[Total: 0    Average: 0/5]

This is the ultimate validation tool, which will allow us to check if the main cause is the main one or if the solution put in place is much better than the initial situation.

Introduction

Method B vs C is actually a comparison technique of 2 situations: b The ” better ” situation, and c the “current situation”. This is the ultimate validation tool, which will allow us to check if the main cause is the main one or if the solution put in place is much better than the initial situation.

The method part of the observation that most of the time we put in place improvements, we perform a test or two then we validate. And a few days or weeks later, the problem reappears.

The principle is rather simple: starting from this observation, the method will consist of comparing the current situation with the news several times by alternating new situation and old.

1-Define the 2 situations

We will start by defining the production situations: the B situation which is the one where we put our solution in place, and the C which is the original situation.

2-Perform a first Test

We go into a first product 3 elements in situation B and 3 elements in situation C. This is called the Six pack test.

3 – Repeat 2 times step 2

To make sure of the result of the previous step, we will reproduce it twice.

4-Analysis

The challenge here is to be able to reproduce the defect. So when we produced in the conditions of situation B, we did not generate the defect. And vice versa.

If on the 3 production cycles, we find these results, then we can conclude that our solution and/or the cause of our defect are well validated.

It is simply noted that performing 3 cycles is statistically low-power. If a doubt persists, it will be possible to increase the statistical power by increasing the number of pairs of samples.

5 – Statistically validate the results

To statistically validate the results, the Shainin method proposes to use the Tukey test. This test will allow us to assess the level of confidence that we can bring to our outcome. The graphical representation of this validation is as follows:

Statistically unreliable result

Statistically reliable result

The process for carrying out this test is as follows:

  1. Store 8 Good and 8 bad individuals in order of best at least good
  2. Count the number of good and bad on each side
  3. Make the sum of these 2 numbers
  4. Determine the level of confidence of the result

Total

Confidence level

<6

No confidence

6

90%

7

95%

10

99%

13

99,9%

Example

We take the example of the thickness material developed for the pairwise comparison. We had concluded that the braking quality was dependent on the thickness of the brake.

We’re going to validate this conclusion a priori by doing the Tukey test. We take at random 8 good and 8 bad parts. The following table is obtained:

Thickness of good parts

Thickness of bad parts

0,015

0,019

0,018

0,018

0,014

0,016

0,022

0,023

0,017

0,024

0,019

0,023

0,011

0,021

0,007

0,017

Individuals are classified by order and the following table is obtained:

Values

Good/Bad

0,007

Good

0,011

Good

0,014

Good

0,015

Good

0,016

Bad

0,017

Bad

0,017

Good

0,018

Bad

0,018

Good

0,019

Bad

0,019

Good

0,021

Bad

0,022

Good

0,023

Bad

0,023

Bad

0,024

Bad

We see in this ranking, that we have in a row 4 good heads and 3 bad at the end. By doing the addition, we get 4 + 3 = 7. It is concluded that our cause is well identified with a confidence level of 95%.

Source

R. D. Shainin (1990) – Analysis of experiments

K. R. Bhote, A. K. Bhote (2000) – World class quality

K. S. Vinay, P. Gowda, H. Ramakrishna (2004) – Industrial scrap reduction using Shainin technique

Share This