Intelligent system architecture for process operation support

This paper presents a generic intelligent system platform, namely INTEMOR (INTElligent Multimedia system for On-line Real-time application), for process operation support. INTEMOR is developed by combining artificial intelligence, computer system and information technology into a unified environment. It integrates various function modules to perform operation support tasks, including communication gateway, data processing and analysis, on-line process monitoring and diagnosis, on-line operation manual, equipment maintenance assistance, reasoning system, knowledge-base creator and multimedia interface. The industrial applications of INTEMOR on a boiler system and a chemical pulping process are illustrated.     

Actuators fault diagnosis for robot manipulators with uncertain model

In this paper a fault diagnosis approach for robotic manipulators, subject to faults of the joints driving systems, is developed. A model-based diagnostic observer is adopted to detect, isolate and identify failures. Compensation of unknown dynamics, uncertainties and disturbances is achieved through the adoption of a class of neural interpolators, the support vector machines and trained off-line. Interpolation of unknown faults is performed by adopting an on-line neural interpolator based on radial basis functions, whose weights are adaptively tuned on-line. The effectiveness of the approach is experimentally tested on an industrial robot manipulator.     

基于在线支持向量回归的非线性模型预测控制方法

针对非线性模型预测控制中离线模型难以适应非线性对象实时变化的缺点, 提出一种基于在线支持向量回归的非线性模型预测控制方法. 该方法通过在线支持向量回归离线训练与在线学习相结合的方式, 建立具有在线校正特性的预测模型, 同时采用最速下降原理滚动优化非线性模型预测控制的目标函数, 求得多步控制量. 通过对非线性对象的控制结果表明, 所提出方法有效且具有良好的自适应性.

     

Expert system for analysis of quality in production of electronics

Quality issues have become increasingly important in the production of electronics, especially when dealing with electronic products not assimilated to the mainstream of consumer electronics, but rather to the group of industrial electronic devices and machinery designed to last for years or even decades. In this paper, an intelligent optimization and modeling system for electronics production is demonstrated. The system exploits real production data and can be used to diagnose and optimize the manufacturing processes. It contains three modules consisting of appropriate mathematical tools specifically tailored to each task: (1) preprocessing, (2) variable selection, and (3) optimization modules. Moreover, concrete examples are presented from the latter two modules, by using a wave soldering process as a case study. Currently, the system works on the Matlab platform, but can be programmed into standalone software and automated in the future. The results illustrate that the system can offer an efficient tool for diagnostics and process optimization in the electronics industry.     

在线KPLS建模方法及在磨机负荷参数集成建模中的应用

针对过程非线性、基于历史数据构建的离线模型泛化性差以及基于滑动窗口 和每样本递推更新的在线建模方法难以均衡建模精度和建模速度等问题, 提出了一种在线 核偏最小二乘(On-line kernel partial least squares, OLKPLS)建模方法. 该方法依据新样本与建模样本间的近似线性依靠(Approximate linear dependence, ALD)值和代表工业过程特性漂移幅度的 阈值, 选择有价值样本更新KPLS模型, 并采用合成数据和Benchmark平台数据对该方法进 行了仿真验证. 针对基于离线历史数据建立的融合多传感器信息的磨机负荷参数集成模型难以适应磨 矿过程时变特性的问题, 提出了基于OLKPLS和在线自适应加权融合算法的在线集成建模方 法, 并通过实验球磨机的实际运行数据仿真验证了方法的有效性.     

A traverse inspection system for high precision visual on-loom fabric defect detection

A self-contained inspection system for vision-based on-loom fabric defect detection is presented in this paper. Design and loom integration of a traversing camera sled, a camera vibration damper and a complementary back-light illumination are presented and discussed. Image acquisition strategies and traverse control are described to complete the discussion on hardware and mechanics. The main part of the paper focuses on a novel algorithmic framework for woven fabric defect detection in highly resolved (1,000+ ppi) image data. Within this scope, single yarns are tracked and measured in terms of position, size, and appearance in real time. An inspection prototype has been mounted onto an industrial loom. Extensive on-line and off-line evaluations for various fabric materials gave precise and stable detection results with few false alarms. A brief cost analysis for the prototype system is provided and completes the presentation of the system.     

