A support vector machine based control application to the experimental three-tank system

This paper presents a support vector machine (SVM) approach to generalized predictive control (GPC) of multiple-input multiple-output (MIMO) nonlinear systems. The possession of higher generalization potential and at the same time avoidance of getting stuck into the local minima have motivated us to employ SVM algorithms for modeling MIMO systems. Based on the SVM model, detailed and compact formulations for calculating predictions and gradient information, which are used in the computation of the optimal control action, are given in the paper. The proposed MIMO SVM-based GPC method has been verified on an experimental three-tank liquid level control system. Experimental results have shown that the proposed method can handle the control task successfully for different reference trajectories. Moreover, a detailed discussion on data gathering, model selection and effects of the control parameters have been given in this paper.     

基于支持向量机的钢淬透性预测研究

支持向量机是建立在统计学理论之上的强有力的机器学习技术。本文提出了基于支持向量机模型预测钢淬透性的方法，并分析了核函数的选择对支持向量机建模的影响。以江阴兴澄钢铁公司的实际数据进行实验，结果表明，支持向量机方法有着良好的泛化能力，优于人工神经网络建模方法。     

基于仿真和正交试验的优化调度研究

针对生产系统多点协调调度问题,提出了一种基于仿真和正交试验的优化调度方法。在动态生产环境中,为了支持快速仿真调度分析,以模型元素和控制逻辑分离的原则为指导,采用面向对象方法,实现了基于模块化控制逻辑的快速仿真建模。以仿真模型作为系统目标求解函数,各种决策参数作为系统输入变量,通过正交试验优化仿真实验,确定最佳调度决策参数组合,将其应用于生产系统的调度控制。以某研究所的数控车间为例,验证了该调度方法的可行性和有效性。     

Makespan minimization for m-machine permutation flowshop scheduling problem with learning considerations

Increased complexity of current manufacturing systems together with dynamic conditions and permanent demands for flexible and robust functionality makes their management and control very difficult and challenging. Workflow simulation is an effective approach to investigate dynamic workflow scheduling policies and evaluate the overall manufacturing system performance. The results attained in simulation model can give directions on how to maximize system output when selecting an appropriate scheduling practice for a real system. In this paper, we investigate the abilities of multi-agent systems in combination with dynamic dispatching rules and failure handling mechanisms to manage dynamic environment conditions (such as machine failures) for systems in the production automation domain. We measure system robustness by systematically assessing the total system performance (e.g., number of finished products) in a number of representative test cases. We use an agent-based simulation environment, MAST, which has been validated with real-world hardware to strengthen the external validity of the simulation results. We investigated the performance of a re-scheduling component which uses four different policies that define how to adjust the system schedule in case of machine disturbances/failures. In the context of the empirical study the Complete Rerouting re-scheduling policy outperformed all other policies.     

面向代理的半导体生产线建模技术研究

为了有效分析和评价基于多代理的半导体生产线调度与控制性能,提出了面向代理的有色赋时Petri网的半导体生产线建模方法。将设备代理内部行为进行封装,建立了设备代理的有色赋时Petri网模型。同时,将半导体生产线多代理模型按层次划分为设备代理层、设备组层和系统层,有效地降低了模型的复杂性,提高了模型的可重用性。建立了分布式半导体生产线仿真平台,以实现基于面向代理的着色赋时Petri网的半导体生产线仿真定量分析与评价。以上海某半导体制造企业6英寸晶圆生产线为例,其仿真结果验证了建模与仿真的有效性。     

FAULT DIAGNOSIS APPROACH BASED ON HIDDEN MARKOV MODEL AND SUPPORT VECTOR MACHINE

Aiming at solving the problems of machine-learning in fault diagnosis, a diagnosis approach is proposed based on hidden Markov model (HMM) and support vector machine (SVM). HMM usually describes intra-class measure well and is good at dealing with continuous dynamic signals. SVM expresses inter-class difference effectively and has perfect classify ability. This approach is built on the merit of HMM and SVM. Then, the experiment is made in the transmission system of a helicopter. With the features extracted from vibration signals in gearbox, this HMM-SVM based diagnostic approach is trained and used to monitor and diagnose the gearbox's faults. The result shows that this method is better than HMM-based and SVM-based diagnosing methods in higher diagnostic accuracy with small training samples.     

