关键词: 示功图, 动态诊断, 过程神经网络, 学习算法, 应用

Abstract: Aiming at the pattern diagnosis problem of the indicator diagram for pumping wells，a dynamic diagnosis method based on process neural networks is presented in this paper.The inputs and connection weights of Process Neural Networks（PNN） can be time-varying functions.The process pattern characteristics of the set for time-varying functions can be extracted automatically by the learning of the sample set for functions，and the combination of several process pattern characteristics can generate the output types.The method on mechanism has good adaptability to solve classification problems of time-varying signals.Combined with gradient descent algorithm，a dynamic diagnosis model based on PNN is built and the learning algorithm based on function basis expansion is given，which are applied to identify the working states of the indicator diagram for the observed pumping wells and show good effects.

Key words: indicator diagram, dynamic diagnosis, Process Neural Networks（PNN）, learning algorithm, application