On-line optimization of the Tennessee Eastman challenge problem

On-line, optimal steady state operating conditions were determined for the six modes of operation (with and without disturbances) for the industrial recycle reactor problem proposed by Downs and Vogel (J.J. Downs, E.F. Vogel, A plant-wide industrial process control problem, Computers and Chemical Engineering 17 (1993) 245–255.). The optimization and modeling problem was formulated as a sequential solution of steady state material balances through the process. Optimal operating setpoints were obtained using NPSOL (P.E. Gill et al., User's Guide for NPSOL: A FORTRAN Package for Nonlinear Programming, Technical Report SOL86-2, Stanford University, CA, 1986.) and reactor feed material balances as equality, non-linear constraints. On-line optimization results, without disturbances, compared favorably with the results obtained by Ricker (N.L. Ricker, Optimal steady-state operation of the Tennessee Eastman challenge process, Computers and Chemical Engineering 19 (1995) 949–959.). The optimization objective function tended to be broad and flat around the optimal operating conditions for all six modes of operation. The on-line, steady state, optimization algorithm compared favorably with the more complicated optimization structure designed by Yan (M. Yan, N. L. Ricker, On-line optimization of the Tennessee Eastman challenge process, in: Proceedings of the 1997 American Control Conference.). However, the algorithm presented here required less computational effort and exhibited greater convergence reliability than the work of Yan. On-line optimization was performed every 8 h and required less than 5 min calculation time. Updated model parameters were calculated every minute and filtered using a first order filter with 15 min time constant. Net profit was introduced as a tool to compare economic performance of the plant operating under a knowledgeable operator and operating under an off-line/on-line optimization algorithm for all six modes of operation. For modes 1–3, operating at the setpoints generated by the optimization algorithm provided significant increases in production rate and net profit that amounted to a 16–45% decrease in product operating costs when compared to operation of the plant at the setpoints specified by an operator. By decreasing operating costs and increasing production rate while maintaining a specified G/H ratio, the optimization algorithm increased net profit by 9.3–0.5% for modes 4–6, respectively, when compared to knowledgeable operator optimization of the plant. Also for sustained disturbances, the optimization algorithm decreased the error in the desired G/H ratio and increased process stability when compared to knowledgeable operator optimization of the plant. On-line optimization provided a maximum 1% relative increase in production rate and 1.5% relative increase in net profit compared to off-line optimization for modes 4–6 only when certain sustained, disturbances occurred. The economic justification of on-line optimization over off-line optimization depends upon the type, magnitude, and frequency of the disturbances.     

A STEP-compliant Industrial Robot Data Model for robot off-line programming systems

Recently, various robot off-line programming systems have promoted their own robot data models, resulting in a plethora of robot representation methods and unchangeable data files among CAx and robot off-line programming systems. The current paper represents a STEP-compliant Industrial Robot Data Model (IRDM) for data exchange between CAx systems and robot off-line programming systems. Using this novel representation method, most resources involved in a robot manufacturing system can be represented. The geometric and mathematic aspects of industrial robots have been defined in IRDM, so that the robot off-line programming system could have abundant information to represent robots’ kinematic and dynamic behaviors. In order to validate the proposed models and approaches, a prototype robot off-line programming system with 3D virtual environment is presented. The functionalities of IRDM not only have significant meaning for providing a unified data platform for robot simulation systems, but also have the potential capability to represent robot language using the object-oriented concept.     

Neural network modelling and control for underwater vehicles

Neural networks are currently finding practical applications ranging from ‘soft’ regulatory control in consumer products to accurate control of non-linear plant in the process industries. This paper describes the application of neural networks to modelling and control of a prototype underwater vehicle, as an example of a system containing severe non-linearities. The most common implementation strategy for neural control is model predictive control, where a model of the process is developed first and is used off-line to design an appropriate compensator. The accuracy and robustness of this control strategy relies on the quality of the non-linear process model, in particular its ability to predict the plant response accurately multiple-steps ahead. In this paper, several neural network architectures are used to evaluate a long-range model predictive control structure, both in simulation and for on-line control of vehicle depth, achieving accurate control with a smooth actuator signal.     

Optimization-based tuning of nonlinear model predictive control with state estimation

Nonlinear model predictive controllers determine appropriate control actions by solving an on-line optimization problem. A nonlinear process model is utilized for on-line prediction, making such algorithms particularly appropriate for the control of chemical reactors. The algorithms presented in this paper incorporates an extended Kalman filter, which allows operations around unstable steady-state points. The paper proposes a formalization of the procedure for tuning the several parameters of the control algorithm. This is accomplished by specifying time-domain performance criteria and using an interactive multi-objective optimization package off-line to determine parameters values that satisfy these criteria. Three reactor examples are used to demonstrate the effectiveness of the proposed on-line algorithm and off-line tuning procedure.     

