本文针对一类含有不确定性的非线性系统，对于执行器故障情形，提出了一种新的鲁棒故障诊断方法。首先设计了自适应观测器结构，并利用最小二乘法给出了故障估计递推算法。对系统中的不确定性，算法中采用域值处理技术以实现鲁棒故障估计。在此基础上，本文分析了该方法的鲁棒性，可检测性和稳定性。最后，给出了仿真实例，结果证明了该方法的有效性。

很多学者对故障诊断和容错控制问题给予很大关注。在很多应用上，特别是对于安全系统，故障诊断固然重要，最快地调节故障系统也是很重要的。例如当今的高性能飞行器即使发生故障仍需保持基本的运行状态。对于非线性系统提出一种故障调节控制器的设计方法，通过修正控制律补偿故障所带来的影响。故障发生后使用的神经网络用于逼近故障函数并提供故障的修正行为，即主动容错。故障调节后闭环系统是稳定的。仿真算例证明了此方法的有效性。

针对一类含模型不确定性的非线性系统，提出了执行器故障检测与诊断的在线估计器设计方法。系统只有输入输出可检测，执行器故障是关于输入和状态的函数。非线性在线估计器用来监视系统是否发生故障，并且估计故障的大小和特征。文中给出了故障诊断结构与算法，并分析了鲁棒性，灵敏度和稳定性。最后，仿真结果验证了该方法的正确性。

针对PCA（Principal Components Analysis）技术中，由于重叠信息会严重影响主成分的正确提取这一问题，提出了一种改进的数据降维处理方法。首先，利用标准化变量间的相关系数大小找到重叠信息。然后，将重叠信息进行加权综合。最后，利用改进的数据降维处理方法以及分布式网络技术，建立了基于PCA-DRBFN（Principal Components Analysis-Distributed Radial Basis Function Network）的软测量模型，并将其应用到某钢厂的精苯精馏过程，对精苯干点进行估计。通过仿真证明，所建立的模型具有较好的泛化效果。

This paper is concerned with the modeling and control of processes with input dynamic nonlinearity. Rather than modeling the overall process with a nonlinear model, it is proposed to represent the process by a composite model of a linear model (LM) and a feedforward neural network (FNN). The LM is to capture the dominant linear dynamics, while the FNN is to approximate the remaining nonlinear dynamics. The controller, in correspondence, consists of two sub-controllers: a linear predictive controller (LPC) designed based on the LM, and an iterative inversion controller (IIC) designed based on the FNN. These two sub-controllers work together in a cascade fashion that the LPC computes the desired reference input to the IIC via a analytic predictive control algorithm and the IIC then determines the process manipulated variable. Since the neural network is used to model the nonlinear dynamics only, not the overall process, a relatively small size network is required, thus reducing computational requirement. The combination of linear and nonlinear controls results in a simple and effective controller for a class of nonlinear processes, as illustrated by the simulations in this paper.

Sliding mode control (SMC) has excellent robustness to model uncertainties and disturbances. This would make SMC an ideal scheme for process control applications where model uncertainties and disturbances are common. The existing SMC, however, has the major drawback of control chattering, i.e., the controller output is a discontinuous high frequency switching signal. This makes SMC not suitable for most chemical processes where the manipulated variables are continuous and where high frequency changes are not permitted. To eliminate chattering, a new PID based sliding mode control (PIDSMC) suitable for chemical process is proposed here. The proposed control system consists of three components: a compensation of process nonlinearity, a linear feedback of state tracking errors, and a PID control of sliding surface function. The chattering is eliminated via the replacement of the discontinuous switching in the SMC by a continuous input determined by a PID scheme. An adaptive strategy is proposed to tune the PID parameters on-line to control the process states onto a sliding surface that characterizes the closed-loop performance. The proposed algorithm has been shown to be effective in controlling an inverted pendulum system and a typical pH neutralization process.

Many processes in the chemical industry have modest nonlinearities, i.e. linear dynamics play a dominant role in governing the process output behaviour in the operating range of interest, but the linearization errors may be significant. For this type of processes, linear-based control may yield a poor performance, while nonlinear-based control results in computation complexity. We propose to model this type of process with a composite model consisting of a linear model (LM) and a multilayered feedforward neural network (MFNN). The LM is used to capture the linear dynamics, while the MFNN is employed to predict the LM's residual errors, i.e., the process nonlinearities. Effective off-line and on-line algorithms are proposed for the identification of the composite model. With this model structure, it is shown that a simple analytical predictive control law can be formulated to control a nonlinear process. Simulation examples are also given to illustrate the effectiveness of the model identification and the proposed predictive control.

Identification and control of a nonlinear process in the presence of unmeasured load disturbances is important, because most chemical processes are perturbed by load disturbances that are often not measured. In this paper, the absorption principle is first extended for the development of an effective identification strategy for feedforward neural network representation of the process input-output relation in the presence of unmeasured load disturbance. This developed model can give an accurate output prediction, irrespective of the load disturbances, as long as they can be reasonably approximated by piecewise polynomials. Secondly, a predictive control scheme is developed based on a genetic algorithm optimization, using the above-identified model, for the nonlinear process under the influence of unmeasured loads. Finally, simulations are given to illustrate the effectiveness of the proposed identification and control scheme.

A simple control law analogous to the linear generalized minimum variance (GMV) control is presented for general unknown nonlinear dynamic processes. With this control law, the iterative search of the control input, which is often encountered in nonlinear control, can be eliminated, resulting in an efficient computation for real-time implementation. The implementation of this control law requires two key quantities to be calculated: the input-output sensitivity function and the quasi-one-step-ahead predictive output. The selection of a diagonal recurrent neural network (DRNN) as the process identifier allows a direct estimation of these two quantities, resulting in the proposed control law to be implemented in a straightforward manner. Both simulation and experiment are given to emonstrate the effectiveness of the proposed control algorithm.

High product quality and consistent operation safety are important to industry processes, particularly to those with a large number of correlated process variables. Principal component analysis (PCA) has been widely used in multivariate process monitoring for its ability in reducing the process dimension. PCA and other statistical techniques, however, have difficulties in differentiating the faults with similar time-domain process characteristics. A wavelet-based time-frequency approach is developed in this paper to improve PCA-based methods by extending the time-domain process features into time-frequency information. Subsequently, a similarity measure is presented to compare process features for on-line process monitoring and fault diagnosis. xperimental results show that the proposed multivariate time-frequency process feature is effective in both fault detection and diagnosis, illustrating the potentials for real world applications.