基于概率密度PCA的多模态过程故障检测

为了提高故障检测和分类能力，提出基于概率密度PCA的多模态过程故障检测算法。对各模态的训练数据建立PCA模型，计算各个模型的控制限和匹配系数。根据匹配系数计算各模态统一的控制限。对新来的数据，运用概率密度确定其模态。新来数据向对应模态的模型上投影并计算统一的统计量，比较统计量与控制限进行多模态过程故障检测。把该方法应用到数值例子和半导体过程中，仿真结果表明，该算法在分类及多模态过程故障检测方面具有很高的准确性。

PCA based on probability density for fault detection of multimodal processes

In order to improve the ability of fault detection and classification, this paper proposed PCA based on probability density for fault detection of multimodal processes. It established PCA model for training data of each mode, and calculated the control limits and matching coefficients of each model. It calculated the unified control limit of each mode according to the matching coefficients. For a new data, it determined its mode by the probability density. It projected the new data to PCA model of the corresponding mode and calculated the unified statistics. It performed fault detection of multimodal processes by comparing the statistics with control limit. We applied the method to a numerical example and the semiconductor process. Simulation results show that the proposed algorithm has high accuracy in classification and fault detection of multimodal processes.