基于改进PSO-PNN的大螺旋钻机故障诊断系统研究

Fault diagnosis system of large auger based on improved PSO-PNN

为解决当前大螺旋钻机故障诊断方法存在的准确率较低问题,提出一种将改进PSO(粒子群优化算法)和PNN(概率神经网络)相结合的大螺旋钻机故障诊断系统。首先通过减小惯性因子和学习因子,间接实现粒子速度由大到小的调整,实现对粒子群优化算法的改进,通过基准函数测试证明改进PSO的收敛速度、精度、全局寻优能力均优于GA(遗传算法)、WOA(鲸鱼优化算法)、PSO等常规优化算法。然后利用改进PSO搜索PNN可满足整个样本空间预测需求的全局最优平滑因子,并加载到PNN。实验结果表明,在诊断精度和实时性方面,分别与经验法选取平滑因子的PNN和由GA、WOA、PSO优化后的PNN进行对比,通过改进PSO优化后的PNN故障诊断准确率达到97.5%,同时优化后的PNN运行速度较快,对单组故障数据分析时间为0.785s,以上说明基于改进PSO-PNN的大螺旋钻机故障诊断系统可满足大螺旋钻机对故障诊断准确率和实时性的需求。

Aiming at the low accuracy in the current fault diagnosis method for 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%.And the optimized PNN runs faster,and the analysis time of single group of fault data is 0.785 seconds.The fault diagnosis system of large screw drill based on improved PSO-PNN can meet the requirements of accuracy and real-time performance.