论文部分内容阅读
研究了一种应用于精馏塔的多源信息融合的故障诊断方法。选择影响精馏塔系统生产质量的主要参数底温、顶温等作为融合对象。首先对参数进行归一化处理,并利用主元分析法对提取的底温、顶温等特征数据进行处理,这样既降低了输入的维数,同时又提高了输入参量特征的相互独立性。然后通过神经网络对主元分析法处理后的特征矢量进行推理分类,得到精馏塔的故障诊断结果。以在一定压力下,间接反映氯乙烯精馏塔产品浓度的塔板温度为诊断对象,仿真结果表明该方法具有很好的诊断效果。
A fault diagnosis method for multi-source information fusion applied to the rectification tower was studied. Select the main parameters affecting the quality of the distillation column system bottom temperature, the top temperature as a fusion target. Firstly, the parameters are normalized, and the principal component analysis (PCA) is used to process the characteristic data such as the bottom temperature and the top temperature, which not only reduces the input dimension but also improves the independence of the input parameters. Then, the feature vectors processed by the principal component analysis method are reasoned and classified by the neural network, and the fault diagnosis results of the rectifying tower are obtained. Under the certain pressure, the plate temperature, which indirectly reflects the product concentration of vinyl chloride rectification column, is the diagnostic object. The simulation results show that the method has a good diagnostic effect.