论文部分内容阅读
实际工业过程中往往包含不同运行工况,且每种工况数据一般不服从同一种分布.数据的多分布性和分布的不确定性使得传统的故障诊断方法难以获得满意的效果,因此提出一种基于局部邻域和贝叶斯推断的多工况故障诊断方法.首先,通过局部邻域标准化算法对多工况数据进行预处理;再利用ICA-PCA(independent component analysis and principal component analysis)方法分别对该数据集的高斯特性和非高斯特性进行分析处理,获得全局模型;然后结合贝叶斯推断将多个统计量组合成一个监测统计量,实现多工况过程的在线监测;最后通过数值例子和TE过程的仿真研究,验证了提出方法的可行性和有效性.
The actual industrial process often contains different operating conditions, and each working condition data generally does not obey the same distribution.The data distribution and the uncertainty of the distribution makes the traditional fault diagnosis method is difficult to obtain satisfactory results, therefore proposed a Based on the local neighborhood and Bayesian inference multi-condition fault diagnosis method.First, through the local neighborhood normalization algorithm for preprocessing multi-case data; and then use the independent component analysis and principal component analysis (ICA-PCA) method Then the Gaussian and non-Gaussian characteristics of the data set are analyzed and processed respectively to get the global model. Combined with Bayesian inference, a number of statistics are combined to form a monitoring statistic to realize the on-line monitoring of the multi-working process. Finally, The simulation of the example and the TE process verify the feasibility and effectiveness of the proposed method.