks like a bell. A Distribution which looks like a bell is a Normal Distribution. Normal Distributions are the most common type of distribution found in nature - but they are not the ONLY type of distribution.
1.59 1.66 1.73 1.87 1.94 2.01

Heights of men in the military: Average height: 1.80 meters Shortest: 1.59 meters Tallest 2.01 meters Sixty-eight percent are between 1.73 and 1.87 meters. Ninety-five per cent are between 1.66 and 1.94 meters.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 22

Standard Deviation
Human Proportions Men Women Height X-Bar sigma X-Bar sigma Standing 1.8 0.07 1.6 0.05 Sitting 0.9 0.03 0.86 0.02

X = X-Bar = Mean

The most common measure of dispersion. The standard deviation is the square root of the variance, which is the sum of squared distances between each datum and the mean, divided by the sample size minus one. For a tiny sample of three values (1,2,15), the mean is: (1 + 2 + 15)/3 = 6 and the variance is ((1 - 6)^2 + (2 - 6)^2 + (15 - 6)^2) / (3 - 1) = 61 The standard deviation () is not a very helpful measure of spread for distributions in general. Its usefulness is due to its intimate connection with a special type of distribution, namely to the normal distribution. The standard deviation is more sensitive to a few extreme observations than is the mean. A skewed distribution with a few values in the 'long tail' will have large standard deviation and it will not provide much helpful information in such cases.

= sigma = Standard Deviation

Mean

Mean

= 0.070

= 0.80

Finding the Standard Deviation

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 23

Standard Deviation

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 24

Basic Terms
X = X-Bar = Average This is X-Bar. It tells us the Average of a group of numbers (in this case average height). X-Bar is the middle of the curve where we have the largest percentage of men.

= sigma = Standard Deviation is the Greek letter Sigma. It

is the distance from the center of the curve to the point where the curve stops curving downward and starts curving outward. Were interested in points at 1, 2 and 3 Standard Deviations from the Mean (the center).

If we measure 1 Standard Deviation on each side of the center of the curve , 68% of the Area will be between the lines drawn through these points. The Human Proportions table tells us the Standard Deviation ( dispersion) of mens heights is 0.07 meters, so by simple addition and subtraction we know that 68% of the men are between 1.73 and 1.87 meters.

Mean

1.8 + 0.07 = 1.87 1.8 - 0.07 = 1.73

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 25

More On Standard Deviation

Two Standard Deviations Two standard deviations would be 0.14 meters. Ninety-five percent of men are between 1.66 and 1.94 meters. Three Standard Deviations are = 0.07 x2 = 0.14 0.21 meters and 99.73% of the Area of the Normal curve is between +3 1.8 - 0.14 = 1.66 sigma (+3) and -3. We can say that practically all of the men are 1.8 + 0.14 = 1.94 between 1.59 and 2.01 meters. Only about 13 in 10,000 men are taller than 2.01 meters. The same is true of men below 1.59 meters. Three Standard Deviations

= 0.07 x3 = 0.21
1.8 - 0.21 = 1.59 1.8 + 0.21 = 2.01

But
What does this have to do with making nuts, bolts, transmissions and aspirin tablets? If we have a machine which will measure every man who comes down the street, what are the odds that any individual man is 1.8 meters tall? What are the odds that the next man who walks by will be between 1.59 and 2.01 meters? How about between 1.73 and 1.87 meters?

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 26

Dispersion Revisited
Sixty-eight % of all men are between 1.73 and 1.87 meters tall. So, there are 68 chances in 100 that the next man will be between these Limits. Id have to give you odds of 68 to 32 (2:1). Ninety-five % of all men are between 1.66 and 1.94 meters tall. That gives us odds of 95 to 5 (19:1) that the next man wont be shorter than 1.66 nor taller than 1.94 meters. There are only 2 1/2 chances in 100 that the next man will be taller than 1.94 meters and 13 1/2 chances in 10,000 (1:740) that he will be taller than 2.01 meters. The gadgets we make are supposed to be exactly 20 millimeters long. We are going to make a Lot of 500 widgets. We measure the Lot with a micrometer and find: As you can see, they are in the same Bell Shaped Pattern as the pieces of pie and the men.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 27

Population vs Sample
The Average is 20 millimeters, but individually they vary from 19.10mm to 20.90mm. This is OK with our customers because their specification calls for and Average of 20mm with a Tolerance of +/-1.5mm. That means they will accept anything between 18.5 and 21.5mm. These points are know as the Limits or the Tolerance Limits. Our problem is to keep the widget cutting machines at the center of 20.0mm and to not let individuals vary (disperse) more than they did in this lot. All of the gadgets we measured were made at one time (the same Run) on the same machine by the same operator and from the same material lot. Because of all these things being the same, we can be quite sure that the Pattern of their Variation will almost exactly fit the Normal Curve. We would also know that the greatest Individual Variations will be quite close to +/-3 . This is a Spread of 6 Standard Deviations.

