摘要
在应用神经网络来判断透平在线状态的工作中，面对众多的相关过程参数及环境参数，如何从中为神经网络选择适当的输入参数是非常重要的。根据透平热力性能在线诊断的需要，本文研究了如何用遗传算法为神经网络寻找优化的输入组合以期达到少输入、快训练和准确回忆的目的。结果表明，由遗传算法选定的较少的参数作为诊断网络的输入即可判断透平性能状态的好坏。
In applying neural networks for the diagnosis of turbine online performance it is of vital importance to select proper input variables for the neural networks from a variety of related processing variables and environmental ones in order to ensure success.Genetic algorithms have been employed in this paper to guide the search for an optimal combination of inputs for the neural networks used to diagnose the turbine online performance with a view to achieving the criteria of fewer inputs,faster training and more accurate recall.The results of the present study have shown that the neural networks with fewer inputs selected by the genetic algorithms are capable of making an accurate diagnosis of the turbine online performance.
出处
《热能动力工程》
CAS
CSCD
北大核心
1998年第2期115-117,共3页
Journal of Engineering for Thermal Energy and Power
关键词
神经网络
遗传算法
透平
故障诊断
neural network,genetic algorithm,turbine performance,failure diagnosis
