基于粒子群最小二乘支持向量机的故障诊断算法研究

For the problems of neural network for fault diagnosis, a method of fault diagnosis based on least squares support vector machines is researched. On the basis of this method, the least squares support vector machines fault diagnosis method based on particle swarm optimization is studied deeply.And then, this method is applied to the fault diagnosis of railway switch control circuit. The thesis focuses on the following aspects: First, a method of fault diagnosis based on neural network is researched. First of all,the neural network structure and learning algorithm are selected according to experience. And then the fault information date is used for training neural network. The neural network will achieve a certain degree of accuracy.The inputting fault date is classed by the trained neural network and the function of fault diagnosis can complete. Simulation result shows that the trained neural network can achieve high rate for fault diagnosis. But the inherent shortcomings of neural network make the fault diagnosis result unsatisfactory. Second, a method of fault diagnosis based on least squares support vector machines is researched. Least squares support vector machines is a machine learning method which based on structural risk minimization principle. It can solve the problem which emerged in neural network fault diagnosis better. By building the multi-class fault classifier, the least squares support vector machines can classify the inputting feature vector, determine the fault types and complete the function of fault diagnosis.The simulation result show that least squares support vector machines is better than neural network for fault diagnosis in fault recognition accuracy and anti-disturbance ability. Third, a method of fault diagnosis based on particle swarm optimization least squares support vector machines is researched. The two adjustable parameters of least squares support vector machines play an important role in fault diagnosis. The effects of finding optimal parameter combination through a lot of experiments are unsatisfied. On the basis of further analysis and many simulation studies for particle swarm optimization algorithm, the improved particle swarm optimization algorithm are applied to optimize the parameter combination of least squares support vector machines.The optimal parameter combination will be found by particle swarm optimization algorithm which makes the least squares support vector machines to achieve higher classification accuracy. A large number of simulations show that the classification accuracy rate of fault diagnosis based on particle swarm optimization least squares support vector machines algorithm has improved significantly. Four, particle swarm optimization least squares support vector machines is applied to the fault diagnosis of railway switch, control circuit. Based on the introduction of working principle of switch control circuit, take the fault diagnosis of five-wire high speed switch control circuit as an example.The attractive and repulsive particle swarm optimization is used for optimizing the parameter of least squares support vector machines and diagnosing the fault of railway switch control circuit. The result is satisfactory. The theoretical study and a large number of simulations prove that the method of the least squares support vector machines based on particle swarm optimization algorithm show a higher fault recognition accuracy in fault diagnose than other methods. It proves the effectiveness and superiority of this method.