针对传统BP神经网络在故障诊断中存在的易陷入局部极值,对初值要求高等缺陷,将基于双链编码的量子遗传算法(Doub-le Chains Quantum Genetic Algorithm,DQGA)进行了改进,直接针对量子相位进行种群更新,优化空间限制在二维Hilbert的[0,π/2]区间内,改进算法在时间、存储量性能上有了明显改进;将该算法用于优化BP神经网络,提出了一种结合DQGA算法与BP神经网络对整流电路故障进行诊断的方法;仿真结果表明,与单BP、GA-BP算法相比,该方法在整流电路故障诊断中诊断精度高,收敛速度快,避免了BP算法易陷入局部极值的缺陷,适合故障自动诊断系统的建立.%Traditional BP neural network was easily trapped in local minimum and high requirements on the initial value in fault diagnosis, Improved the quantum genetic algorithm which based on double-chains coding, updated the population with quantum phase , limited the optimization space in [0, π/2] of two-dimensional Hilbert, and the algorithm got apparent improvement of time and storage , then used the improvement algorithm to optimize the weights and thresholds of BP neural network. Proposed a fault diagnosis method for rectifier circuit which combined with neural network and quantum genetic algorithm . Simulation results show that the method is accurate and reliable by compared the simulation results with BP algorithm, GA-BP algorithm, and it is suit for building of the automatic fault diagnosis system.
展开▼
