基于双能γ射线的煤质灰分软测量技术研究

为提高煤质灰分测量精度,提出了基于双能γ射线的煤质灰分智能软测量方法,该方法以1377Cs和241Am作为中能和低能的γ射线源,并以探测器检测到的γ计数作为辅助变量,利用混沌算法优化的函数链神经网络实现灰分软测量辨识建模,最后对煤质灰分进行软测量预测和验证.研究结果表明:混沌算法优化的函数链神经网络预测方法的预测精度高,具有较强的泛化能力;基于混沌算法优化函数链神经网络的灰分智能软测量值与实测值的平均误差为0.7％,最大误差为0.9％,煤质灰分测量准确度高.     

神经网络逆软测量方法的拓展及在生物浸出过程中的应用

在前期工作中,提出了基于"内含传感器"的逆软测量方法,其中逆软仪表的构造仅仅是基于直接可测的状态来实现的。对该方法进行了拓展,首先将用于构造逆软仪表的直接可测量由直接可测的状态拓展为函数变量,然后对逆软仪表的建模算法进行了改进。这种拓展和改进不仅增加了逆软仪表构造成功的可能性,而且可以降低构造逆软仪表所需的直接可测量的导数的阶数,便于工程实现和应用。另外,采用神经网络来逼近理论上存在的逆软仪表,并得到了神经网络逆软仪表,从而解决了解析逆软仪表难以实现的工程应用瓶颈。最后将神经网络逆软仪表应用于生物浸出过程,实现了其不直接可测状态的在线软测量。仿真结果表明神经网络逆软仪表的软测量值与真实值非常接近,从而验证了该方法的有效性。     

基于PCA-BP神经网络的精馏塔产品组成软测量模型

依据工艺机理和操作经验，初选了醋酸精馏塔产品组成的神经网络预测模型的输入变量，运用主元分析方法对输入变量进行主元分解，降低输入变量维数且消除了输入变量之间的线性相关性，再通过基于LM优化算法的BP神经网络进行建模。仿真结果表明，该模型具有较快的训练速率和较高的预测精度，可以满足精馏过程对出口物料组成的在线软测量要求。     

基于改进核模糊聚类算法的软测量建模研究

针对发酵过程软测量建模采用单模型建模方法存在计算量大和精度较差的问题,提出一种基于改进核模糊聚类算法的多模型神经网络软测量建模方法.该方法首先使用主元分析方法对样本数据进行数据处理,所得主元变量作为模型的输入变量,然后使用基于粒子群优化算法的核模糊C均值聚类算法(PSKFCM)对数据集作聚类划分,最后针对每个聚类建立局部神经网络模型,多个局部神经网络模型估计结果的融合即为软测量模型的输出.将所提建模方法应用于红霉素发酵过程生物量浓度软测量建模,结果表明所建软测量模型具有较高的精度和良好的泛化能力.     

基于并行核径向基神经网络的软测量与控制模型

转炉提钒过程中存在大量多元非线性因素,难以从统计学和机理上建立各操作参数与生产目标的优化控制模型,为优化转炉的操作参数,建立了基于径向基神经网络的半钢钒含量软测量和控制模型.径向基神经网络常用于非线性回归预测和控制,但是高维的核函数矩阵运算需要花费巨大计算资源.为了缩短计算时间,本文设计了并行算法用于计算径向基网络核函数矩阵,并将它用于转炉提钒软测量和控制模型,在以MPI构建的工作站机群上执行该算法,利用实际数据验证了该算法的加速性和准确性.     

龙门式多轴自动锁螺丝设备横梁结构优化

为实现龙门式多轴自动锁螺丝设备横梁的轻量化,结合BP神经网络与粒子群优化算法对其横梁进行结构优化。以横梁质量为目标函数,数个关键尺寸为设计变量,变形量及固有频率为约束条件建立数学模型。利用BP神经网络拟合设计变量与约束变量的映射关系,结合已建立的神经网络模型,应用基于Deb可行性规则改进的粒子群算法,在满足约束要求的条件下,寻求各关键尺寸的最优值。优化结果表明,优化后的横梁质量减少29.61%,实现横梁轻量化。     

基于IQPSO优化ELM的熟料质量指标软测量研究

水泥熟料游离氧化钙(f Ca O)含量是水泥生产过程的重要质量指标。针对难以建立其精确的数学模型和难以实时在线测量的问题,首先采用序列二次规划方法增强量子粒子群算法的局部搜索能力,提出了一种局部区域可调的改进量子粒子群优化(IQPSO)算法,并采用提出的IQPSO算法优化超限学习机(ELM)的输入层权值和隐层阈值参数,在优化过程中同时兼顾均方根误差和隐层输出矩阵条件数最小的原则,建立了基于IQPSO优化ELM的水泥熟料f Ca O软测量模型,仿真验证结果表明,IQPSO算法具有较高的搜索精度以及较快的收敛速度,建立的软测量模型精度高、泛化能力强。最后基于该模型,通过软件编程的方法给出了水泥熟料质量指标软测量仪表,实现了f Ca O含量的在线软测量。     

一种新型混沌PSO算法在板形智能控制中的应用简

提出一种新型混沌PSO算法优化RBF神经网络并对板形进行识别,使用神经网络预测和效应矩阵控制器对板形进行预测控制.仿真过程表明,新型混沌PSO算法对优化神经网络的结构和参数都有明显的效果,使用板形识别模型和带预测过程的效应函数可以有效控制板形系统.     

