Normalization approaches to manufacturing data

I’m hoping someone would be able to help with insight to a normalization approach, I am tasked with identifying the relative difficulty between stages in a manufacturing process.

The processes can be multi dimensional, where you might have steps 1,2,3 but within these steps there are different types of variants that can be produced. Each manufacturing process may consists of difference steps and variant combinations (not all steps need to be performed and each process may have different step, variant combinations).

Additionally. the output data may be reported in different unit of measures, e.g. early steps may be in KGs where later steps in piece counts (# no. of units).

What I am trying to do is create a relative difficulty analysis that will tell me how much more or less difficult {step 1 variant 1} is compared to any other step, e.g. {step 1 variant 2} or {step 3 variant 3} for what ever process is being run.

So I can easily say {step 1 variant 1} is +/- difficult than {step 4 variant 6}

I have a large data set of historical output performance, e.g kg/hour, units/hour etc. for each step and variant.

So far I have been able to normalize the variants within the steps but the difficulty is with comparing the variants across the steps.

Does anyone have any previous experience with an approach that could work here or any existing libraries/algorithms that could work?

• I do not know I understand your problem as well but, if you are interested to use a descriptive method instead of the mathematical method, you could try MCDM methods to compare steps and whose variants. I hope it will be helpful. – A.Omidi May 10 '20 at 0:16

If I understood your question correctly, you are trying to find the relative weight of each individual process-variant combination. It reminds me of the Analytical Hierarchy Process(AHP) in which we try to choose the best option(out of a given set of options) based on the importance or value of the options. In AHP a matrix of relative weights for each attribute has been calculated and using those matrices the decision-maker proceeds the process. You can easily find details of the process, for example in (1). Another approach that you can try (I doubt if it's technically feasible) is TOPSIS in which all the options are ranked based on their similarity to the ideal option.

On the other hand, you can normalize whatever weight you calculated to avoid dealing with the units.

(1) Goepel, Klaus D. "Implementing the analytic hierarchy process as a standard method for multi-criteria decision making in corporate enterprises–a new AHP excel template with multiple inputs." Proceedings of the international symposium on the analytic hierarchy process. Vol. 2013. Creative Decisions Foundation Kuala Lumpur, 2013.