【作者】 郑文建；

【导师】 杨煜普；

【作者基本信息】 上海交通大学 ， 控制工程， 2012， 硕士

【摘要】 模糊辨识就是采用模糊集合理论，根据系统的输入输出值来辨识系统的模糊模型。目前，它被广泛地应用于非线性系统的辨识中。现有辨识算法仍存在一些难题，例如避免“维数灾难”和提高模型泛化能力的问题。模糊模型辨识主要分为结构辨识和参数辨识两个部分，其中最重要的便是结构辨识，目前尚未形成对结构辨识完善的理论。而且，目前的一些模糊模型辨识方法很难应用到实际生产过程中，其中一个主要的原因就是传统辨识方法存在计算复杂度高与庞大的规则库的问题。因此，本文研究的主要出发点就是如何设计简单有效的辨识算法，以及减少其计算复杂度，使其适用于工业生产过程中。本文研究了一种基于最小二乘支持向量机（LSSVM）的T-S模糊建模新方法，提出了一种基于LSSVM的模糊内模控制策略，然后将其应用到先进铸造设备的定量浇铸控制。主要做了以下工作：（1）针对标准支持向量机模糊建模方法的计算复杂度高问题，引入LSSVM算法的等式约束，明显的提高了建模效率。（2）通过LSSVM算法对模糊模型进行结构划分，实现模糊模型的结构辨识；在不改变训练参数的情况下，通过剪枝算法得到具有稀疏性的支持向量，依据支持向量的个数来划分模糊空间，从而使得模型结构简单，便于应用推广。（3）将LSSVM模糊模型引入内模控制中，将其作为系统的内部模型，并且根据该模型设计了逆模型控制器。（4）设计了整个定量浇铸控制系统。主要分析了定量浇铸加压控制系统的控制特点，建立了仿真模型。仿真结果表明，在定量浇铸的液面加压系统中，基于LSSVM的模糊模型的内模控制方法在控制精度上和抗干扰能力方面都具有一定的优越性。更多还原

【Abstract】 Fuzzy Identification is to identify the fuzzy model of the system from the measured values of system inputs and outputs, using fuzzy set theory, it has a very wide range of applications in the field of nonlinear system identification. However, the existing identification algorithm is still facing the problems of how to avoid the curse of dimensionality and improve the model generalization capability. Fuzzy Model Identification includes structure identification and parameter identification. One of the most important is the structure identification, and it has not formed a theory on the structure identification. Moreover, there are still many difficulties in the application of fuzzy model identification in actual industrial processes. One of reasons is the huge rule base generated by the traditional identification methods, and consumption of identification calculation. Therefore, the main purpose of this paper is how to design a simple and effective identification algorithm, how to reduce the identification algorithm computing complexity.In this paper, we do some research on TS fuzzy modeling based on least squares support vector machine (LSSVM), and proposed a method of LSSVM-based fuzzy model control. Then we applied it in advanced casting equipment constant casting control system and mainly done the following work:1) For the high computational complexity problem in fuzzy modeling method based on standard support vector machine, the introduction of the LSSVM with equation constraints, significantly improved the efficiency of modeling.2) The structure of fuzzy model is identified using the LSSVM algorithm; the sparse support vectors are gotten through the pruning algorithm without changing the training parameters; the Fuzzy space is divided based on the number of support vectors. It simplifies the structure of the fuzzy model, and promotes its application.3) The LSSVM fuzzy model is introduced into the internal model control, and used as the system’s internal model. The inverse model controller is designed according to this modeling method.4) A constant casting control system is designed. A simulation model is built after the analysis of features of the constant casting pressure control system. The simulation results show that quantitative casting surface pressure systems, internal model control method based on of LSSVM fuzzy model has certain advantages in the control precision and anti-interference capability.更多还原