具有预测校正的固定床反应器状态估计

本文建立了基于后集中方法的固定床反应器的非线性状态估计算法，并分别用正交配置法集中偏微分方程、Gear法积分常微分方程、Hybrid法求解非线性代数方程，实现状态估计器的在线估计．针对工业实际过程中浓度测量滞后较大、采样时间间隔较长的特点，提出了浓度估计的预测校正，进一步提高了状态估计器对模型误差的克服能力．同时，在实验和仿真两方面考察了固定床反应器状态估计器的性能，无论是收敛性还是鲁棒性，其品质都令人满意．     

具有预测校正的固定床反应器状态估计

本文建立了基于后集中方法的固定床反应器的非线性状态估计算法，并分别用正交配置法集中偏微分方程、Gear法积分常微分方程、Hybrid法求解非线性代数方程，实现状态估计器的在线估计．针对工业实际过程中浓度测量滞后较大、采样时间间隔较长的特点，提出了浓度估计的预测校正，进一步提高了状态估计器对模型误差的克服能力．同时，在实验和仿真两方面考察了固定床反应器状态估计器的性能，无论是收敛性还是鲁棒性，其品质都令人满意．     

化学反应器的神经元自适应观测和非线性控制

提出了一种基于神经网络的自适应观测和非线性控制策略,证明了自适应观测器的收敛件和非线性控制系统的稳定性,将其用于连续搅拌釜式放热反应器的浓度控制。根据可在线测量的反应温度,在线估计不可在线测量的反应物浓度和辨识Arrhenius指前因子,并利用重构的状态信息设计出带约束的非线性控制策略。仿真结果表明,观测器/控制器的组合提供了满意的闭环特性,证实了本文方法的有效性。     

非线性分布参数系统状态估计的最佳测量位置

研究了分布参数系统状态估计中特有的最佳测量位置问题．建立了基于后集中方法的分布参数系统的非线性状态估计器,包括状态估计偏微分方程和微分灵敏度矩阵偏微分方程,并用适当的数值计算方法实现状态估计器的求解;以一个最小化的空间域上积分函数表达最佳测量位置的目标函数,并相应地用非线性约束优化方法求解系统具有一个或多个测量时的最佳测量位置．还以壁冷式单管固定床反应器为例,讨论了各种因素对最佳测量位置的影响及其灵敏度,并得出了一些有普遍意义的结论．     

基于工业色谱仪的精馏脱炔烃塔软测量建模

提出一种基于工业色谱仪的软测量建模方法,并针对碳五馏分分离过程中的精馏脱炔烃塔塔底成分估计问题,建立了合适的工业软测量模型。介绍了工业色谱仪在线质量检测原理和LM-BP神经网络模型的建立,并利用工业色谱仪在线检测的质量数据进行系统的在线和周期性模型更新,提高了软测量模型的在线估计精度。研究结果表明,基于工业色谱仪的LM-BP神经网络模型是一种有效的软测量建模方法。     

异构化机理软测量模型在工业装置中的在线应用

讨论芳烃异构化机理模型的性能及其在线应用。基于已开发的动力学模型,针对大量工业数据,进行离线模拟,结果发现模型的＂老化＂问题。为了兼顾模型的估计精度和计算效率,提出将软测量模型拆分成在线模拟计算模块和离线参数估计模块,实时更新模型参数。然后将模型应用于工业异构化装置,在线估计系统组分的浓度,结果表明,该软测量模型具有良好的性能。     

基于状态预测的塑料薄膜张力反馈控制器设计

塑料薄膜工业生产过程中,张力控制系统存在明显的非线性和滞后性,严重影响了塑料薄膜产品的质量。为了提高塑料薄膜张力控制系统的控制精度以及动态性能,基于神经网络对系统进行状态预测,并利用预测的系统状态设计了反馈控制器。通过神经网络在线辨识动态非线性模型,构建了神经网络动态辨识器;运用泰勒级数展开法计算预测未来时刻的神经网络权值,并建立状态预测器;根据预测状态设计了塑料薄膜张力系统的反馈控制器;通过对比仿真实验验证了所设计反馈控制器能够明显改善塑料薄膜张力控制系统的控制精度及控制性能。     

一种扰动自适应的鲁棒预测控制算法

针对实际生产中扰动的时变性,提出了一种扰动自适应的鲁棒预测控制（RAMPC）算法以提高扰动抑制性能。采用时间序列（ARMA）模型在线辨识系统的不可测扰动,通过基于多次迭代思想的递推辨识算法（multi-iteration pseudo-linear regression,MIPLR）来保证在线辨识的质量和收敛速度。考虑到数据与辨识模型的不确定性,改用min-max形式描述MPC算法的控制作用优化命题,并将在线辨识过程中的误差数据引入min-max命题,使在线辨识与控制作用鲁棒优化求解紧密结合起来,提高算法鲁棒性。进一步将此min-max问题转换为一个等效的非线性min问题,并采用多步线性化方法实现快速求解,解决了传统min-max方法在线计算负荷高的问题。仿真结果表明了该算法的有效性。     

