Theory and application of nonlinear system identification, central and its integration based on Wiener model

本文研究了一类维纳型非线性系统的辨识与控制问题。采用子空间法和模型参数优化相结合的方法，提出了一种新的基于Wiener模型的参数识别算法。辨识结果表明，维纳模型在捕捉非线性系统的动力学特性方面优于线性模型。在假设已识别的维纳模型是完全可控和可观察的前提下，设计并实现了基于ekf的维纳模型控制器。数值算例验证了所提辨识算法和控制设计的有效性。然后，我们将该算法应用于半导体制造中快速热过程的温度控制。根据RTP系统所要求的温度范围，设计了鉴定实验。然后根据输入-输出数据估计出维纳模型的参数，根据得到的维纳模型设计了…More控制器，以达到期望的温度响应。控制实验验证了我们的控制设计，结果表明，控制器能够在较宽的工作包络范围内控制晶圆上的动态响应和热均匀性。针对模型参数变化缓慢的情况，对控制系统进行了改进。尝试了一种基于滑模技术的鲁棒观测器。使用滑模观测器的另一个优点是避免了Riccati方程的在线计算，节省了控制系统的计算成本。故障检测算法的开发与系统辨识和控制设计同步进行。为了捕获模型参数方差，提出了一种基于维纳模型的在线识别算法。然后利用奇异值分解技术，得到模型内部特征的缓慢变化。建立了鲁棒控制系统的故障标准。少

This study is concerned with the problem of identification and control of a Wiener type nonlinear systems. A new parametric identification algorithm based on the Wiener model is derived by using both subspace method and model parameter optimization. The identification result shows that the Wiener model is superior to a linear model in capturing the dynamics of nonlinear system. An EKF-based controller for proposed Wiener model is designed and implemented subject to the assumption that an identified Wiener model is completely controllable and observable. The numerical example demonstrates the validity of both proposed identification algorithm and control design.Next, we applied the algorithms to the temperature control of rapid thermal process in semiconductor manufactory. The identification experiment was designed with respect to the desired temperature range in The RTP system. The parameter of Wiener, model was then estimated from input-output data Based on obtained wiener model, the … More controller was designed to achieve the desired temperature response. The control experiment was performed to verify our control design and the results shows that the con roller is capable of steering the dynamic response and thermal uniformity across a wafer over a wide operating envelope.We also improved the control system with respect to the case that the model parameter is varying slowly. A robust observer based on the sliding mode technique was attempted. The other advantage to use sliding mode observers is to avoid online calculation of Riccati equation, which- saves the calculation cost in the control system.The development of the algorithm for fault detection has been studied at the same pace with system identification and control design. An online identification algorithm based on the wiener model was developed for capturing the model parameter variance. Then by using singular value decomposition technique, we obtain the slow variation of model inner character. A standard for system fault has been established for robust control system. Less