|M.Sc Student||Eilam Rachel|
|Subject||Manufacturing Process Characterization Using Comparative|
Analysis of Poisson and Binomial Models
|Department||Department of Quality Assurance and Reliability||Supervisor||DR. Pavel Grabov|
The postulate in this work is that the manner by which quality reports are analyzed shall serve as the basis for manufacturing process characterization, and derive the proper recommendations for the process improvement.
It is plausible to characterize a manufacturing process by using comparative analysis of indices from two main families: one based on “Defectives”, using the binomial model, and the other based on “Defects”, using the Poisson model.
In this work it is shown that when ‘actual rate of fitting units’ represented by “First Time Yield” (FTY) is greater than the ‘expected rate of fitting units’ represented by “Throughput Yield” (TPY), a chain of defects exists, where one defect causes another. Otherwise, (when TPY > FTY), a permanent damage of numerous items exists. It is named “Regular Defects” process.
An index for defects concentration is developed in this work, named: 'Total Defects per Defective' (TDD). When this index takes a high value, it reveals the existence of correlation between defects in the process. This is an indication for a “chain of defects”; therefore TDD may be used as a pre-processing tool for Pareto analysis.
Simulations performed in order to assess the use of proposed model, showed that while dependency exists between defects, FTY>TPY and TDD>1.2. On the other hand, when there is a regular defect then TPY>FTY, and <1.2. Simulations have also shown high correlation between TDD index and the difference between FTY and TPY.
The model was tested on true data of Israeli industry, and the results, presented in this work, uphold the proposed model.