论文部分内容阅读
为及时跟踪过程的突变特性,提出一种改进的JITL-PLS软仪表建模方法。该方法与传统的JITL-PLS方法不同,对参与建模的每个样本基于相似度分配不同的权值,以区别不同样本对模型的贡献,此外,为提高计算效率,采用PLS核算法而不是NIPALS算法建立局部模型。对脱丁烷塔的软仪表建模研究表明,提出的方法的预报精度较传统JITL-PLS方法有明显提高。
In order to track the mutation characteristics of the process in time, an improved JITL-PLS soft instrument modeling method is proposed. This method is different from the traditional JITL-PLS method. Each sample involved in the modeling is assigned a different weight based on similarity to distinguish the contribution of different samples to the model. In addition, in order to improve the computational efficiency, PLS is used instead of NIPALS algorithm to establish a local model. The modeling of the soft instrument of the debutanizer tower shows that the proposed method has a better prediction accuracy than the traditional JITL-PLS method.