论文部分内容阅读
针对复杂化工过程具有的非线性、非高斯性和动态特征,提出了基于核独立成分分析(KICA)的模式匹配方法,用于动态过程监控和诊断。首先,利用滑动窗建立基准集与测试集的KICA模型,提取各自的核独立元:其次,融合余弦函数绝对值度量和距离度量,提出新的不相似度监控指标,识别训练与测试操作期间的相似模式,进行故障检测:最后,基于两类数据的核子空间之间的差异子空间,获得每个过程变量方向与该差异子空间之间的互信息,并定义新的非线性非高斯贡献度指标,进行故障诊断。基于污水处理过程的仿真结果表明,与主成分分析不相似度因子的方法、标准的独立成分分析(ICA)统计指标方法及标准的ICA T~2/SPE指标融合的贡献度方法相比,本文提出的方法具有更好的检测能力与故障诊断效果。
In view of the nonlinear, non-Gaussian and dynamic characteristics of complex chemical processes, a KICA-based pattern matching method is proposed for dynamic process monitoring and diagnosis. Firstly, the KICA model of dataset and test set is established by using sliding window, and their respective kernel independent elements are extracted. Secondly, the absolute measure and distance measure of cosine function are fused, and a new index of dissimilarity monitoring is proposed to identify the KICA model during training and test operation Similarity model for fault detection: Finally, we obtain the mutual information between the direction of each process variable and the difference subspace based on the difference subspace between the two types of data and define the new non-linear non-Gaussian contribution Indicators for troubleshooting. The simulation results based on the wastewater treatment process show that, compared with the method of the principal component analysis dissimilarity factor, the standard ICA statistical index method and the standard ICA T 2 / SPE index fusion contribution method, The proposed method has better detection ability and fault diagnosis effect.