The difference between the Longest (20.9mm) and the Shortest (19.1mm) is 1.80mm. One sixth of 1.80 is 0.3. So. The standard deviation is 0.3mm. Since we measured every widget, we measured the entire Population. We can do the same thing with a Sample.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 28

Samples
To take a sample, you take gadgets out of a bucket. A Sample is usually more than 1 piece. Well use a Sample Size of 5. BUT WAIT!!! We want a Random Sample. To get a Random Sample we must dig around in the bucket were taking gadgets from. If we do this 5 times, we shold see something like the table below.
Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 20.16 20.25 20.02 + 19.73 20.33 20.27 + 19.63 19.88 20.46 + 20.03 19.94 20.15 19.99 19.41 20.37 + 19.54 20.59 + 19.96 20.01 19.68 Grand 19.85 19.90 19.78 20.09 20.13 Average Average 19.95 20.10 19.93 19.94 20.11 20.01 Range 0.73 0.96 0.24 1.05 0.69 0.734 + Largest Measurement in sample Smallest Measurement in sample

These 5 Sample Averages vary quite a bit from the Lot Average of 20.0mm, but the Grand Average is 20.01mm. We wouldnt always come so close, but if we have enough Samples (say 20 to 25) the Average of a number of Samples will be a good Estimate of the Actual Average of the Lot. Youll notice the table has another number labeled Range. The range is the difference between the largest and smallest measurement in each sample. Like the Averages, the Ranges vary in each sample.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 29

Samples
In our Lot of 500 gadgets we divided the Range by 6 to get the Standard Deviation. In that case we were measuring a Population. We cant do that with Samples of 5 because we had 500 chances of getting some of the extreme values - not just 5 chances. In Samples of 5 there is less chance of getting some of the extreme values so the Divisor will be smaller. We call this divisor the d2 (d two) factor. We can find (look up) the appropriate d factor in a Table. For a sample size of 5, the d2 factor is 2.326 (from the table). Dividing the Average Range (0.734) by the d2 factor (2.326) gives us a Standard Deviation of 0.316. This doesnt differ much from the figure of 0.3 which we got when we measured 500 pieces. This shows that by taking relatively small samples we can Estimate two very important things about the Distribution of the Measurements of any product: The Average and the Standard Deviation. X = X-Bar = Average

= sigma = Standard Deviation

The Standard Deviation () of the gadgets is 0.3mm and the Average is 20.00mm. Knowing these 2 values, we can now put some % figures on our Normal Distribution Curve. If were doing a good job of cutting the gadgets to the Average of 20mm, 68% should be between 19.70 and 20.30mm. Ninety-five percent should be between -2 s and +2 (19.40mm and 20.90mm). 99.73%

should be between -3 and +3 (19.10mm and 20.90mm).

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 30

Samples

X-Bar +/-1 sigma 20.00 + 0.3 = 20.30 20.00 - 0.3 = 19.70. 68%

X-Bar +/-2 sigma 20.00 + 0.6 = 20.30 20.00 - 0.6 = 19.40 95%

X-Bar +/-3 sigma 20.00 + 0.9 = 20.90 20.00 - 0.9 = 19.10 99.73%

Even if something very unusual happens to give me an occasional gadget shorter than 19.10mm, or longer than 20.90mm, as long as the gadget machine is properly adjusted to make an Average of 20.00mm I could give you odds of 997 to 3 that you cant find one (At Random) outside of these Limits. But - You and I know that machines, materials and operators do not always do what they are supposed to do.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 31

Samples
Bill Jones works on machine #2. Bills mother-in-law has been camping at his house for the last 6 weeks. Bill has been feeling sort of frustrated. His mind just isnt focused on cutting gadgets. Yesterday his machine was off about four tenths (4/10) of a millimeter. How did we find out? We cannot if we just look at a few individual gadgets. Lets assume that his Average was 20.40mm. The Pattern of the Variation of Individual Measurements about the center of the 20.40mm was the same as in any good lot Except - the Bell Shaped Curve was shoved over +0.40mm. Sigma (Standard Deviation) is the same (0.3mm). The 3 Limits of his Distribution are now changed to 19.50mm and 21.30mm. They are within the customers specification of 18.50mm through 21.50mm But - the customer will probably complain about the higher Average.