提高动态流量软测量实时性的RBF中心优化算法

针对液压伺服系统动态流量软测量模型中神经网络训练精度和训练速度难以同时提升的问题,引入减聚类(SCM)算法将原训练样本集映射成初始径向基函数(RBF)中心集,并确定基函数宽度;利用敏感性分析算法(SenV)对基函数的中心进行优化,从而减少神经网络隐层节点数目;在根本上为同时提升神经网络训练精度和训练速度提供保障.实验表明,神经网络的隐层节点数可降低至少30%.     

正则化学习算法在软测量建模中的应用

软测量技术的核心是建立软测量模型。基于过程可测信息集建立软测量模型即逼近建模过程是不适定的。以径向基函数神经网络作为软测量模型,在软测量建模中引入正则化学习算法。以广义交叉验证作为正则化参数估计方法,讨论了径向基函数神经网络软测量逼近建模的全局与局部正则化学习算法,给出的实例说明了其有效性。     

化工精馏塔的PLC温度控制系统设计

针对精馏塔温度具有大滞后、大惯性时间常数,难以准确控制的特点,为某化工厂设计了基于PLC的温度控制系统.在该控制系统中采用了串级控制,主控制器中采用积分分离PID算法,改善了精馏过程的动态特性,对负荷变化的适应性也较强,在实践中取得了稳定的控制效果.     

SOFT SENSING MODEL BASED ON SUPPORT VECTOR MACHINE AND ITS APPLICATION

Soft sensor is widely used in industrial process control. It plays an important role to improve the quality of product and assure safety in production. The core of soft sensor is to construct soft sensing model. A new soft sensing modeling method based on support vector machine (SVM) is proposed. SVM is a new machine learning method based on statistical learning theory and is powerful for the problem characterized by small sample, nonlinearity, high dimension and local minima. The proposed methods are applied to the estimation of frozen point of light diesel oil in distillation column. The estimated outputs of soft sensing model based on SVM match the real values of frozen point and follow varying trend of frozen point very well. Experiment results show that SVM provides a new effective method for soft sensing modeling and has promising application in industrial process applications.     

精馏塔温度模糊解耦控制系统的研究

精馏过程属于多变量、时变、强耦合和具有分布参数的非线性过程,为了实现精馏塔塔顶和塔底的产品均需达到一定纯度的要求,在分析精馏塔工作特性以及塔顶和塔底的温度耦合特征的基础上,提出了一种塔顶和塔底温度模糊解耦控制方案。通过对塔顶和塔底温度的PID控制、解耦控制、模糊解耦控制,进行了设定值扰动测试、过程扰动测试的仿真。结果表明,模糊解耦优于PID控制和解耦控制,验证了模糊解耦控制的可行性。     

Neuro-estimator based GMC control of a batch reactive distillation

In this paper, an artificial neural network (ANN)-based nonlinear control algorithm is proposed for a simulated batch reactive distillation (RD) column. In the homogeneously catalyzed reactive process, an esterification reaction takes place for the production of ethyl acetate. The fundamental model has been derived incorporating the reaction term in the model structure of the nonreactive distillation process. The process operation is simulated at the startup phase under total reflux conditions. The open-loop process dynamics is also addressed running the batch process at the production phase under partial reflux conditions. In this study, a neuro-estimator based generic model controller (GMC), which consists of an ANN-based state predictor and the GMC law, has been synthesized. Finally, this proposed control law has been tested on the representative batch reactive distillation comparing with a gain-scheduled proportional integral (GSPI) controller and with its ideal performance (ideal GMC).     

