基于LE与ICROA-RVM的瓦斯传感器故障诊断

摘要

In order to solve the problem that the gas sensor diagnosis speed is slow and diagnosis accuracy is not high, this paper takes the common type gas sensor fault such as impact, drift, offset, and periodic fault as research object and proposes a pattern classification and identification of fault diagnosis of gas sensor method based on Laplacian eigenmaps (LE) and improved chemical reaction optimization algorithm (ICROA) optimized relevance vector machine (RVM) to achieve gas sensor fault diagnosis.Firstly, the manifold learning method LE is used to extract the nonlinear dimensionality reduction features from the high-dimensional original data space and extract fault features.The method largely preserves the overall geometric information in the original fault data; then the fault features are used as the RVM model input. The ICROA algorithm performs global optimization on the kernel function parameters of RVM mode, and uses the trained ICROA-RVM model to perform fault diagnosis on the test samples. The results show that the diagnosis method has the characteristics of high training speed and high identification accuracy. The recognition accuracy is above 96％, which can effectively improve the speed and accuracy of fault diagnosis.%针对瓦斯传感器故障诊断速度慢、诊断精度不高的问题,以常见的冲击型、漂移型、偏置型和周期型传感器输出故障作为研究对象,提出一种基于拉普拉斯特征映射(LE)和改进化学反应优化算法(ICROA)优化的相关向量机(RVM)进行模式分类与辨识,实现瓦斯传感器故障诊断.首先采用流形学习方法LE对高维原始数据空间进行非线性降维特征提取,提取故障特征,该方法极大地保留了原始数据中的整体几何信息;然后将故障特征作为RVM模型训练输入,利用ICROA算法对RVM模型的核参数进行全局寻优,将训练好的ICROA-RVM模型对测试样本进行故障诊断.实验结果表明:该诊断方法具有训练速度快,故障辨识精度高的特点,故障诊断正确率在96％以上,能够有效地提高瓦斯传感器故障诊断的速度和准确性.