基于深层神经网络的多输出自适应软测量建模

在污水处理运行过程中,多个重要的难测过程变量的存在,不仅妨碍了生产过程的监控,而且阻碍了过程控制策略的调整或优化。即使软测量模型得到合理的构建,在投入运行后仍然遭受性能的退化和同时带来的高昂的维护成本。此外,合适辅助变量的选取直接影响后续建模的效果。因此,文中提出了一种基于深层神经网络的多输出自适应软测量模型,用于污水处理过程中多个目标变量的同步在线预测。其中,深层神经网络基于一种栈式自编码而构建,在极端复杂场景下具有优异的在线预测性能;并在建模中引入时差建模和变量重要性投影（VIP）这两种算法,以应对性能退化问题和实现辅助变量的精选。最后,通过一个实际案例对所提出模型进行验证。结果表明,所提出的软测量模型不仅具有较好的多输出预测性能,且在单目标预测结果上也有不错的表现。     

基于深层神经网络的多输出自适应软测量建模

在污水处理运行过程中,多个重要的难测过程变量的存在,不仅妨碍了生产过程的监控,而且阻碍了过程控制策略的调整或优化。即使软测量模型得到合理的构建,在投入运行后仍然遭受性能的退化和同时带来的高昂的维护成本。此外,合适辅助变量的选取直接影响后续建模的效果。因此,文中提出了一种基于深层神经网络的多输出自适应软测量模型,用于污水处理过程中多个目标变量的同步在线预测。其中,深层神经网络基于一种栈式自编码而构建,在极端复杂场景下具有优异的在线预测性能;并在建模中引入时差建模和变量重要性投影(VIP)这两种算法,以应对性能退化问题和实现辅助变量的精选。最后,通过一个实际案例对所提出模型进行验证。结果表明,所提出的软测量模型不仅具有较好的多输出预测性能,且在单目标预测结果上也有不错的表现。     

On-line identification of interacting two-tank system

In order to demonstrate the effectiveness of the process identification algorithm, on-line parameter estimator is evaluated experimentally by using two-tank system with interaction. On-line parameter estimator used in this paper is based on a recursive parameter estimation algorithm. MIMO linear, bilinear and quadratic models based on ARMA model are used to identify two-tank system. A quadratic model for two-tank system with interaction is developed to confirm the propriety of MIMO quadratic model used in identification of two-tank system. The results of on-line identification experiments on the two-tank system show that the estimated parameters of each model converge and the output tracking errors are bounded by disturbance bound. But, the quadratic model showed the best convergence.     

Model identification of a twin screw extruder for thermoplastic vulcanizate (TPV) applications

Multi‐input multi‐output (MIMO) models of a twin‐screw co‐rotating extruder for thermoplastic vulcanizate (TPV) are developed using the process identification techniques. The process inputs are screw speed (SS) and barrel temperature (WT). The three outputs are motor load (ML), melt temperature (MT), and melt pressure (MP). Two appropriate rubbers for TPV applications with different physical and mechanical properties are used for the experimentation. The process model is obtained from the experimental input–output data using various identification techniques such as least squares and prediction error. Recursive online model identification is performed on the process to update the model parameters in real time. To perform the identification studies, the process data was transferred via OPC server from the local PLC (Programmable Logic Controller) to the Advanced Control and Identification toolbox in MATLAB software. The effect of rubber properties and two curative agents (Peroxide and Phenolic) in the TPV experiment are studied on the final identified models. This comprehensive model identification study provided sufficient accurate models for further model based process analysis and control for TPV applications. POLYM. ENG. SCI., 2010. © 2009 Society of Plastics Engineers     

基于快速滑窗QR分解的自适应子空间辨识

子空间辨识直接由输入输出数据辨识得到过程状态空间模型,在多变量系统的辨识中取得广泛应用。实现在线子空间辨识算法的关键在于快速、高效的QR分解及SVD分解更新算法。通过将Updating和Downdating操作有效结合,提出了一种快速的滑窗QR分解算法,减少了不必要的重复步骤,进一步提高了计算效率。复杂度分析结果表明,随数矩阵行数增加,快速滑窗QR分解算法比Updating、Downdating两步法可以减少8.3%的计算量。将快速滑窗QR分解算法用于PO-MOESP子空间辨识算法的自适应更新,并通过数值仿真实例验证了算法的有效性。     

Optimization-based predictive control of a simulated moving bed process using an identified model

For optimization-based dynamic control of simulated moving bed (SMB) process, a novel control strategy based on process identification, which is an extension of the earlier work (Song et al., 2006a. Identification and predictive control of a simulated moving bed process: purity control. Chemical Engineering Science 61, 1973-1986), is proposed. A linear output prediction model is obtained by the method of subspace identification and used for the dynamic control. The controller is designed for optimizing the production cost while maintaining the specified product purities. For all of these, the average purities over one switching period of the target components in extract and raffinate streams, the reciprocal productivity and the solvent consumption are selected as output variables, while the flow rates in 1, 2, 3 and 4 are chosen as the manipulated variables. The realization of this concept is discussed and assessed on a virtual eight column SMB unit for a system following a bi-Langmuir isotherm. The identified prediction model is proven to be in good agreement with the first principles model considered as the actual SMB process. For typical control objectives encountered in actual operation, i.e., disturbance rejection and set-point tracking, it is shown that the proposed controller exhibits excellent performance, hence it is an effective tool for optimization-based control of SMB process.     

