摘要
对于常规BP神经网络,其收敛过程存在两个很大的缺陷:收敛速度慢;存在所谓“局部最小值”问题。文章采用了一种自适应学习速率动量梯度下降反向传播算法对BP神经网络进行训练,建立了智能诊断模型,并应用于某型坦克发动机的故障诊断,测试结果表明该方法相比常规BP算法能够更有效诊断发动机油路和气路的故障,从而为故障诊断及判定的自动化提供了一个新思路。
There are two major defects in the convergence process of a conventional BP nerve network, namely the slow convergence rate and the problem of local minimum. This paper adopts a reverse transmission calculation of self-adapting learning rate with a momentum gradient reduction, which sets up a new fault diagnosis model and then applies the reverse transmission algorithm to the fault diagnosis of a tank engine. The test result with improved BP algorithm is more efficient than that with the conventional one, thus providing a new approach to the fault diagnosis.
出处
《工业仪表与自动化装置》
2007年第3期45-48,共4页
Industrial Instrumentation & Automation
