重油混烧控制系统的智能故障诊断

Intelligence Failure Detection for Heavy-oil Mixture-burning Control System

提出用离散Hopfield神经网络预测模型算法对传感器的故障进行检测;采用Hopfield神经网络与模糊评判结合方法,对阀门进行了故障诊断;提出一种改进离散Hopfield神经网络算法,该改进算法中,为使得样本构造记忆矩阵中包含抑制信息,将故障输入矢量和输出诊断矢量对由二值状态矢量对转换为两极对形式,构成离散Hopfield神经网络的输入-输出两极对.实时检测重油混烧控制系统的故障和异常,构造输入故障矢量,通过判断是否与神经网络原训练样本完全匹配或由计算贴近度确定一个最靠近的样本与之对应,给出混烧控制系统的故障诊断和故障信息提示.

By using the characters of discrete Hopfield neural networks, the paper put forward an algorithm from a prediction model of such networks, which is capable of detecting sensor failures. And by making a fault diagnosis to valves by adapting Hopfield neural networks and Fuzzy judgment, the paper then used this algorithm to improve its failure detection. In order to make the sample memory-matrix contain suppression information, this improved algorithm transforms the pair of fault input vector and output diagnosis vector from the. pair of two-value condition vectors into the form of two-pole pair, thus constructing a two-pole pair of input and output. This method was applied to the realization of fault and abnormity detection for the mixture-burning control system on line and for the structure input fault vector. By judging whether the results match with those of the original sample and determining the nearest sample to correspond, the paper finally presented some fault information of the adjustors and the values in the control system.