Nonlinear model-based control algorithm for a distillation column using software sensor

This paper presents the design of model-based globally linearizing control (GLC) structure for a distillation process within the differential geometric framework. The model of a nonideal binary distillation column, whose characteristics were highly nonlinear and strongly interactive, is used as a real process. The classical GLC law is comprised of a transformer (input-output linearizing state feedback), a nonlinear state observer, and an external PI controller. The tray temperature based short-cut observer (TTBSCO) has been used as a state estimator within the control structure, in which all tray temperatures were considered to be measured. Accordingly, the liquid phase composition of each tray was calculated online using the derived temperature-composition correlation. In the simulation experiment, the proposed GLC coupled with TTBSCO (GLC-TTBSCO) outperformed a conventional PI controller based on servo performances with and without measurement noise as well as on regulatory behaviors. In the subsequent part, the GLC law has been synthesized in conjunction with tray temperature based reduced-order observer (GLC-TTBROO) where the distillate and bottom compositions of the distillation process have been inferred from top and bottom product temperatures respectively, which were measured online. Finally, the comparative performance of the GLC-TTBSCO and the GLC-TTBROO has been addressed under parametric uncertainty and the GLC-TTBSCO algorithm provided slightly better performance than the GLC-TTBROO. The resulting control laws are rather general and can be easily adopted for other binary distillation columns.     

基于支持向量分类器PCA方法及其在过程监控和故障检测中的应用

提出了基于支持向量分类器对过程进行性能监控和故障检测的改进 PCA方法 ,该方法避免了多元统计过程控制(MSPC)假设主元必须服从正态分布的前提。此外 ,通过对苯 -甲苯两组分精馏分离过程的仿真研究表明 ,该方法是有效的 ,并具有比传统多元统计过程控制更为优越的性能。     

Nonlinear control of a multicomponent distillation process coupled with a binary distillation model as an EKF predictor

The work is devoted to design the globally linearizing control (GLC) strategy for a multicomponent distillation process. The control system is comprised with a nonlinear transformer, a nonlinear closed-loop state estimator [extended Kalman filter (EKF)], and a linear external controller [conventional proportional integral (PI) controller]. The model of a binary distillation column has been used as a state predictor to avoid huge design complexity of the EKF estimator. The binary components are the light key and the heavy key of the multicomponent system. The proposed GLC-EKF (GLC in conjunction with EKF) control algorithm has been compared with the GLC-ROOLE [GLC coupled with reduced-order open-loop estimator (ROOLE)] and the dual-loop PI controller based on set point tracking and disturbance rejection performance. Despite huge process/predictor mismatch, the superiority of the GLC-EKF has been inspected over the GLC-ROOLE control structure.     

A hybrid feedback linearizing-Kalman filtering control algorithm for a distillation column

This paper studies the design of a discrete-time multivariable feedback linearizing control (FLC) structure. The control scheme included (i) a transformer [also called the input/output (I/O) linearizing state feedback law] that transformed the nonlinear u-y to a linearized v-y system, (ii) a closed-loop observer [extended Kalman filter (EKF)], which estimated the unmeasured states, and (iii) a conventional proportional integral (PI) controller that was employed around the v-y system as an external controller. To avoid the estimator design complexity, the design of EKF for a binary distillation column has been performed based on a reduced-order compartmental distillation model. Consequently, there is a significant process/predictor mismatch, and despite this discrepancy, the EKF estimated the required states of the simulated distillation column precisely. The FLC in conjunction with EKF (FLC-EKF) and that coupled with a measured composition-based reduced-order open-loop observer (FLC-MCROOLO) have been synthesized. The FLC structures showed better performance than the traditional proportional integral derivative controller. In practice, the presence of uncertainties and unknown disturbances are common, and in such situations, the proposed FLC-EKF control scheme ensured the superiority over the FLC-MCROOLO law.     

Multi-objective optimization for model predictive control

This paper presents a technique of multi-objective optimization for Model Predictive Control (MPC) where the optimization has three levels of the objective function, in order of priority: handling constraints, maximizing economics, and maintaining control. The greatest weights are assigned dynamically to control or constraint variables that are predicted to be out of their limits. The weights assigned for economics have to out-weigh those assigned for control objectives. Control variables (CV) can be controlled at fixed targets or within one- or two-sided ranges around the targets. Manipulated Variables (MV) can have assigned targets too, which may be predefined values or current actual values. This MV functionality is extremely useful when economic objectives are not defined for some or all the MVs. To achieve this complex operation, handle process outputs predicted to go out of limits, and have a guaranteed solution for any condition, the technique makes use of the priority structure, penalties on slack variables, and redefinition of the constraint and control model. An engineering implementation of this approach is shown in the MPC embedded in an industrial control system. The optimization and control of a distillation column, the standard Shell heavy oil fractionator (HOF) problem, is adequately achieved with this MPC.     

混沌算法在成型机节能中的应用

电火花成型机加工工艺复杂,加工过程中受影响的因素众多,其中电参数的选择对加工结果有很大的影响,但影响规律很难用精确的数学模型来表达.采用基于混沌神经网络算法的Bp网络对电火花成型机加工工艺效果进行了预测,结果表明:该预测模型能有效地避免BP算法在能量的最小化过程中陷入局部极小化的问题而得到最优解,最终能够很好地映射出电参数与该加工网络电火花加工工艺效果之间的关系.