CASCADE NEURAL-NETWORK-BASED FAULT CLASSIFIER FOR THREE-PHASE INDUCTION MOTOR
Induction motors are subject to different faults which, if undetected, may lead to serious machine failures. The objective of this design is to develop an alternative NN based fault-detection scheme that overcomes the limitations of the present schemes. The present schemes are costly and applicable for large motors furthermore, many design parameters are requested. Concerning long-time operating machines, these parameters cannot be available easily. In some existing schemes, either a detail mathematical model is required, or many features must be extracted, for which costly instrumentation is needed. In this scheme, only stator current is captured, and simple statistical parameters of the current waveform are used as inputs to detect the four conditions of the motor. Comparing with existing methods i.e NN-based fault-detection, a single network is used, the proposed method is simple, accurate, reliable, economical and uses a cascade connection of RBF–multilayer-perceptron (MLP) (RBF-MLP) network is developed to achieve a classification better accuracy.
In the design, the first layer of cascade NN, which is the RBF with conscience full competitive rule and Boxcar metric with nearly 36 cluster centers. For the second layer of the network, Momentum learning rule and Tanh transfer function give the optimal results. For generalization, the network is trained and tested rigorously. It has been found that network is able to detect the faults in induction motor with average classification accuracies above 90% when tested on testing data and CV data, respectively. Other performance measures are shown in the graphs. The training time required per epoch per exemplar is fast enough since the proposed classifier is to be used in real time, where measurement noise is anticipated, the robustness of the classifier to the noise is verified.
For a demonstration of cascade NN-based fault classifier, experimental results are used instead of simulation to make the classifier more practical. For design, the Matlab 2018a version is used.
Reference Paper: Cascade Neural-Network-Based Fault Classifier for Three-Phase Induction Motor
Author’s Name: Vilas N. Ghate and Sanjay V. Dudul
Source: IEEE-IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Year: 2011
Request source code for academic purpose, fill REQUEST FORM or contact +91 7904568456 by WhatsApp or info@verilogcourseteam.com, fee applicable.
SIMULATION VIDEO DEMO
