本科毕业设计--PLS在TE过程中的故障诊断中的应用研究.docVIP

摘要 随着现代流程工业不断地向大型化、集成化方向发展，工业过程的复杂性不断提高，工程变量也越来越多，一旦发生故障，将会造成人员和财产的巨大损失，因此过程的故障检测和诊断已经成为控制领域研究的重点之一。及时准确地检测和诊断出过程的故障，不仅可以减少事故、增加过程运行的安全性，而且可以降低生产管理成本，提高产品的质量。基于数据驱动的过程架空方法由于依赖于易得的过程数据，不依赖于精确的数学模型，正受到越来越多的关注。 本文以部分最小二乘法（Partial Least Squares, PLS）为基础，针对TE过程中的21个故障数据中的几个作为重点研究对象，进行故障建模。建模以后再通过交叉验证的方法诊断出主元的个数，用最小二乘法来判断故障诊断限，并求出回归矩阵B。从而使预测矩阵和被预测矩阵之间建立起关系，通过已经建立的模型和求出的回归矩阵对以知的故障进行检测，判断是否达到系统所能够允许的数值。 两个主程序都是用MATLAB编写的，一个是交叉验证法程序，一个是部分最小二乘法验证程序。交叉验证程序实现了主元个数的选取，最小二乘法验证程序能够证明PLS在TE过程故障诊断中对数据进行检测，诊断率能够达到标准。 关键词：部分最小二乘法，故障诊断，TE过程，交叉验证法。 Abstract Because of the facts that the complexity in the industrial process improves continuingly and the engineering changes become more and more as the modern flow industry progresses in the direction of greatness and integrity, and that man power and assets will suffer from great losses if there are malfunctions, therefore the testing and diagnosing of the malfunctions in the process have already become be of importance in the field of controlling study. Correctly testing and diagnosing the malfunctions in time can, not only reduce the accidents and improve the safety in the running of the process, but also reduce the price of production and management, as well as improve the quality of products. Stemming from the process figures that are easy to get and not relying on the accurate mathematic models, the non-process methods based on figure controlling are getting more and more attention. This paper, based on the Partial Least Squares, deals with the malfunction model establishment for the 21mulfunction figures in the TE process. After the model being established, the number of the main element is diagnosed by the method of reciprocal proving and the limit of malfunction diagnosis is discriminated by PLS, then the round matrix B is got. In that way, the predicting matrix and the predicted matrix are related in some degree. The model that has been established and the round matrix thus got could test if the malfunctions are in the limit of th