KDnuggets : News : 2004 : n23 : item5 < PREVIOUS | NEXT >

Features

From: Lincoln Evans-Beauchamp
Date: 24 Nov 2004 20:14:55 -0800
Subject: Time-series data mining - relationship to manufacturing flaws?

We were just looking at your Time Series Poll Results, and saw the question on Multiple Time Dimensions.

This "Multiple Time Dimension" Time Series analysis is exactly how one searches for "problem" tools, ones that are responsible for manufacturing flaws, in a manufacturing process that has one or more "single tool" manufacturing steps. These "single tool" steps introduce places where there is no entropy in the analysis, in that all products flow through the same tool and no tool-to-tool comparisons can be made. What one ends up looking for the the "cleanness" of the introduction of the problem source. The tool with the cleanest problem introduction tends to be the source of the problem.

Comparing to Joe Beeden's question, each tool in the process is just another time steam to be analyzed. So in his analysis, there are two tools, one with real time and the other using book time. For example, the same analysis for a semiconductor fabrication process could be doing the same analysis across hundreds of tools and thus time streams.

Lincoln Evans-Beauchamp
CEO
EWA Systems


KDnuggets : News : 2004 : n23 : item5 < PREVIOUS | NEXT >

Copyright © 2004 KDnuggets.   Subscribe to KDnuggets News!