Font Size: a A A

Analog Circuit Fault Diagnosis Based On Neural Network&Fruit Fly Optimization Algorithm

Posted on:2016-05-26Degree:DoctorType:Dissertation
Country:ChinaCandidate:W X YuFull Text:PDF
GTID:1108330473967103Subject:Electrical engineering
Abstract/Summary:PDF Full Text Request
With the development of modern electronic technology, people have become increasingly demanding on Analog Circuit Fault Diagnosis, In general, a large part of circuit fault from the analog circuit fault circuit module in an actual circuit system. According to the survey shows that more than 80% of the proposed circuit fault from the analog circuit blocks in the electronic equipment, along with increasingly high degree of integrated circuits, it is difficult to diagnose the difficulty gradually increasing. But the method of analog circuit fault diagnosis has seriously lagged behind, so the theory and technology research analog circuit fault diagnosis not only has important theoretical significance, but also has high practical utility value. Given the existence of analog circuit fault diversity and complexity, which resulted in the traditional fault diagnosis technology has been unable to meet the requirements of diagnostic results. Ensemble learning and Bionic intelligent computing, which have been successfully applied in pattern identification, provide a new solution for fault diagnosis for analog circuit. Combined with the modern test theory, signal processing and pattern identification theory, in this paper, in-depth study of several analog circuit fault diagnosis method. The main research contents and achievements of the dissertation are as follows:1.In this paper, we describe a kind of wavelet function based on Shannon function and Gauss function, there is a faster convergence rate and a strong approximation capability, so the construct wavelet basis could instead of function neural network activation function. At the same time, we have analyzed the advantages of structural wavelet network for fault diagnosis of analog circuits, and given the general process of fault diagnosis. From the experimental results, the construction of wavelet network convergence has achieved relatively good performance, gotten a smooth training performance curve, which showed the data of fault circuit were set to learn effectively by structure of wavelet network, in a few iteration it could achieve the specified performance indicators, but it should be noted that the fault diagnosis accuracy rate of only 94.44%,diagnostic accuracy should be improved.2. Based on construct wavelet neural network structure to analog circuit fault diagnosis method, we have proposed the use of FOA and construct wavelet neural network to establish a FOA-structure wavelet neural network algorithm to Diagnostic Fault Circuit. First of all, we give each group Fruit fly individual random initial direction and distance, and then we iterate the algorithm according to FOA, the constantly of iterative process to looking for the optimal value by the fitness function, the weights and thresholds optimized into the construct wavelet Neural networks. Finally, the test data would be brought into the construct wavelet neural network to test. Weights and thresholds could be optimized by FOA to iteration into the construct wavelet neural network, such as training samples showed better, faster network convergence and diagnostic accuracy rates are higher.3. FOA have a good ability to global optimization, LSSVM performance a superior recognition pattern, we have proposed FOA--LSSVM method of analog circuit fault diagnosis. After compared with PSO--LSSVM, we have found that there is difference in both mathematical algorithms, but they have the same fault diagnosis mode, Experiments have proved that the diagnostic accuracy of both methods is closed.So they are more efficient and reliable analog circuit fault diagnosis method, and have broad application prospects in analog circuit fault diagnosis.4. With powerful optimization algorithm function in FOA, the ergodicity, randomness, and etc. in chaos, we have proposed a method to analog circuit fault diagnosis by chaos theory and FOA--LSSVM mention. It is a further enhance to the FOA--LSSVM in circuit fault diagnosis. Simulation results show that this method is better than FOA--LSSVM apply to circuit fault diagnosis method.
Keywords/Search Tags:Fault diagnosis, Wavelet, FOA, LSSVM, Chaos
PDF Full Text Request
Related items