| With the rapid development of electronic industry technology,the research on the related subjects of circuit system detection and diagnosis has been highly valued.The analog circuit is highly integrated,and if there is a problem,the circuit system will not work properly.In addition,it has obvious nonlinear and tolerance characteristics.If the conventional diagnostic methods and principles are used,it cannot meet the needs of fault diagnosis.Therefore,it is of great practical significance to explore more reasonable analog circuit fault diagnosis technology.The research in this paper is mainly based on Extreme Learning Machine(ELM),Auto Encoder(AE),Stacked Extreme Learning Machine(SELM),Stacked Kernel Extreme Learning Machine(SKELM)and Sailfish Optimizer(SFO).The Sallen-Key band-pass filter circuit,the two-stage four-op-amp biquad low-pass filter circuit,the Leap-frog filter circuit,and the logarithmic amplifier circuit are used as the circuit under test,and the simulated experimental environment is on Candence16.5 and Matlab2014 a carry out and set the resistance and capacitance tolerance of the circuit,build a fault set after sensitivity analysis,and then perform Monte Carlo analysis,collect the output waveform and save it as the circuit output data,and divide it into training set samples and test set samples as diagnostic models.A method for fault diagnosis of analog circuits based on the optimized SELM is proposed.The specific work and innovative methods involved are as follows:1.Specific work: This paper proposes a fault diagnosis algorithm for analog circuits based on SFO-optimised SELM,using the initial weights and biases of the SFO-optimised Extreme Learning Machine-Auto Encoder(ELM-AE)to construct an optimal SELM model,which is used to classify faults.Finally,the Sallen-Key band-pass filter circuit and the two-stage quad op-amp dual second-order low-pass filter circuit were used as the simulation experimental circuit,and the SELM optimized by Genetic Algorithm(GA)and Particle Swarm Optimization(PSO)were compared.The experimental results show that the SFO has a strong optimisation capability and can accurately diagnose faults,which proves the feasibility of the algorithm.For SELM,the optimisation of multilayer network parameters greatly increases the model training time and does not guarantee the generalisation capability of the whole model,so this paper proposes an SFO-optimised SKELM-based fault diagnosis algorithm for analogue circuits.A multilayer Kernel Extreme Learning Machine(KELM)is constructed by introducing a kernel function,and the weights of each layer of the SKELM are obtained by training the Kernel Extreme Learning Machine-Auto Encoder(KELM-AE)based on the KELM.However,the kernel parameters and the regularization coefficients of KELM-AE are artificially set,so the SFO is used to find the optimum of these two parameters and then the optimal SKELM model is built by layer-by-layer training.Finally,the Leap-frog filter circuit and logarithmic amplifier circuit are used as the simulation experimental circuit and further compared with SELM,SKELM and SFO optimised SELM.The results show that KELM-AE has powerful generalisation capability and can map fault features to high-dimensional feature space through non-linear mapping without separate feature extraction of faults,thus improve the accuracy of classification.2.Innovative method: In order to improve the accuracy of analog circuit fault diagnosis and to solve the problem of difficult selection of network hidden layer parameters,a new swarm intelligence optimization algorithm is proposed to optimize the SELM analog circuit fault diagnosis method based on SFO.SFO has the characteristics of fast convergence and high accuracy of optimization search.Therefore,the initial weights and biases of the SFO-seeking ELM-AE are used to form a SELM network by training the ELM-AE,so as to construct an optimal SELM model and improve the fault diagnosis accuracy of the analogue circuit.In order to improve the generalization ability as well as the learning speed of the forward neural network,an SFO-optimized SKELM-based fault diagnosis algorithm for analogue circuits is proposed.The KELM-AE is constructed by introducing kernel functions,then the kernel parameters and regularization coefficients of the KELM-AE are optimised by SFO,and finally the SKELM structure is built by layer-by-layer training to realise the classification of faults. |
