Extreme Learning Machine Model For Multi-Fault Diagnosis of Analog Circuits
摘要
Due to analog circuits’ complexity and characteristics, their fault diagnosis has become a bottleneck problem. In this paper, we have investigated multiple fault diagnosis using Extreme Learning Machine technique to determine which part of an analog signal conditioning circuits is malfunctioning. To check its viability we have compared the performance of this model with the other existing neural network models such as radial basis function neural network and support vector machine. As a case study we have considered sallen-key bandpass filter circuit to comparatively investigate the above mentioned algorithms. We made multiple passive elements faulty in that circuit to find the potential of multiple-fault diagnosis using Extreme Learning Machine. For training and testing, we applied a single pulse as an input to that analog circuit and collected raw input-output data. The outcome of the computational experiments give