Nonlinear internal model control strategy for neural network models

A nonlinear internal model control (NIMC) strategy based on neural network models is proposed for SISO processes. The neural network model is identified from input—output data using a three-layer feedforward network trained with a conjugate gradient algorithm. The NIMC controller consists of a model inverse controller and a robustness filter with a single tuning parameter. The proposed strategy includes time delay compensation in the form of a Smith predictor and ensures offset-free performance. Extensions for measured disturbances are also presented. The NIMC approach is currently restricted to processes with stable inverses. Two alternative implementations of the control law are discussed and simulations results for a continuous stirred tank reactor and pH neutralization process are presented. The results for these two highly-nonlinear processes demonstrate the ability of the new strategy to outperform conventional PID control.     

SIMULATION AND EXPERIMENTAL IMPLEMENTATION OF A NEURAL-NETWORK-BASED INTERNAL-MODEL CONTROL STRATEGY ON A REACTOR SYSTEM

The use of inverse-model-based control strategy for nonlinear system has been increasing lately. However it is hampered by the difficulty in obtaining the inverse of nonlinear systems analytically. Since neural networks has the ability to model such inverses, it has become a viable alternative. Although many simulations using neural network inverse models For controls have been reported recently, no actual experimental application has been reported on a reactor system. In this paper we describe a novel experimental application of a neural network inverse-model based control method on a partially simulated pilot plant reactor, exhibiting steady state parametric sensitivity and designed to test the use of such nonlinear algorithms. The implementation involved the control of the reactor temperature under set point changes, disturbance rejection and set point regulation with plant/model mismatches. Simulation tests on the model of the system were also carried out to enable better design of the neural network models and to highlight the differences between simulation and actual online results. The online implementation results obtained were sufficient to demonstrate the capability of applying these neural-network-based control methods in real systems.     

Nonlinear adaptive switching control for a class of non-affine nonlinear systems

An improved nonlinear adaptive switching control method is presented to relax the assumption on the higher order nonlinear terms of a class of discrete-time non-affine nonlinear systems. The proposed control strategy is composed of a linear adaptive controller, a neural network (NN) based nonlinear adaptive controller and a switching mechanism. An incremental model is derived to represent the considered system and an improved robust adaptive law is chosen to update the parameters of the linear adaptive controller. A new performance criterion of the switching mechanism is designed to select the proper controller. Using this control scheme, all the signals in the system are proved to be bounded. Numerical examples verify the effectiveness of the proposed algorithm.     

基于神经网络的pH中和过程非线性预测控制

针对pH中和过程这一化工过程系统中的典型非线性对象特点，应用神经网络建模思想和模型预测控制方法，并结合Hammerstein模型特点，研究pH中和过程非线性系统的两种新型模型预测控制手段，分别建立基于神经网络的非线性预测控制系统整体求解策略和基于Hammerstein模型的两步法预测控制策略，并用MATLAB对其进行仿真。控制仿真结果表明，建立的神经网络预测控制策略和非线性Hammerstein模型预测控制均优于传统PID控制方法，具有良好的设定值跟踪效果和抗干扰控制响应，说明这两种控制策略是非线性过程的有效控制方法。     

Batch-to-batch optimal control of a batch polymerisation process based on stacked neural network models

A neural network based batch-to-batch optimal control strategy is proposed in this paper. In order to overcome the difficulty in developing mechanistic models for batch processes, stacked neural network models are developed from process operational data. Stacked neural networks have enhanced model generalisation capability and can also provide model prediction confidence bounds. However, the optimal control policy calculated based on a neural network model may not be optimal when applied to the true process due to model plant mismatches and the presence of unknown disturbances. Due to the repetitive nature of batch processes, it is possible to improve the operation of the next batch using the information of the current and previous batch runs. A batch-to-batch optimal control strategy based on the linearisation of stacked neural network model is proposed in this paper. Applications to a simulated batch polymerisation reactor demonstrate that the proposed method can improve process performance from batch to batch in the presence of model plant mismatches and unknown disturbances.     

