Statistics abound on the cost of sudden plant shutdowns due to sensor/actuator/equipment malfunction or other accidents. It is estimated that such unexpected outages can cost in excess of $1M per day and on the average they rob the plant of 7% of its annual capacity. A majority of such disruptions are due to preventable and common faults such as plugged lines, faulty sensors or actuators, inoperative alarm systems etc. They render the most sophisticated control policies useless.

Over the last decade the fields of multivariate statistics, machine learning methods developed by computer scientists and Bayesian inferencing methods have merged to develop powerful condition based monitoring systems for predictive fault detection and diagnosis. These sensor fusion methods with embedded digital intelligence are at a stage where such strategies are being considered for off-line and on-line deployment.    

This presentation will outline the field of sensor fusion wherein sensor data and process knowledge are combined to obtain a holistic picture of the plant. In summary sensor fusion is the application of signal processing methods, in the temporal as well as spectral domains, on a multitude and NOT singular sensor signals to detect incipient process abnormality before catastrophic breakdown is likely to occur.

Industrial case studies ranging from sheet-break prevention in the pulp and paper industry, monitoring polymer production facilities for reactor decomposition, and detection and diagnosis of plant wide oscillations will serve as illustrative examples to demonstrate the success of these methods and complement the technical presentation.