缺失数据下基于EM算法的非线性过程建模

线性时变参数系统(LPV)将多阶段、非线性的过程建模转化为线性多模型的辨识问题,近年来得到了极大关注。考虑缺失数据下LPV系统的离线建模问题,首先引入一个二进制变量表征输出样本缺失状态,选取过程关键变量作为调度变量,确定主要工况点;然后围绕不同工况点建立局部子模型,将输出缺失部分和采样数据的模型归属当作隐藏变量,利用EM算法进行参数估计,再采用高斯权重函数融合各子模型。最后分别针对典型二阶过程和连续搅拌反应釜(CSTR),运用提出的多模型和算法进行仿真实验,表明有效性。     

On-line process identification and PID controller autotuning

We propose a new and simple on-line process identification method for the automatic tuning of the PID controller. It does not require a special type of test signal generators such as relay or P controller only if the signals are persistently exciting. That is, a user can choose arbitrary signal generators such as relay, a P controller, the controller itself, pulse signal and step signal generator because it needs only the measured process output and the controller output. It can incorporate nonlinearities due to actuator saturation or manual mode operation during identification work and shows a good robustness to measurement noises, nonlinearity of the process and disturbances. The proposed autotuner combined with the identification method and tuning rule using a model reduction shows good control properties compared with previous autotuning methods.     

Identification of uncertain MIMO Wiener and Hammerstein models

Several approaches can be found in the literature to perform the identification of block oriented models (BOMs). In this sense, an important improvement is to achieve robust identification to cope with the presence of uncertainty.In this work, two special and widely used BOMs are considered: Hammerstein and Wiener models. The models herein treated are assumed to be described by parametric representations. The approach introduced in this work for the identification of the multiple input-multiple output (MIMO) uncertain model is performed in a single step. The uncertainty is described as a set of parameters which is found through the solution of an optimization problem.A distillation column simulation model is presented to illustrate the robust identification approach. This process is an interesting benchmark due to its well-known nonlinear dynamics. Both Hammerstein and Wiener models are used to represent this plant in the presence of uncertainty. A comparative study between these models is established.     

Process identification for an SOPDT model using rectangular pulse input

A new method of process identification for a second-order-plus-dead-time model is proposed and tested with two example systems. In the activation of the example processes for the identification, a rectangular pulse input is applied to open loop systems. The model parameters are estimated by minimizing sum of modeling errors with the least squares method. The estimation performance is examined by comparing the output pulse responses from the example system and the estimated model. The performance comparison of the proposed method and two existing techniques indicates that satisfactory parameter estimation is available from the proposed procedure. In addition, the role of sampling time and the shape of input pulse is evaluated and it is found that the sampling time of less than 0.01 minute gives good estimation while the shape of input pulse does not affect the estimation performance. Finally, the robustness of the estimation in noisy process is proved from the investigation of the performance in the processes having various levels of noise.     

Improved Fourier transform for processes with initial cyclic‐steady‐state

A new process identification method is proposed to estimate the frequency responses of the process from the activated process input and output. It can extract many more frequency responses and guarantees better accuracy than the previous describing function analysis algorithm. In addition, the proposed method can be applied to the case that the initial part of the activated process data is periodic (cyclic‐steady‐state), which is not possible with any previous nonparametric identification methods using the modified Fourier transform or Fourier analysis. Furthermore, it can incorporate all the cases in which either the initial part is steady‐state and the final part is cyclic‐steady‐state or both the initial and final parts are steady‐state. © 2010 American Institute of Chemical Engineers AIChE J, 2010     

Improved system identification method for Hammerstein-Wiener processes

We propose a new system identification method for Hammerstein-Wiener processes, in which an input static nonlinear block, a linear dynamic block, and an output static nonlinear block are connected in a series. The proposed method can estimate the model parameters in a very simple way without solving the full-dimensional nonlinear optimization problem by activating the process with a specially designed test signal, composed of a relay feedback signal, a binary signal and a multi-step signal. The proposed method analytically identifies the output nonlinear static function and the input nonlinear static function from the relay signal and the multi-step signal, respectively. The linear dynamic subsystem is identified from the relay feedback signal and the binary signal with existing well-established linear system identification methods. We demonstrate with a simple example that the proposed method can be successfully applied to identify the Hammerstein-Wiener-type nonlinear process.