论文部分内容阅读
工业间歇过程数据普遍具有多阶段、动态和非高斯特性,且轨迹不同步是其固有特征,针对上述问题,提出一种基于高斯混合模型-动态偏最小二乘(GMM-DPLS)的故障监测与质量预报新策略。采用GMM对过程数据进行聚类,客观反映不同阶段操作模态的数据分布特点,实现子阶段划分;针对子阶段不等长问题,采用动态时间规整(DTW)算法同步阶段轨迹,最后对同步后的子阶段分别建立DPLS模型。间歇发酵过程的应用实例表明该策略相比传统单一模型的DPLS方法,能显著提高故障监测效率和质量预报准确性。
The industrial intermittent process data generally have multi-stage, dynamic and non-Gaussian characteristics, and the trajectory is not synchronized is its inherent characteristics. In view of the above problems, a Gaussian mixture model based on dynamic partial least squares (GMM-DPLS) fault monitoring and Quality forecast new strategy. The process data are clustered by using GMM to objectively reflect the data distribution characteristics of the operating modes at different stages and to realize the sub-stage division. In the case of sub-stage unequal length problems, the dynamic time warping (DTW) algorithm is used to synchronize the stage trajectories. Finally, D sub-stages of the establishment of DPLS model. The application of batch fermentation process shows that this strategy can significantly improve the efficiency of fault monitoring and the accuracy of quality forecasting compared with the traditional single model DPLS method.