基于粒子群优化粒子滤波和CUDA加速的故障诊断方法

FAULT DIAGNOSIS METHOD BASED ON PSO PARTICLE FILTER AND CUDA ACCELERATION

在非线性系统中,粒子滤波需要大量粒子才能保证状态估计的准确度,这降低了算法的实时性,导致故障诊断的准确率和实时性不佳。针对该问题,提出基于GPU平台的粒子群优化粒子滤波(PSOPF)并行算法。通过分析PSOPF算法的并行性,设计并实现一种基于CUDA并行计算架构的PSOPF并行算法,利用大量的GPU线程对算法进行加速。为解决拒绝重采样对GPU全局内存的非合并访问带来的执行效率低问题,通过改进拒绝重采样并行算法,使线程束中的线程对同一内存区段中的粒子进行重采样,提高了其执行效率。通过对风力机组变桨距系统故障诊断验证了算法的有效性。实验结果表明,该方法可满足故障诊断准确率和实时性的要求。

In non-linear systems,particle filter requires a large number of particles to ensure the accuracy of state estimation,which reduces the real-time performance of the algorithm,resulting in poor accuracy and real-time performance of fault diagnosis.Aiming at this problem,this paper proposes a particle swarm optimization particle filtering(PSOPF)parallel algorithm based on GPU platform.By analyzing the parallelism of PSOPF algorithm,we designed and implemented a PSOPF parallel algorithm based on CUDA parallel computing architecture,which used a large number of GPU threads to accelerate the algorithm.In order to solve the problem of low execution efficiency caused by the non-coalescing access of the GPU global memory,we improved the parallel algorithm of rejection resampling.It made the threads in the thread bundle by resampling the particles in the same memory segment,and improved the execution efficiency.Finally,we validated the effectiveness of the algorithm through fault diagnosis of wind turbine pitch system.The experimental results show that this method can meet the requirements of fault diagnosis accuracy and real-time performance.