A Simple Process Validation Example


Consider the bread making process shown in the figure below. Someone developed it for making bread. The bread made using this process has various characteristics that consumers find desirable: look, feel, taste, etc. Each of these characteristic can be measured and will have some target value (based on consumer research) such that if a loaf is made with all its characteristics on target, there is a high probability that it will meet the consumer’s expectation and the consumer will enjoy eating it. The question that process validation seeks to answer is will this process consistently produce bread loaves of the specified quality.

A fundamental assumption in manufacturing is that if the inputs to the process remain constant e.g. you use exactly 6 cups of bread flour each time, and the process itself is constant e.g. the oven generates 350 F of heat every time, then each output of the process will be the same as the previous with no discernable difference. However, nothing is constant: there is natural variability in the quantity of the flour used; sometimes you might use as little as 5.5 cups; other times you might use as much as 6.5 cups. Even the oven periodically turns its heating mechanism on and off to provide a mean temperature of 350 F, but the actual temperature at any given instant is more likely than not to be above or below the mean. So in the physical world each output of the process i.e. each loaf of bread will be different from the previous.

The question then becomes is the loaf to loaf variability in the output, the result of the variability in the inputs and the process, noticeable by the consumer? Each characteristic of the output not only has a target value but also a range about the target that is considered acceptable. The bread may be okay if its crust is slightly more or less brown, but rejected if it is significantly dark (suggesting burnt) or light (suggesting underdone). What exactly are the limits of acceptability for each characteristic? That is decided through consumer research. Assuming, for our purposes that these limits are already specified, then if the measured value of a particular characteristic for a given loaf of bread falls within its upper specification limit and lower specification limit, it is considered acceptable.

During process validation the process is kept constant i.e. step sequence, parameter settings, etc. are fixed, while its inputs are varied between their extreme possible conditions. The thought is if the output of the process subjected to such extreme conditions of its inputs is within acceptable limits, then the output of the process with normal conditions of inputs will also be acceptable. The intent of this exercise is to demonstrate the robustness of the process to the natural variations in its inputs.

The design of experiments provides an efficient way to simultaneously vary every input between its extremes. For the bread making process in this example, there are 6 inputs: amount of bread flour, salt, vegetable oil, active dry yeast, white sugar and water. If we assume that each of these inputs will vary from their specified quantity as shown in the table below, then we can construct a two level six factor experiment for the process validation study.

 

 

Low (-)

High (+)

A

Bread flour (cups)

5.75

6.25

B

Salt (teaspoon)

1.25

1.75

C

Vegetable oil (cups)

3/16

5/16

D

Active dry yeast (tablespoon)

1.25

1.75

E

White sugar

5/9

7/9

F

100F warm water

1.75

2.25

Such an experiment is referred to as a full factorial experiment i.e. one where every combination of high and low values of every factor is made. Each combination will then be run through the process in randomized order. And each resulting loaf of bread will have various quality characteristics measured e.g. look (I), feel (II), and taste (III). These measured values will be plotted on separate run charts with their respective specification limits drawn in. The expectation is that the actual values will all fall within the spec limits. If that is the case, we can state with confidence that as long as the input variables remain within the upper and lower limits of their respective specifications, the quality characteristics of the resulting output will also be within their respective specification limits. And, thus we can conclude that the process is validated… for the set of inputs specifications defined.

Note: A full factorial experiment is not necessary. I will have more to say about this in another post.

References

Appendix – Full factorial experiment design (order not randomized)

A

B

C

D

E

F

I

II

III

1

+

+

+

+

+

+

2

+

+

+

+

+

-

3

+

+

+

+

-

+

4

+

+

+

+

-

-

5

+

+

+

-

+

+

6

+

+

+

-

+

-

7

+

+

+

-

-

+

8

+

+

+

-

-

-

9

+

+

-

+

+

+

10

+

+

-

+

+

-

11

+

+

-

+

-

+

12

+

+

-

+

-

-

13

+

+

-

-

+

+

14

+

+

-

-

+

-

15

+

+

-

-

-

+

16

+

+

-

-

-

-

17

+

-

+

+

+

+

18

+

-

+

+

+

-

19

+

-

+

+

-

+

20

+

-

+

+

-

-

21

+

-

+

-

+

+

22

+

-

+

-

+

-

23

+

-

+

-

-

+

24

+

-

+

-

-

-

25

+

-

-

+

+

+

26

+

-

-

+

+

-

27

+

-

-

+

-

+

28

+

-

-

+

-

-

29

+

-

-

-

+

+

30

+

-

-

-

+

-

31

+

-

-

-

-

+

32

+

-

-

-

-

-

33

-

+

+

+

+

+

34

-

+

+

+

+

-

35

-

+

+

+

-

+

36

-

+

+

+

-

-

37

-

+

+

-

+

+

38

-

+

+

-

+

-

39

-

+

+

-

-

+

40

-

+

+

-

-

-

41

-

+

-

+

+

+

42

-

+

-

+

+

-

43

-

+

-

+

-

+

44

-

+

-

+

-

-

45

-

+

-

-

+

+

46

-

+

-

-

+

-

47

-

+

-

-

-

+

48

-

+

-

-

-

-

49

-

-

+

+

+

+

50

-

-

+

+

+

-

51

-

-

+

+

-

+

52

-

-

+

+

-

-

53

-

-

+

-

+

+

54

-

-

+

-

+

-

55

-

-

+

-

-

+

56

-

-

+

-

-

-

57

-

-

-

+

+

+

58

-

-

-

+

+

-

59

-

-

-

+

-

+

60

-

-

-

+

-

-

61

-

-

-

-

+

+

62

-

-

-

-

+

-

63

-

-

-

-

-

+

64

-

-

-

-

-

-

About these ads

5 responses

  1. Really good notification to understand the word”validation”

  2. Perfect article. Just wondering, were to find an unswer to this one:
    Note: A full factorial experiment is not necessary. I will have more to say about this in another post.

  3. Hi Tomas,

    Thanks for your comment. I’m glad you found the post useful.

    I haven’t written the follow-up to the article yet. It sort of fell off my radar. However, I can provide a brief answer for you.

    As I understand things, for process validation, if you have identified the critical inputs to your process, then it is necessary to conduct the full factorial experiment. This is because every extreme (or ‘corner’) of the inputs must be investigated.

    Full factorial experiments are not necessary if you are designing a product or process. The better way is to run a sequence of partial factorial experiments. An excellent guide to industrial experimentation is Dr. Donald J Wheeler’s book “Understanding Industrial Experimentation” (http://www.spcpress.com/book_understanding_indust_exp.php)

    Best regards,
    Shrikant Kalegaonkar

  4. Hi Shrikant, thank You for the comments and suggestions. Also regarding software to help with process validation and DOE ? Would you be able to suggest or recommend ?

    Kind Regards,
    Tomas

    1. Hi Tomas,

      I have to tell you I am not a big fan of software for statistics. I personally use a spreadsheet program like MS Excel or Libre Calc.

      Having said that, I’ve used Minitab and JMP. They are both powerful programs with comparable capabilities. However, they are not cheap and I am not convinced they are worth the cost for something you can easily do yourself in a spreadsheet program.

      Best wishes on your choice,
      Shrikant

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

%d bloggers like this: