A wavelet neural network framework for diagnostics of complex engineered systems

Abstract

This paper introduces a new model-free diagnostic methodology to detect and identify machine failures and product defects. The basic module of the methodology is a novel multidimensional wavelet neural network construct used as the failure mode classifier. Validated sensor data are preprocessed and a vector of appropriate features is extracted. The feature vector becomes the input to the wavelet neural network which is trained off-line to map features to failure causes. An example is employed to illustrate the robustness and effectiveness of the proposed scheme.