|
Abstract: A trend predicting is a key technology to achieve the
advanced predictive maintenance. A prediction of neural network model is
a new way to achieve the trend predicting. The present predicting models
of neural networks for working conditions prediction to rotary machinery
are comparatively poor in adaptability to environment and in accuracy of
predicting; In view for the problems, a new way of on-line optimization
for trend predicting is put forward. Using the capability of parallel
search with genetic algorithm (GA) dynamically optimizes the structure
parameters of BP network. The predicting model improved may dynamically
optimize the structure parameters according to different conditions. The
more satisfactory results of the on-line predicting are gained.
Key words: Rotary machinery Predicting model GA optimization
CLC No: TH113
国家自然科学基金资助项目(59775002). Received 20010816, received in revised form
20020425
|