Value Driven Design Logo

6.19 - Rank Order Clustering (ROC) Analysis - Parts

In Modular Function Deployment (MFD) the aim is to design from scratch functional and physical modules (see guide 6.18). However, we may have products that are going to be in production for many years to come and we need to do what we can to manage the current situation efficiently. Rank Order Clustering can used to find common parts for the products we have now. The process is similar to that described in guide 5.35 for finding common processes. This time we have assemblies A, B, C, D, & E.

Assemblies Right Arrow


A

B

C

D

E

Totals

↓ Parts

Binary

Dec.

1

1

 

1

1

 

10110

22

2

 

1

   

1

01011

9

3

1

 

1

1

 

10110

22

4

1

 

1

1

 

10110

22

5

       

1

00001

1

6

1

1

1

1

1

11111

31

7

 

1

   

1

01001

9

8

1

   

1

 

10010

18

9

1

 

1

1

 

10110

22

10

1

 

1

1

 

10110

22

11

 

1

   

1

01001

9

12

1

 

1

1

 

10110

22

13

 

1

   

1

01001

9

14

 

1

   

1

01001

9

Binary Total

             

Decimal

             

We can quickly see that parts 1, 3, 4, 9, 10, & 12 have the same binary score indicating they are common to some of the assemblies (A, C, & D).

The next stage is to re-order the parts according to their binary totals thus:-

Assemblies Right Arrow


A

B

C

D

E

Totals

↓ Parts

Binary

Dec.

6

1

1

1

1

1

11111

31

1

1

 

1

1

 

10110

22

3

1

 

1

1

 

10110

22

4

1

 

1

1

 

10110

22

9

1

 

1

1

 

10110

22

10

1

 

1

1

 

10110

22

12

1

 

1

1

 

10110

22

8

1

   

1

 

10010

18

2

 

1

   

1

01011

9

7

 

1

   

1

01001

9

11

 

1

   

1

01001

9

13

 

1

   

1

01001

9

14

 

1

   

1

01001

9

5

       

1

00001

1

Binary Total

             

Decimal

16320

8254

16256

16320

8255

   

Sort the columns by their totals and we get the result on the next chart. We have part groups (brown, green and yellow) highlighted as shown.

Notice that ALL the assemblies contain part 6. This is a common part. The brown group is common to assemblies A C & D and the green group is common to assemblies B & E.

Assemblies Right Arrow


A

D

C

E

B

Totals

↓ Parts

Binary

Dec.

6

1

1

1

1

1

11111

31

1

1

1

1

   

10110

22

3

1

1

1

   

10110

22

4

1

1

1

   

10110

22

9

1

1

1

   

10110

22

10

1

1

1

   

10110

22

12

1

1

1

   

10110

22

8

1

1

     

10010

18

2

     

1

1

1011

9

7

     

1

1

1001

9

11

     

1

1

1001

9

13

     

1

1

1001

9

14

     

1

1

1001

9

5

     

1

 

1

1

Binary Total

             

Decimal

16320

16320

16256

8255

8254

   

Part 8 is common to A & D. Part 5 is unique to assembly E. After sorting, the binary and decimal values for the rows actually change their values, but it does not affect the results. By using text concatenation and binary to decimal conversion add-ins in Excel it is possible to manipulate some fairly large matrices in this manner. For information on how to manage common part groups in similar assemblies to assist with part rationalisation and modular design see guide 6.20.

BOM Efficiency Improvements

The product structure for the original arrangement would be:-

Assembly A – 1, 3, 4, 6, 8, 9, 10, 12

Assembly B – 2, 6, 7, 11, 12, 13, 14

Assembly C – 1, 3, 4, 9, 10, 12

Assembly D – 1, 3, 4, 6, 8, 9, 10, 12

Assembly E – 2, 5, 6, 7, 11, 12, 13, 14

Assembly A and D are identical – this is not an unusual finding where little, or no, attention has been paid to modularisation. So of the total of 37 structure records by modularisation we can eliminate assembly A or D all together. If we create sub assemblies green and brown, we get the following:-

Sub-assembly (or phantom) Brown – 1, 3, 4, 9, 10, 12

Sub-assembly (or phantom) Green – 2, 7, 11, 13, 14

The original assemblies now look like this:-

Assembly A – Brown, 6, 8

Assembly C – Brown, 6

Assembly E – Green, 6, 5

Assembly B – Green, 6

The total product structures have been reduced to 21 from 37 – almost a 50% reduction. This affects computer processing time to generate requirements and BOM displays etc. Assembly kitting is easier and the BOM maintenance time is reduced and engineering change is much easier to manage.

Some BOM systems have a reporting capability called a matrix BOM. This is similar to ROC method described here. The user schedules the report with the part numbers of the assemblies s/he wishes to compare and the number of levels the bill should be exploded too. A matrix report is generated with components down the page and assemblies across the page. However, it is up to the user to do the clustering to determine what the natural option groups could be. The matrix report could be imported into a spreadsheet tool for clustering.

The next guide, modularising the BOM (guide 6.20), shows what else can be done to improve the BOM structure.

 
Further Reading