Comprehensive experimental evaluation of a systematic approach for cost effective and rapid design of condition monitoring systems using Taguchi’s method

This paper provides an extensive experimental and analytical evaluation of a previously presented approach to the systematic design of condition monitoring systems for machining operations [J. Mater. Process. Technol. 107 (2000) 243]. The methodology termed automated sensor and signal processing selection (ASPS), is based on Taguchi’s orthogonal arrays in order to provide cost effective and speedy selection of sensors and signal processing methods that are ultimately used for monitoring process conditions. The evaluation using tool damage in end milling operations shows that ASPS methodology can successfully achieve its objectives without significantly affecting the system’s capability for fault detection. The experiments investigate two new types of cutting tools each with three distinct conditions which are processed by four different and independent neural network paradigms—two supervised and two unsupervised. Thus, the results confirm the feasibility and efficiency of the proposed ASPS methodology and show that it can be applied to condition monitoring systems without the need for implementing pattern recognition tools during the design phase.