【摘要】 针对间歇过程数据的多阶段特性及复杂非线性特性,提出一种基于多阶段多核支持向量数据描述(MPMK-SVDD)的间歇过程故障检测方法。为充分挖掘间歇过程数据的多阶段信息,首先提出一种基于互信息相似矩阵的改进谱聚类方法,解决间歇过程数据集的多阶段划分问题。进一步考虑到单一核函数难以充分描述过程数据的复杂非线性问题,设计一种基于多重核函数和核参数的SVDD监控模型,并通过贝叶斯推理构造全局监测统计量,以实现过程故障的有效监控。以青霉素发酵过程为仿真研究对象,验证方法的有效性。结果表明,提出的方法比传统的SVDD方法能更有效地检测过程故障,具有更高的故障检出率。更多还原

【Abstract】 Aiming at the multi-phase characteristics and complex nonlinear characteristics of batch process data, a fault detection method using multi-phase and multi-kernel support vector data description(MPMK-SVDD) is proposed. In order to fully mine the multi-phase information of batch process data, an improved spectral clustering method based on mutual information similarity matrix is proposed to solve the problem of multi-phase partition of batch process data set. Considering the complexity and non-linearity of the process data that cannot be fully described by a single kernel function, a SVDD monitoring model based on multiple kernel functions and kernel parameters is designed, and the global monitoring statistics are constructed by Bayesian inference to detect process faults. The proposed method is verified by the simulation of penicillin fermentation process. The results show that the proposed method can detect process faults more effectively than the traditional SVDD method and has a higher fault detection rate.更多还原