基于Hammerstein-Wiener模型的连续搅拌反应釜神经网络预测控制

针对化工过程中广泛使用的连续搅拌反应釜(CSTR),提出一种基于神经网络的模型预测控制策略,采用分段最小二乘支持向量机辨识Hammerstein-Wiener模型系数的方法,在此基础上建立线性自回归模式〖DK〗(ARX)结构和高斯径向基神经网络串联的非线性预测控制器。利用BP神经网络训练预测控制输入序列和拟牛顿算法求解非线性预测控制律,从而实现一种基于支持向量机Hammerstein-Wiener辨识模型的非线性神经网络预测控制算法。对CSTR的仿真结果表明,该方法能够更有效地跟踪控制反应物浓度。     

Additive scale control optimization and operation modes

Along the Gulf Coast there are three Saudi main desalination sites. These are, South to North, Al-Khobar, Al-Jubail and Al-Khafji. In these plants some 300 MIGD of distillate could be produced when all 58 MSF evaporators are run at a low top brine temperature (TBT) of 90°C. An additional 50 MIGD could be produced when the 52 evaporators designed as well, for high TBT of 113°C or more are run near this temperature. All these 58 evaporators are designed with additive scale control. Therefore additive treatment and cost are vital to East Coast Saudi Arabian plants.This paper will review original design operating parameters and performance. It will also report on some revisions reviewed or developed and executed at these plants. The changes were developed and adopted in an attempt to optimize production cost particularly scale control additive dose levels at different top temperatures. These plants were designed and undergone performance and trial runs at additive dose levels of 12 to 3 ppm. These were then reduced after the warranty periods to extremely low values of 2.2 to 0.8 ppm at different top temperatures.     

Machine-learning-based construction of barrier functions and models for safe model predictive control

In this paper, we propose a control Lyapunov-barrier function-based model predictive control method utilizing a feed-forward neural network specified control barrier function (CBF) and a recurrent neural network (RNN) predictive model to stabilize nonlinear processes with input constraints, and to guarantee that safety requirements are met for all times. The nonlinear system is first modeled using RNN techniques, and a CBF is characterized by constructing a feed-forward neural network (FNN) model with unique structures and properties. The FNN model for the CBF is trained based on data samples collected from safe and unsafe operating regions, and the resulting FNN model is verified to demonstrate that the safety properties of the CBF are satisfied. Given sufficiently small bounded modeling errors for both the FNN and the RNN models, the proposed control system is able to guarantee closed-loop stability while preventing the closed-loop states from entering unsafe regions in state-space under sample-and-hold control action implementation. We provide the theoretical analysis for bounded unsafe sets in state-space, and demonstrate the effectiveness of the proposed control strategy using a nonlinear chemical process example with a bounded unsafe region.     

Dynamic On‐Line Reoptimization Control of a Batch MMA Polymerization Reactor Using Hybrid Neural Network Models

A hybrid neural network model based on‐line reoptimization control strategy is developed for a batch polymerization reactor. To address the difficulties in batch polymerization reactor modeling, the hybrid neural network model contains a simplified mechanistic model covering material balance assuming perfect temperature control, and recurrent neural networks modeling the residuals of the simplified mechanistic model due to imperfect temperature control. This hybrid neural network model is used to calculate the optimal control policy. A difficulty in the optimal control of batch polymerization reactors is that the optimization effort can be seriously hampered by unknown disturbances such as reactive impurities and reactor fouling. With the presence of an unknown amount of reactive impurities, the off‐line calculated optimal control profile will be no longer optimal. To address this issue, a strategy combining on‐line reactive impurity estimation and on‐line reoptimization is proposed in this paper. The amount of reactive impurities is estimated on‐line during the early stage of a batch by using a neural network based inverse model. Based on the estimated amount of reactive impurities, on‐line reoptimization is then applied to calculate the optimal reactor temperature profile for the remaining time period of the batch reactor operation. This approach is illustrated on the optimization control of a simulated batch methyl methacrylate polymerization process.     

