Application of Neural Networks on the Detection of Sensor Failure During the Operation of a Control System

Neural computing is one of the fastest growing branches of artifical intelligence. Neural Nets, endowed with inherent parallelism hold great promise owing to their ability to capture highly nonlinear relationships. This paper discusses the use of the back-propagation neural net for failure cognition in chemical process systems. The backpropagation. paradigm along with traditional fault detection algorithms such as the finite intgral square error method and the nearest neighbor method are discussed. The algorithm is applied to an IMC controlled first order linear time invariant plant subject to high model uncertanity. Compared to traditional methods, the backpropagation technique is shown to be able to accurately discern the supercritical failures from their subcritical counterparts. The use of backpropagation fault detection systems in on-line adaptation of nonlinear plants has been investigated.

[1]  A. T. Miller,et al.  An Integrated Approach to Controls and Diagnostics: The 4-Parameter Controller , 1988, 1988 American Control Conference.

[2]  Robert L. Kosut,et al.  Robust Fault Detection: The Effect of Model Error , 1984, 1984 American Control Conference.

[3]  Robert Hecht-Nielsen,et al.  Applications of counterpropagation networks , 1988, Neural Networks.

[4]  T. McAvoy,et al.  Interpreting biosensor data via backpropagation , 1989, International 1989 Joint Conference on Neural Networks.

[5]  Terrence J. Sejnowski,et al.  Analysis of hidden units in a layered network trained to classify sonar targets , 1988, Neural Networks.

[6]  Robert A. Jacobs,et al.  Increased rates of convergence through learning rate adaptation , 1987, Neural Networks.