摘要
BP算法在故障诊断领域已取得广泛应用,但其存在收敛速度慢且容易陷入局部最小值的缺陷,限制了其进一步的发展;ACO(Ant colony optimization)算法是一种模拟进化算法,已很好地应用于解决旅行商和资源两次分配等经典的优化问题,具有启发式收敛、正反馈以及分布式计算等优点;为此,将ACO算法引入BP算法故障诊断方法中,使用ACO算法对BP网络中的参数即权值、阈值以及学习率等进行优化,定义了一种结合ACO算法和BP算法能对故障进行诊断的新算法,并将其应用于具体的故障诊断实例中,最后,通过100组样本中的95组进行训练,并对剩余5组进行故障诊断,实验证明结合ACO算法和BP算法的新算法较传统的仅使用BP算法的诊断方法具有收敛速度快、诊断精确高以及训练性能好的优点。
BP atgorism has been widely applied in the fault diagnosis area, but it has the defect of slow convergence rate and getting the local minimum and can not be developed farther. ACO （Ant colony optimization） algorism is a simulated evolutionary algorism applied in sol- ving the classic problems such as the traveling salesman and twice resource allocation. ACO has the virtue of heuristic convergence, positive feedback and distribution computation. Therefore, ACO algorism is introduced to the fault diagnosis method, using ACO algorism to opti- mize the parameters of BP network such as weight and threshold, and then a new algorism combined ACO algorism and BP algorism is de- fined. Finally, 100 groups of samples are trained, and the experiment result shows the new method for fault diagnosis has the quick conver- gence speed, high precise diagnosis and the good training performance.
出处
《计算机测量与控制》
CSCD
北大核心
2012年第6期1460-1462,1466,共4页
Computer Measurement &Control
基金
中国地震局教师基金2011年度项目资助(20110115)
