一种用于动态过程故障诊断的子波-S型神经网络

FAULT DIAGNOSIS IN DYNAMIC PROCESSES USING A NEW TYPE OF WAVELET - SIGMOIDAL FUNCTION BASED NEURAL NETWORKS

在分析子波变换的基础上,将子波理论引入神经网络,提出了一种新的神经网络模型─—子波-S型神经网络(WSNN),并将其成功地应用于动态过程的故障诊断,同时根据其特性给出训练方法, 实验证明,与S形作用函数的前向阶层型神经网络(SBFN)相比,子波-S型神经网络提高了故障早期诊断的正确诊断率。

For improving the correctness rate of dynamic fault diagnosis, a new artificial neuralnetwork——wavelet-sigmoidal function based neural network(WSNN) is proposed after introducing wavelet transforms, and is successfully applied to dynamic process fault diagnosis. The training algorithm is also presented based on its error surface properties. In this paper, WSNN is compared with the traditional sigmoidal function based neural network (SBFN) in terms of the correctness rate of fault diagnosis.