Reconfigurable multi-stage neural networks in monitoring industrial machines

A two-stage reconfigurable neural networks (NN) is described for real-time monitoring of onset of faults in a coolant system of a CNC machine. The measured variables in the system are current and pressure signals. The steady state values of these parameters when out of healthy range, are used as stimulus for initiating a non-destructive test. This causes the closure of a flow control valve and results in the transient response of the pump outlet pressure. The transient signal is used as input to the NN which accurately identifies inception of any faults in the system. If the system is faulty, an interprocess communication system (IPC) activates the second stage of the two-stage NN which then tests the transient pattern against the known types of failure and identifies severity of the fault. The double stage design of neural network results in achieving a high accuracy of over 99 percent in fault identification and isolation.

[1]  Jing Peng,et al.  LDA/SVM driven nearest neighbor classification , 2003, IEEE Trans. Neural Networks.

[2]  Rolf Isermann,et al.  Model Based Fault Diagnosis and Supervision of Machines and Drives , 1990 .

[3]  Jean-Pierre Martens,et al.  An equalized error backpropagation algorithm for the on-line training of multilayer perceptrons , 2002, IEEE Trans. Neural Networks.

[4]  S Z Qin,et al.  Comparison of four neural net learning methods for dynamic system identification , 1992, IEEE Trans. Neural Networks.

[5]  James L. McClelland.  Explorations In Parallel Distributed Processing , 1988 .

[6]  Hosein Marzi,et al.  Real-time fault detection and isolation in industrial machines using learning vector quantization , 2004 .

[7]  K. F. Martin,et al.  Neural Network Solution to Coolant System Diagnostics , 1993 .

[8]  Mo M. Jamshidi,et al.  Neural networks in robotics: A survey , 1990, J. Intell. Robotic Syst..

[9]  Richard S. Sutton,et al.  Neuronlike adaptive elements that can solve difficult learning control problems , 1983, IEEE Transactions on Systems, Man, and Cybernetics.

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

[11]  K. F. Martin,et al.  Coolant system health monitoring and fault diagnosis via health parameters and fault dictionary , 1990 .

[12]  Janos J. Gertler,et al.  Analytical Redundancy Methods in Fault Detection and Isolation , 1991 .

[13]  James L. McClelland,et al.  Explorations in parallel distributed processing: a handbook of models, programs, and exercises , 1988 .

[14]  Venkat Venkatasubramanian,et al.  On the nature of fault space classification structure developed by neural networks , 1992 .

[15]  O. J. Murphy,et al.  Nearest neighbor pattern classification perceptrons , 1990, Proc. IEEE.

[16]  Xiaoli Li,et al.  Utilization of information maximum for condition monitoring with applications in a Machining Process and a water pump , 2004, IEEE/ASME Transactions on Mechatronics.

[17]  K F Martin,et al.  Diagnostics of a coolant system via neural networks , 1999 .