This paper introduces a measure for selecting an appropriate pre-processing strategy for use in neural network-based condition monitoring and fault diagnosis (CMFD) applications. The proposed selection measure is derived from a non-parametric separability matrix: no knowledge of the underlying distribution of the data is required. The effectiveness of this measure is explored on a problem of engine misfire detection.
|Journal||International Journal of Knowledge-Based Intelligent Engineering Systems|
|Publication status||Published (in print/issue) - 2002|