关键词: 改进粒子群优化算法, 概率神经网络, 平滑因子, 大螺旋钻机, 故障诊断

Abstract: In order to solve the problem of low accuracy in the current fault diagnosis method of large screw drilling rig, a fault diagnosis system of large screw drilling rig based on the combination of improved PSO (Particle Swarm Optimization Algorithm) and PNN (Probabilistic Neural Network) is proposed. Firstly, by reducing the inertia factor and learning factor, the particle velocity is indirectly adjusted from large to small, and the particle swarm optimization algorithm is improved. The benchmark function test shows that the convergence speed, accuracy and global optimization ability of the improved PSO are better than GA (Genetic Algorithm), WOA (Whale Optimization Algorithm), PSO and other conventional optimization algorithms. Then, the improved PSO is used to search the global optimal smoothing factor of PNN that can meet the prediction demand of the whole sample space, and loaded into PNN. The experimental results show that in terms of diagnosis accuracy and real-time performance, it is compared with the PNN with smoothing factor selected by empirical method and the PNN optimized by GA, WOA and PSO respectively. The fault diagnosis accuracy of PNN optimized by PSO is 97.5%. At the same time, the optimized PNN runs faster, and the analysis time of single group of fault data is 0.785 seconds, The above shows that the fault diagnosis system of large screw drill based on improved PSO-PNN can meet the requirements of large screw drill for fault diagnosis accuracy and real-time.