摘要
提出了一种采用改进型的BP神经网络诊断船舶主柴油机冷却系统故障的新方法。介绍了包括附加动量法和自适应学习速率法在内的改进型的BP算法,其中附加动量法是在每一个权值的变化上加上一项正比于前次权值变化量的值,不仅考虑了误差在梯度上的作用而且考虑了在曲面上变化趋势的影响;自适应学习速率根据误差函数值的变化对学习率进行实时调整,可以保证网络总是以最大的学习速率学习。给出了仿真实例,实验证明所提出的方法与传统方法相比有更好的实时性和诊断效率,也具有较好的故障诊断能力。
Some methods are put forward to improve Algorithm of BP Neural Network, and are applied to the diagnosis of hest diesel engine cooling system. This method includes the additional momentum algorithm and self- adaptive study speed algorithm. The additional momentum algorithm is adding an item proportional to the value of fore weights variation. The algorithm consists of not only the impact of error in the gradient, but also the impact of change trend of error. Self- adaptive study speed algorithm is real- time adjustment of study speed according to the variation of the error function value, which ensures that the network studies at maximal study speed. Finally, simulation results show that Algorithm of BP is more feasible and efficient in contrast to traditional methods, and has better fault diagnosis capacity.
出处
《茂名学院学报》
2007年第4期52-55,共4页
Journal of Maoming College
关键词
柴油机
神经网络
故障诊断
BP算法
diesel engine
neural networks
fault diagnosis
BP algorithm