An approach to process production reactive scheduling

Process manufacturing has some characteristics that make it different from other types of industry. In practice, there are many examples of process industrial settings where production resources are shared and there is no intermediate storage between two processing units. Perhaps the most important issue usually to be excluded from the academic discussion of process production scheduling/control problems is the uncertainty of the duration of real chemical processes. These properties of process manufacturing result in blocking, no-wait, and other constraints that must be taken into account during designing the production scheduling system. The first part of the paper addresses some features of process manufacturing and their influence on production scheduling. The main part of the paper analyzes the scheduling problems under the presence of typical process constraints and unpredictable process duration times and introduces a simple and reliable approach to avoid these problems. The proposed concept is presented on a case study of an industrial two-stage washing batch subprocess. It is considered that the presented approach would be relatively easy to implement in a process production scheduling system and that it can contribute to more effective process production scheduling.     

Performance Modelling and Evaluation of Dynamic Tool Allocation in Flexible Manufacturing Systems using Coloured Petri Nets: An Object-Oriented Approach

Manufacturing systems are faced with ever-increasing customisation and unstable demand. The traditional hierarchical control structures for shop floor (pre-release planning, scheduling, dispatching and activity control) are often inflexible in responding to unexpected scenario changes and are thus not robust to system disturbances. In this paper, an object-oriented approach to modelling of FMS dynamic tool allocation and control under a non-hierarchical shop floor control scheme using coloured Petri nets is presented. A client–server paradigm is used in the proposed modelling method. The complete FMS model is partitioned into individual classes (machines, magazines, tool transport system, SGVs, tool storage, etc.) thereby significantly reducing the complexity of the model to a tractable size. The system performance under different tool request selection rules is also evaluated using coloured Petri net simulation. The proposed method can provide the designer of a tool management system with a high-level and structured representation of the tool-sharing control. It also provides an effective method for prototyping and evaluating performance of object-oriented shop floor control software.     

CIM环境下连续生产过程通用虚拟生产调度系统

介绍了一种通用的面向连续生产过程分析和调度的原型系统,以及实现这类系统的一些思想、技术和方法。提出了一种连续生产装置抽象模型,并基于此模型通过面向对象编程方法,建立了通用的生产装置分析调度平台。这个平台的主体是一个复杂的生产装置分析调度对象类。     

The impact of release times, lot size, and scheduling policy in an A&T facility

A number of recent research works have focused on how to improve the performance of production systems. This paper examines the system based on a simulation model with two manufacturing cells under a re-entrant environment. With the model a set of experiments has been carried out to study how the factors influence the system performance. Different release times and lot sizes have been compared, and scheduling heuristics for both bottleneck and non-bottleneck have been discussed to capture the essence of the production system. We used ANOVA to analyze the experimental results and achieved the conclusions that: interval releasing is better than beginning releasing; lot size can improve one of the performance indicators, but deteriorate the other under interval releasing; NC policy can obtain both good due-date performance and high throughput.     

Concurrent optimisation of a computer vision system’s multiple responses

A number of recent research works have focused on how to improve the performance of production systems. This paper examines the system based on a simulation model with two manufacturing cells under a re-entrant environment. With the model a set of experiments has been carried out to study how the factors influence the system performance. Different release times and lot sizes have been compared, and scheduling heuristics for both bottleneck and non-bottleneck have been discussed to capture the essence of the production system. We used ANOVA to analyze the experimental results and achieved the conclusions that: interval releasing is better than beginning releasing; lot size can improve one of the performance indicators, but deteriorate the other under interval releasing; NC policy can obtain both good due-date performance and high throughput.     

