一种基于局部模型的多工况过程质量预测方法

Process quality predict method of multiple loading conditions based on local models

根据不确定、非线性复杂生产过程的质量预测难点,提出一种基于局部模型的多工况过程质量预测方法。首先利用K-均值聚类算法对过程进行工况划分,结合支持向量机回归原理建立各局部质量预测模型,再利用改进粒子群算法求解最优的各局部模型权重,使全局模型的输出与预期输出之间的误差达到最小,以得到其全局模型,进而实现生产过程的质量预测。该方法较好地解决了复杂生产过程的复杂不确定的问题,同时有良好的全局适应性。最后以Tennessee Eastman(TE)为例,实现其生产过程的建模和质量预测,结果表明了该方法的可行性和有效性。

As for quality prediction difficulty that complex production process has the characteristics of uncertainty and nonlinear,this paper proposed a process quality prediction method of multiple loading conditions based on local models.Firstly,it classified the operation condition using the K-means clustering algorithm.Then it established the local quality prediction models of multiple loading conditions using support vector machine.Finally,it got the local model weights using the improved particle swarm optimization,to make minimum error between the model output with the expected output,so that obtained the global model to realize the production process quality prediction.This method solved complexity and uncertainty problem of complex production process and had great global adaptive.It used a case study of the TE process to verify the model,implemented the modeling of complex production processes and advanced prediction.The result shows that the method is feasible and efficient.