Fault diagnosis in thermoplastic injection molding via mold cavity pressure signal analysis

This dissertation presents the development of a novel approach for process monitoring and diagnostics in a thermoplastic injection molding process via cavity pressure signal analysis using the pattern recognition techniques. In this application of pattern recognition techniques for process fault diagnostic system development, principle component analysis is applied to reduce the dimensionality of the original signals; Process "fingerprints" were developed via wavelet decomposition of the "reduced" signal using multi-resolution analysis. Feature elements of the "fingerprints" were defined element-by-element statistical index comparison. The final feature elements were classified and interpreted with the aid of artificial neural networks, for process condition monitoring and fault diagnosis.; In thermoplastic injection molding, mold cavity pressure signals facilitate process monitoring and diagnosis. The "healthy" process condition and parameter settings which lead to a "healthy" cavity pressure profiles are obtained using design of experiments and response surface analysis. The diagnostic system developed is shown to be fast, robust and adaptable. This methodology is suitable for on-line use for fault diagnostics and the run-to-run process control.