Weve already seen that a good Lot can be expected to vary between 19.10mm and 20.90mm. Five per cent of the gadgets Bill was making were over the limit of 20.90mm. NOTE: You have top use a Statistical Table to figure this out! If we were to pick up single pieces from Bills machine, there are only 5 chances in 100, or 1 chance in 20, that we would get one of the pieces which is larger than 20.90mm.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 32

So What?
Remember a while back we gave Odds of 740 to 1 that the next man to walk by the height machine would not be taller than 2.01 meters. We were betting that he would not exceed the 3 sigma Limit. What if we look at the next 2 men who walk by. Its a safe bet that the Average of the two will be nearer the Grand Average of 1.80 meters than to the extreme height of 2.01 meters. For the same Odds of 740:1, Id have to use narrower 3 sigma () Limits. If we bet on the Average of the next 4 men, wed have to use still narrower Limits? No! These Limits are NOT reduced in proportion to the number of men (the Sample Size). Limits for the Average of 2 men would be about 2/3 of the Limits for Individuals. For the Average of 4 men, the Limits would be 1/2 the Limits for Individuals. How do we get these figures? Not only do measurements of Individuals vary in the Bell Shaped Pattern of the Norman Distribution Curve, but Sample Averages vary in the same Pattern, about the same Center. However, the Standard Deviation (sigma or ) of Sample Averages is the of Individuals divided by the Square Root of the Sample Size. In the case of our men, the Sigma of Individual Heights is 0.07 meters. For each Sample size we divide 0.07m by the Square Root of the Sample Size.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 33

A Square Root?
We get Square Roots from a Table or a Calculator. Lets apply this principle to measuring gadgets from Bills machine. We take Samples of 5 gadgets from the machine every 15 minutes.
Sample Size 2 4 5 10 Square Root 0.07 Divided 3 Times of Sample by Square Sigma of Size Root Acerages 1.414 0.050 0.150 2.000 0.035 0.105 2.236 0.031 0.093 3.162 0.022 0.066 Three sigma Limits Shortest (1.8-3 sigma) 1.65 1.70 1.71 1.73 Longest (1.8 + 3 sigma 1.95 1.91 1.89 1.87

We get Square Roots from a Table or a Calculator. Lets apply this principle to measuring gadgets from Bills machine. We take Samples of 5 gadgets from the machine every 15 minutes and measure them with a dial micrometer. We have seen that Standard Deviation (sigma or ) of this measurement is 0.3mm. The Square Root of 5 (the Sample Size) is 2.236. Dividing 0.3 by 2.236 gives us a Standard Deviation of 0.134mm for Averages of 5 gadgets. 3 Standard Deviations is 0.402mm.

0.3mm / 5 = 0.134mm
(Adjustment for Sample Size)

3 x 0.134mm = 0.402mm
(Yields the Spread)
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 34

A Shift in the Center

If Bills machine is set up right, for an Average of 20.00mm, Averages of 5 should not vary more than 0.4mm above and below this Center (between 19.60mm and 20.40mm). As long as our Sample Averages vary between these limits we can be reasonably sure that the Actual Average of the machine is close to the Specified Average of 20.00mm. Its a 740:1 bet that no Sample Average is higher than 20.40mm. If Bills machine is set up right, for an Average of 20.00mm, Averages of 5 should not vary more than 0.4mm above and below this Center (between 19.60mm and 20.40mm). As long as our Sample Averages vary between these limits we can be reasonably sure that the Actual Average of the machine is close to the Specified Average of 20.00mm. Its a 740:1 bet that no Sample Average is higher than 20.40mm. But If Bills machine setting is 20.40mm, Averages of Samples of 5 will vary 0.4mm on either side of this High Center. There is now a 50:50 chance that any Sample Average will be above 20.40mm. Of course, theres the same 50:50 chance that the Sample Average will be BELOW 20.40mm. But With these Odds in our favour were practically certain to get a high Average in the first 2 or 3 Samples.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 35

Control Chart
Developed in the mid 1920's by Walter Shewhart of Bell labs, this SPC tool has become a major contributor to the quality improvement process. It allows the use to monitor and control process variation. It also allow the user to make the proper corrective actions to eliminate the sources of variation. Even though they require the user to have some statistical background, the are relatively easy to construct. There are two basic types of control charts, the average and range control charts. The first deals with how close the process is to the nominal design value, while the range chart indicates the amount of spread or variability around the nominal design value. A control chart has basically three lines: the upper control limit UCL, the center line CL (average), and the lower control limit LCL. These lines are computed from samples taken from the production line. Each sample represents a point on the control chart. A minimum of 25 points is required for a control chart to be accurate. In the Control Chart at the right, the Center Line says that the gadgets should Average 20.00mm. The Upper and Lower dotted lines say that if this specification is met, Averages of Sample sizes (n) of 5 (n = 5) should not be less than 19.60mm nor more than 20.40mm.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

A Control Chart is a Picture of what we are doing.

Upper Control Limit X Control Chart of Gadgets Sample Size (n) = 5

Lower Control Limit Average

Statistical Analysis and SPC

Whats Wrong With This Picture?

