摘要
针对煤矿主通风机故障与征兆对应关系复杂的特点和利用传统BP网络进行故障诊断存在训练速度慢、易陷入局部极小的缺点,本文提出基于小波和概率神经网络的故障诊断方法。先利用时频两域有紧支撑能力的Mexican Hat小波变换故障信号并提取能量归一化故障特征向量;然后将概率神经网络作为诊断决策分类器,输出故障模式。该方法充分利用了概率神经网络计算简单、收敛快、新增样本无须重新训练的特点,而且通过小波特征提取有效的故障特征,减少了网络输入层节点数,降低网络规模,减少计算复杂度,加快了训练速度。经实际验证,此方法准确的诊断煤矿主通风机故障类型,具有速度快、精确度高的特点。
Owing to the complicated relationship between the faults and the corresponding symptoms of coal mine main ventilator and the general BP neural network’s shortcomings such as low learning speed, probability of local minimum point. This paper introduces a fault diagnosis based on wavelet and PNN, which firstly uses Mexican Hat wavelet to transform diagnosis signal and extract energy normalized vector, then uses PNN to output fault as a classifier. This method utilizes the PNN’s advantages, such as simple structure, fast speed and new trained samples can be added to PNN easily. Moreover, it decrease node numbers, network scale, and computation complexity and accelerate the training speed by way of extracting energy normalized vector. Proved by example that this way can diagnose the main coal mine ventilator’s type quickly and accurately.
出处
《电气技术》
2007年第4期21-24,共4页
Electrical Engineering
基金
辽宁省自然科学基金项目
20051206
辽宁省高校优秀人才基金项目
2005219005
作者简介
付华(1964),女,博士,教授,硕士生导师,辽宁工程技术大学电气系,长期从事故障诊断方面的研究.
