论文部分内容阅读
对具有不等长时段的多时段批次过程进行监测是十分重要而且具有一定难度的.时段在批次间的错位现象导致时间方向的不同过程特性混合在一起,这给时段分析以及在线应用带来了一系列的问题.为了解决不等长所带来的问题,本文提出一种基于不等长时段有序识别及建模的故障检测方法.该方法的主要贡献包括以下方面:1)该方法通过步进地衡量过程的变量相关性对模型精度以及监测性能的影响,自动有序地识别出每个不等长时段;2)在每个时段内,通过对不规则的过程数据进行整合建立了时段模型以捕捉不规则的时段特性;3)本文提供了一种简单而有效的在线判断新样本隶属时段和监测其运行状态的方法.最后,本文通过一个实例-具有不等长批次长度的注塑过程阐述了本方法的有效性.
Monitoring multi-session batches with unequal length of time is of great importance and with some degree of difficulty.Disorders between batches lead to the mixing of different process characteristics in time direction, Come to a series of problems.In order to solve the problems caused by unequal length, this paper presents a fault detection method based on unequal length of time ordered recognition and modeling.The main contributions of this method include the following aspects: 1) The method can automatically and orderly identify each unequal length of time period by measuring the influence of process variables on the precision and monitoring performance step by step. 2) Through the integration of irregular process data in each time period, A period model is built to capture the characteristics of irregular time periods.3) This paper provides a simple and effective method to determine the new sample membership period and monitor its running status.Finally, the paper through an example - Length of the injection process illustrates the effectiveness of the method.