Optimizing the end-point state-weighting matrix in model-based predictive control

In this paper a linear model-based predictive control (MPC) algorithm is presented, for which nominal closed-loop stability is guaranteed. The input is obtained by minimizing a quadratic performance index over a finite horizon plus an end-point state (EPS) penalty, subject to input, state and output constraints. Under certain conditions, the weighting matrix in the EPS penalty enables one to specify an invariant ellipsoid in which the input, state and output constraints are satisfied. In existing MPC algorithms this weighting matrix is calculated off-line. The main contribution of this paper is to incorporate the calculation of the EPS-weighting matrix into the on-line optimization problem of the controller. The main advantage of this approach is that a natural and automatic trade-off between feasibility and optimality is obtained. This is demonstrated in a simulation example.     

Managing complexity and variability of a model-based embedded software product line

This paper presents a framework for model-based product lines of embedded systems. We show how to integrate model-based product line techniques into a consistent framework that can deal with large product lines as they are common in industry. The framework demonstrates the strengths of model-based techniques like abstraction, support for customised representations, and a high degree of automation. In particular, we provide the following contributions: (1) to shift existing product lines towards a model-based approach, we support the (semi-) automated extraction of models from existing requirement, test, and implementation artefacts; (2) to cope with the complexity of artefacts and their interrelations in industrial product lines, we support the generation of context-specific views. These views support developers, e.g., in analysing complex dependencies between different artefacts; (3) finally, we support automated product derivation based on an integrated hardware abstraction layer. Most of the presented concepts have been inspired by challenges arising in the industrial application of product line techniques in the model-based engineering of embedded systems. We report on experiences gathered during the application of the techniques to a prototypical product line (on a rapid prototyping platform in the university lab) and to industrial sample cases (at the industry partner).     

变约束高效预测控制

为改善离线鲁棒预测控制算法的最优性引入了变终端约束集的思想,即在原离线方法基础上通过在线优化方法获得构建当前反馈矩阵的凸组合系数,然后再产生当前控制量.与原方法相比增加了在线求解的自由度,改善了系统的最优性.在此基础上,又提出了变约束预测控制(VC-MPC)算法,根据状态所处的不同稳定椭圆域确定其对应的控制约束,在取得...     

Production modelling for holistic production control

Holistic production control is a concept that introduces production optimisation by employing model-based, closed-loop control of the principal production Performance Indicators (pPIs). The concept relies on the development of a simple black-box model that describes the relation between the main pPIs and the most influential input (manipulative) variables. In this article the modelling aspects of the holistic production control implementation are presented. The main steps of the production modelling procedure are described, such as data preprocessing, the definition of pPIs, the selection of input variables and the derivation of black-box models. Particular emphasis is given to a modelling approach based on neural networks and a corresponding modelling assistant tool, which has been developed to support the modelling procedure. The approach is illustrated on the Tennessee Eastman benchmark process, where neural network models for three main production performance indicators, i.e., costs, quality and production rate, are derived.     

基于CS3000平台的聚酯过程模拟培训系统设计

现代工业生产规范现场人员的操作是提高生产安全性的一个重要方面。对员工进行模拟培训是一种行之有效的方法。为了配合现场运行的控制系统并对员工进行模拟培训,在CS3000开发平台上,进行了聚酯生产过程模拟培训系统的设计。文中以酯化过程为例,根据实验数据和反应机理建立数学模型,讨论了浆料釜液位控制、酯化釜1温度控制、EG／TPA摩尔比控制和酯化釜质量指标检测模块的CS3000实现问题。该系统有效地实现了对员工的离线模拟培训。规范了员工的操作,减少了人为事故的发生,提高了聚酯生产的安全性。     

小型计算机控制系统的设计及应用

针对因经济、技术等因素的制约,当前众多中小型工矿企业自动化水平较低、采用大型计算机系统易造成资源浪费的现状,介绍了自行研制的适合中小型生产过程特点的小型计算机控制系统,包括组态和运行两大部分。使用可视化组态平台,可快速方便地开发出经济实用的监控系统。运行平台是交互式的图形用户界面,具有数据实时采集处理、历史数据存储和浏览、报警显示和对复杂工业对象的远程控制等功能,并内嵌OPC服务器,便于对系统进行远程访问和控制。应用结果表明,该系统具有良好的开放性和可靠性,各项功能正常,能满足实际工业生产的需要。