Dynamic recurrent radial basis function network model predictive control of unstable nonlinear processes

A multistep model predictive control (MPC) strategy based on dynamically recurrent radial basis function networks (RBFNs) is proposed for single-input single-output (SISO) control of uncertain nonlinear processes. The control system consists of two automatically configured RBFNs, a trained network representing the plant model and a network with on-line learning to function as controller. The automatic configuration and learning of the networks is carried out by using a hierarchically self-organizing learning algorithm. This control strategy is structurally simple and computationally efficient since a single output node of each RBFN is configured to provide multistep predictions for plant output and controller. The performance of the proposed RBFNMPC strategy is evaluated by applying to two unstable nonlinear chemical processes, a chemical reactor and a biochemical reactor, and also a stable polymerization reactor. Further, the results of the RBFNMPC is compared with similar RBFN model based control strategies and also with well tuned PID/PI controller. The results show the better performance of the proposed RBFNMPC for the control of open-loop unstable nonlinear processes that exhibit multiple steady-state behavior.     

利用输出反馈方法实现聚合物分子量分布的控制

利用正交多项式组合神经网络对聚合反应分子量分布(MWD)进行建模,MWD被解构为若干矩值组成的矩向量函数,由此二元MWD控制问题可被转换为一元分布矩控制问题。以矩向量为直接被控对象,通过对矩的控制实现MWD的跟踪。为便于求解这类非仿射非线性多变量系统的控制策略,提出了改进型非线性自回归正交多项式网络结构,建立模型输出与U(k)之间的线性映射关系;针对高维被控矩向量,证明了矩向量中独立变量与分布函数参变量间的数量对等关系,给出了矩向量降维准则,将高维输出控制转化为低维问题。基于改进的神经网络模型,利用输出反馈方法对MWD矩向量进行控制,实现了对MWD的形状跟踪,仿真实验证明了方法的有效性。所提出的方法为非线性多变量系统的建模控制问题提供了新思路。     

模糊非线性内模控制算法及其在pH值控制中的应用

pH值控制过程具有较强的非线性,历来是过程控制研究的一大热点,本文针对pH值控制系统提出了一种基于模糊推理网的非线性内模控制算法（FNIMC）。模糊推理网用于辨识对象的模糊模型;FNIMC由一个逆模控制器和具有一个可调参数的鲁棒滤波器组成。仿真结果表明该算法优于非线性PID调节器,且计算效率高。     

Optimal control of batch processes using particle swam optimisation with stacked neural network models

An optimal control strategy for batch processes using particle swam optimisation (PSO) and stacked neural networks is presented in this paper. Stacked neural network models are developed form historical process operation data. Stacked neural networks are used to improve model generalisation capability, as well as provide model prediction confidence bounds. In order to improve the reliability of the calculated optimal control policy, an additional term is introduced in the optimisation objective function to penalize wide model prediction confidence bounds. The optimisation problem is solved using PSO, which can cope with multiple local minima and could generally find the global minimum. Application to a simulated fed-batch process demonstrates that the proposed technique is very effective.     

Control of nonisothermal CSTR with time varying parameters via dynamic neural network control (DNNC)

Dynamic neural network control (DNNC) is a model predictive control strategy potentially applicable to nonlinear systems. It uses a neural network to model the process and its mathematical inverse to control the process. The advantages of single hidden layer DNNC are threefold: First, the neural network structure is very simple, having limited nodes in the hidden layer and output layer for the SISO case. Second, DNNC offers potential for better initialization of weights along with fewer weights and bias terms. Third, the controller design and implementation are easier than control strategies such as conventional and hybrid neural networks without loss in performance. The objective of this paper is to present the basic concept of single hidden layer DNNC and illustrate its potential. In addition, this paper provides a detailed case study in which DNNC is applied to the nonisothermal CSTR with time varying parameters including activation energy (i.e., deactivation of catalyst) and heat transfer coefficient (i.e., fouling). DNNC is compared with PID control. Although it is clear that DNNC will perform better than PID, it is useful to compare PID with DNNC to illustrate the extreme range of the nonlinearity of the process. This paper represents a preliminary effort to design a simplified neural network-based control approach for a class of nonlinear processes. Therefore, additional work is required for investigation of the effectiveness of this approach for other chemical processes such as batch reactors. The results show excellent DNNC performance in the region where conventional PID control fails.     

