Surveillance and analysis of multivariate processes using digital signal processing

Digital signal processing (DSP) techniques are being developed and applied in many industries including; communications systems, aerospace systems, chemical, metals, process industries, and power generating systems. As the choice of techniques increases, it is necessary to integrate the various techniques and provide guidelines to the user. The purpose of this research is to develop an integrated DSP system and demonstrate its applications to monitoring and diagnostics of power and process industry systems. A modular software system that integrates classical and modern signal processing techniques was developed and tested on a personal computer (PC) platform. Both classical and modern digital signal processing methods were included. An approach was developed to systematically use the techniques to detect and isolate process and sensor anomalies. The "modern" digital signal processing techniques include univariate and multivariate autoregression time-series modeling, diagnostics based on spectral features and statistical methods, and nonstationary data analysis using parametric models. The classical signal processing techniques include correlation and Fourier transform approaches. The various methods were first tested using simulated data and then applied to data from nuclear power plants and from industrial rotating machinery. Test data, that include start-up and coast-down sequences, were also used for establishing the performance of nonstationary models. The DSP system was implemented on a PC with a user-friendly interface and graphical display of information.