论文部分内容阅读
针对微生物制药的间歇生产过程中缓变故障难于监测的问题,提出了多向核熵成分分析(multi-way kernel entropy component analysis,MKECA)过程监测的新方法,克服了传统多向核主成分分析(multi-way kernel principal component analysis,MKPCA)方法在监控缓变故障时漏报率高的缺陷.该方法首先将3维历史数据按照本文所提的3步法进行预处理,然后通过核映射将数据从低维空间映射到高维特征空间,解决数据的非线性特性,并在高维特征空间依据核熵的大小对数据进行降维,使降维后的数据分布与原点成一定的角度,能够逼近原始间歇过程的数据分布.通过数值实例和实际工厂数据对方法进行验证.结果表明,MKECA方法具有更可靠的监控性能,能及时、准确地监测出故障,具有广阔的应用前景.
Aiming at the problem that the gradual change of microorganism pharmaceutical is difficult to be monitored in the process of intermittent production, a new method of process monitoring of multi-way kernel entropy component analysis (MKECA) is proposed, which overcomes the traditional multi-directional kernel principal component analysis (Multi-way Kernel Principal Component Analysis, MKPCA) method is used to monitor slow faults with high false alarm rate.The method firstly preprocesses the 3-D historical data according to the 3-step method proposed in this paper and then uses kernel mapping The data is mapped from low-dimensional space to high-dimensional feature space to solve the non-linear characteristic of data. In the high-dimensional feature space, the data is reduced in dimension according to the size of kernel entropy, so that the data distribution after dimension reduction is at an angle with the origin, Which can approximate the data distribution of the original batch process.The method is validated by numerical examples and actual factory data.The results show that the MKECA method has more reliable monitoring performance and can timely and accurately monitor the fault and has broad application prospects.