论文信息 - Process Fault Diagnosis using a Self-Adaptive Neural Network with On-Line Learning Capabilities

Process Fault Diagnosis using a Self-Adaptive Neural Network with On-Line Learning Capabilities

Abstract An approach to on-line learning and diagnosis of process fault conditions using a self-adaptive neural network is described. A radial basis function network structure incorporating local activation functions is utilised. Hidden and output nodes are automatically added to the network to accommodate new process faults after detection. Online adaptation is achieved using recursive linear algorithms to train localised network parameters. The adaptive neural network is applied to the detection and diagnosis of incipient faults on a controlled, multi-variable, chemical process simulation. Results illustrate the network operation and demonstrate the capability of the network to successfully learn and diagnose a range of gradual and abrupt process faults with different fault sizes

J. B. Gomm

[1] J. A. Leonard,et al. Radial basis function networks for classifying process faults , 1991, IEEE Control Systems.

[2] David M. Himmelblau,et al. FAULT CLASSIFICATION WITH THE AID OF ARTIFICIAL NEURAL NETWORKS , 1991 .

[3] Jie Zhang,et al. On-line process fault diagnosis using neural network techniques , 1992 .

[4] Tor Arne Johansen,et al. An Integrated Approach to On-line Fault Detection and Diagnosis - Including Artificial Neural Networks with Local Basis-Functions , 1992 .

[5] Venkat Venkatasubramanian,et al. Representing and diagnosing dynamic process data using neural networks , 1992 .

[6] John Moody,et al. Fast Learning in Networks of Locally-Tuned Processing Units , 1989, Neural Computation.