Application of neural networks to meltblown process control

Process modeling is essential for the control of optimization and an on-line prediction is very useful for process monitoring and quality control. Up to now, no satisfactory methods have been found to model an industrial meltblown process since it is of highly dimensional and nonlinear complexity. In this article, back-propagation neural networks (BPNNs) were investigated for modeling the meltblown process and on-line predicting the product specifications such as fiber diameter and web thickness. The feasibility of this application was successfully demonstrated by agreement of the prediction results from the BPNN to the actual measurements of a practical case. The network inputs included extruder temperature, die temperature, melt flow rate, air temperature at die, air pressure at die, and die-to-collector distance (DCD). The output of the fiber diameter was obtained by neural computing. The network training was based on 160 sets of the training samples and the trained network was tested with 70 sets of test samples which were different from the training data. This research is preliminary and of industrial significance and especially valuable for the optimal control of advanced meltblown processes. © 1996 John Wiley & Sons, Inc.     

基于复合神经网络的离散非线性系统建模与控制

An adaptive inverse controller for nonliear discrete-time system is proposed in this paper. A compound neural network is constructed to identify the nonlinear system, which includes a linear part to approximate the nonlinear system and a recurrent neural network to minimize the difference between the linear model and the real nonlinear system. Because the current control input is not included in the input vector of recurrent neural network (RNN), the inverse control law can be calculated directly. This scheme can be used in real-time nonlinear single-input single-output (SISO) and multi-input multi-output (MIMO) system control with less computation work. Simulation studies have shown that this scheme is simple and affects good control accuracy and robustness.     

基于结构逼近式神经网络的间歇反应器优化控制

利用结构逼近式混合神经网络（SAHNN）建立了一类典型放热液相二级平行间歇反应的数学模型。基于主产物浓度和反应温度的递归神经网络（RNN）模型,使用混合PSO-SQP算法求解该间歇反应主产物产率最大化问题,进而得到反应温度优化曲线。鉴于反应温度实时可测,提出扩展的EISE指标,该指标把实时计算的模型误差引入控制策略,为基于模型的控制增加了反馈通道,增强了控制方法的鲁棒性和抗干扰性能。利用 原理对所提出的一步超前预测控制做了稳定性分析,证明了算法的正确性。研究的结果充分证明了基于SAHNN混合神经网络模型的优化控制策略的有效性。     

Neural network-based predictive control for multivariable processes

In this paper, the systematic derivations of setting up a nonlinear model predictive control based on the neural network are presented. This extends our previous work (Chen, 1998) into a multivariable system to explore the characteristics of the design. There are two stages for the development of nonlinear neural network predictive control: a neural network model and a control design. In the neural network model design, a parallel multiple-input, single-output neural network autoregressive with a model of exogenous inputs (NNARX) is proposed for multistep ahead predictions. In control design, the controller with extended control horizon is developed. The Levenberg-Marquardt algorithm is applied to training the NNARX model. Also, the sequential quadratic programming is used to search for the optimal manipulated inputs. The gradient of the objective function and constraints that require computation of Jacobian matrices are completely derived for optimization calculation. To demonstrate the control ability of MIMO cases, the proposed method is applied through two nonlinear simulation problems.     

Neural‐Networks‐Based Feedback Linearization versus Model Predictive Control of Continuous Alcoholic Fermentation Process

In this work advanced nonlinear neural networks based control system design algorithms are adopted to control a mechanistic model for an ethanol fermentation process. The process model equations for such systems are highly nonlinear. A neural network strategy has been implemented in this work for capturing the dynamics of the mechanistic model for the fermentation process. The neural network achieved has been validated against the mechanistic model. Two neural network based nonlinear control strategies have also been adopted using the model identified. The performance of the feedback linearization technique was compared to neural network model predictive control in terms of stability and set point tracking capabilities. Under servo conditions, the feedback linearization algorithm gave comparable tracking and stability. The feedback linearization controller achieved the control target faster than the model predictive one but with vigorous and sudden controller moves.