UNIT II VARIATION IDENTIFICATION OF VARIATION Variation is an
inevitable condition-- let's learn to use it to ouradvantage.
If we were to
take a box of chalk and lay the pieces on a table, we wouldprobably say that they all look alike.
Maybe they do all look alike, but if we measure them we will be able todetect a difference.
If we use a ruler, we may be able to sort the pieces intogroups according to differences of 1/64 inch.
However it may be impossible to detect a difference among some of the pieces.
If we used a micrometer we could measure to the nearest thousandth of aninch.
This time we would find that the pieces which were alike when measured tothe nearest 1/64 inch vary quite a bit when measured to the nearest thousandthof an inch.
We could go so far as to select the pieces that showed no difference whenmeasured to the nearest thousandth of an inch and measure them to the nearestten thousandth of an inch.
As the increment of measurement becomes morerefined, we will be able to detect more and more variation.
The foregoing is illustrative of the natural laws of variation.
We mayobserve two things that appear to be identical, but closer observation willreveal a difference.
There may be times when things are so much alike that it isdifficult to find any difference.
Variation is important to us only if we canmeasure it, or if it has some affect on the functions for which it was made.
Present day industry is capable of producing parts so much alike that they mustbe measured in millionths of an inch in order to detect any variation.
It must be pointed out that the measurement of length is not the onlymeasure of variation.
It is possible compare variation in grams, tensilestrength, specific gravity, feet per minute, temperature, hardness, or by manyother methods.
Industry is constantly seeking new and improved methods of measurement to help detect variation.
ORDER OF VARIATION One of the most difficult things for us to understand is the order ofvariation.
If we were to measure the parts coming out of a machine and recordthe measurements, it would appear that there is no logical sequence to thevariation.
However if we arrange our data in some logical order, we can notonly identify the variation, but we can predict future outcomes.
CAUSES OF VARIATION In industry we generally attribute variation to four causes.
We refer tothese as the "4-M's".
They are:
1.
Men 2.
Machines 3.
Materials 4.
Methods 1.
Men - The manner in which an operator adjusts the machine, sharpens atool, or controls the feeding mechanism has an affect upon the variation of theproduct coming from the machine.
2.
Machines - The type of machine, the amount of wear, or the conditionof the machine will generate causes that contribute to variation.
3.
Materials - The uniformity or lack of uniformity of the material weuse may cause variation in the product.
4.
Methods - One of the most important factors that contribute tovariation is the method we use to perform the operation.
Industry recognizesthis and uses written instructions and routing sheets to establish standardmethods of operation.
We can not expect an operators to get the best resultsfrom a machine if they do not have instructions to follow.
By the same token,we can not expect to get uniformity if we do not have a standard way of doing things.
CHANCE CAUSE SYSTEM As long as the men, machines, materials and methods continue in a normalfashion, the product will exhibit a normal amount of variation.
We may say thatthe inherent variation generates a CHANCE CAUSE SYSTEM.
A chance cause system is one which is subject only to the inherentvariation of men, machines, materials and methods.
It will produce a certainpattern of variation.
If we can identify the pattern of variation , we can usethis information to predict future outcomes of the same combination of men,machines, materials and methods.
ASSIGNABLE CAUSE An assignable cause is one which changes the pattern of normal variation.
If an operator failed to follow instructions, or kept changing the machinesetting, we would say the operator was the assignable cause.
If a machine toolbroke, the machine would be the assignable cause.
If the material was of poor quality, it would be the assignable cause.
If the instructions were not clearand understandable, the method would be the assignable cause.
As long as a system operates with a "normal" amount of variation we saythat it is "in control".
If only inherent variation exists, there is little wecan do to improve the process without making major changes.
When an assignable cause enters into the system and creates excessivevariation, we say the process is "out of control".
If we can detect anassignable cause, identify it, and correct the condition, we have a means forcontrolling quality.
EVALUATION The primary function of statistics in the control of quality is
to identifythe pattern of variation within a known risk level, or limit.
Statisticalmethods, such as control charts, operate on the assumption that a certainpattern of variation does exist within a chance cause system.
The risk level isthe amount of chance or risk we take when we make an assumption about a process.
he controls and tests we use indicate the event that our assumption is correct.
As an example, control charts usually operate at the 99.
7% risk level.
This means that 3 times out of 1 000, an assignable cause would be sought eventhough one did not exist.
That is, the assumption that the process is "incontrol" would be rejected in favor of the assumption that an assignable causeis present because there is a good chance that it may be.
Experience teaches usthat machines and materials are common sources of trouble.
We would ratherassume that something has gone wrong and look for it, than assume that anunusual sample has occurred in a process that is in control.
HISTOGRAM AND THE NORMAL CURVE A.
CONSTRUCTING A HISTOGRAM We have discussed the importance of being able to observe data in someorderly fashion so that we might identify the pattern of variation.
The following figure illustrates the weight in grams of a plastic product.
In looking at the data it is difficult to detect anything other than a group of numbers.
WEIGHT IN GRAMS 25 25 21 28 21 23 28 24 20 23 26 27 26 24 29 24 25 25 24 27 22 26 27 25 26 22 24 25 27 24 23 27 22 24 22 25 28 29 23 25 30 25 26 22 25 27 26 20 24 26 If we use the technique of constructing a frequency chart, we can see apattern does indeed exist.
The following figure is an example of the above dataas it is tabulated on such a chart.
WEIGHT TALLY 30 I 29 II 28 III 27 IIIII I 26 IIIII II 25 IIIII IIIII 24 IIIII III 23 IIII 22 IIIII 21 II 20 II The following figure shows the data in the form of a frequency histogram,or what is commonly called a bar graph.
HISTOGRAM F 10 R 9 E 8 Q 7 U 6 E 5 N 4 C 3 Y 2 1 20 21 22 23 24 25 26 27 28 29 30