Slide 36

Control Charts
Control Charts Serve 2 Basic Purposes
Control Charts provide information for decision making with respect to a manufacturing process. When a random pattern of variation occurs and process capability has been established, the process should be left alone. When an unstable (nonrandom) pattern of variation is occurring, action should be taken to find and eliminate the disturbing or assignable causes. Control Charts provide information for decision making with respect to recently produced product. The Control Chart can be used as one source of information in the determination of whether product should be released to the customer or some alternate disposition made.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 37

Interpreting Control Charts

Control Charts provide information as to whether a process is being influenced by Chance causes or Special causes. A process is said to be in Statistical Control when all Special causes of variation have been removed and only Common causes remain. This is evidenced on a Control Chart by the absence of points beyond the Control Limits and by the absence of Non-Random Patterns or Trends within the Control Limits. A process in Statistical Control indicates that production is representative of the best the process can achieve with the materials, tools and equipment provided. Further process improvement can only be made by reducing variation due to Common causes, which generally means management taking action to improve the system.
Upper Control Limit Average Lower Control Limit

A. Most points are near the center line. B. A few points are near the control limit. C. No points (or only a rare point) are beyond the Control Limits.
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 38

Interpreting Control Charts

When Special causes of variation are affecting a process and making it unstable and unreliable, the process is said to be Out Of Control. Special causes of variation can be identified and eliminated thus improving the capability of the process and quality of the product. Generally, Special causes can be eliminated by action from someone directly connected with the process. The following are some of the more common Out Of Control patterns:
Change To Machine Made Tool Broke Tool Wear Upper Control Limit Average Lower Control Limit

A. Most points are near the center line.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 39

Interpreting Control Charts

Points Outside of Limits
Upper Control Limit Average Lower Control Limit

Trends

A Run of 7 intervals up or down is a sign of an out of control trend. A Run of 7 intervals up or down is a sign of an out of control trend.

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 40

Interpreting Control Charts

Run Of 7

A Run of 7 successive points above or below the center line is an out of control condition. A Run of 7 successive points above or below the center line is an out of control condition.

Run Of 7

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 41

Interpreting Control Charts

Systematic Variables

Predictable, Repeatable Patterns Predictable, Repeatable Patterns

Cycles

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 42

Interpreting Control Charts

Freaks

Sudden, Unpredictable Sudden, Unpredictable

Instability

Large Fluctuations, Erratic Up and Down Movements Large Fluctuations, Erratic Up and Down Movements

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 43

Interpreting Control Charts

Mixtures

Unusual Number of Points Near Control Limits (Different Machines?) Unusual Number of Points Near Control Limits (Different Machines?)

Sudden Shift in Level

Typically Indicates a Change in the System or Process Typically Indicates a Change in the System or Process

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 44

Interpreting Control Charts

Stratification

Constant, Small Fluctuations Near the Center of the Chart Constant, Small Fluctuations Near the Center of the Chart

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 45

Analysis Summary
There is a wide range of non-random patterns that require action. When the presence of a special cause is suspected, the following actions should be taken (subject to local instructions).

1. CHECK
Check that all calculations and plots have been accurately completed, including those for control limits and means. When using variable charts, check that the pair (x bar, and R bar) are consistent. When satisfied that the data is accurate, act immediately.

2. INVESTIGATE
Investigate the process operation to determine the cause. Use tools such as: Brainstorming Cause and Effect Pareto Analysis Your investigation should cover issues such as: The method and tools for measurement The staff involved (to identify any training needs Time series, such as staff changes on particular days of the week Changes in material Machine wear and maintenance Mixed samples from different people or machines Incorrect data, mistakenly or otherwise Changes in the environment (humidity etc.)
Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 46

Analysis Summary
[Link]
Decide on appropriate action and implement it. Identify on the control chart The cause of the problem The action taken As far as possible,eliminate the possibility of the special cause happening again.

4. CONTINUE MONITORING
Plotting should continue against the existing limits The effects of the process intervention should become visible. If not, it should be investigated. Where control chart analysis highlights an improvement in performance, the effect should be researched in order that: Its operation can become integral to the process Its application can be applied to other processes where appropriate Control limits should be recalculated when out of control periods for which special causes have been found have been eliminated from the process. The control limits are recalculated excluding the data plotted for the out of control period. A suitable sample size is also necessary. On completion of the recalculation, you will need to check that all plots lie within the new limits

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Systematic Variables

Slide 47

Control Charts
The most common type of Control Chart is the X-Bar & R Chart

X = X-Bar = Average

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 48

Attribute Control Charts

Cayman Business Systems USA 513 227-4675 Revision A 19991008 - [Link]

Statistical Analysis and SPC

Slide 49