基于仿真的生产调度优化技术研究

现有的调度方法无法同时达到既能建立精确的调度模型又能获得满意的调度结果.为此,提出了基于仿真的优化调度方法.该方法包括试验设计模块、遗传算法模块和仿真模型.首先由试验设计模块设计遗传算法的参数值的组合方案,然后由遗传算法采用每组参数值与仿真模型进行反复迭代,筛选出一组最佳的调度规则序列,以及与其对应的生产系统性能指标值.将该性能指标值反馈到试验设计模块,并选择下一组方案重新进行上述过程.当所有方案试验完毕后,对试验结果进行分析,确定最佳的遗传算法参数值组合方案,以及与该方案对应的调度结果.最后,以某研究所机加车间为例,验证了该方法的可行性和有效性.     

Work-in-process scheduling by evolutionary tuned fuzzy controllers

In this paper, an evolutionary algorithm (EA) strategy for the optimization of generic work-in-process (WIP) scheduling fuzzy controllers is presented. The EA strategy is used to tune a set of fuzzy control modules that are used for distributed and supervisory WIP scheduling. The distributed controllers objective is to control the rate in each production stage in a way that satisfies the demand for final products while reducing WIP within the production system. The EA identifies those sets of parameters for which the fuzzy controller performs optimal with respect to WIP and backlog minimization. The proposed EA strategy is compared with known heuristically tuned distributed and supervised fuzzy control approaches. Extensive simulation results show that the EA strategy significantly improves the system’s performance.     

A developed production control and scheduling model in the semiconductor manufacturing systems with hybrid make-to-stock/make-to-order products

Semiconductor manufacturing systems are one of the most complex production systems and this complexity increases when these systems produce both make-to-stock (MTS) and make-to-order (MTO) products in order to improve the production system utilization. To deal with this complexity, we present a dynamic production control and scheduling model for a semiconductor shop (fab) with hybrid MTS/MTO production environment. The proposed model encompasses two major modules: release module and dispatching module. The release module deals with two issues: prioritizing the MTS and MTO products in the job pool and determining when and which products can be released into the shop floor. The only considered issue of dispatching module is to prioritize the MTS and MTO products in the queue of each workstation whenever a machine becomes idle. To evaluate the proposed model, different performance measures for MTS and MTO products are considered. Moreover, a number of numerical experiments have been conducted by simulation studies. Simulation studies indicate that the proposed model outperforms other related well-known production control and scheduling policies in the literature.     

An integration architecture for flexible manufacturing cells

A modern flexible manufacturing cell (FMC) usually consists of several distributed control systems, including machining centres and material handling systems. Modelling an FMC to evaluate its design and performance, and translating the resulting models into operational programs are complex, but critical steps in implementing an FMC. This paper represents an architecture for the design, integration, and implementation of an FMC. The primary focus is on defining the process and functional requirements of the system, designing the system using high-level abstraction tools that ensure system integration and effective communication with users and implementors, and using efficient algorithms and effective interfaces for implementing real-time process control.     

Dissimilarity Maximization Method for Real-time Routing of Parts in Random Flexible Manufacturing Systems

This paper presents a dissimilarity maximization method (DMM) for real-time routing selection and compares it via simulation with typical priority rules commonly used in scheduling and control of flexible manufacturing systems (FMSs). DMM aims to reduce the congestion in the system by selecting a routing for each part among its alternative routings such that the overall dissimilarity among the selected routings is maximized. In order to evaluate the performance of DMM, a random FMS, where the product mix is not known prior to production and off-line scheduling is not possible, is selected for the simulation study. A software environment that consists of a computer simulation model, which mimics a physical system, a C++ module, and a linear program solver is used to implement the DMM concept. In addition to DMM, the simulation study uses two priority rules for routing (i.e., machine) selection and seven priority rules for selecting parts awaiting service at machine buffers. The results show (1) DMM outperforms the other two routing selection rules on production rate regardless of the part selection rule used, and (2) its performance is highly dependent on the part selection rules it is combined with.     

A SVM-based pipeline leakage detection and pre-warning system

A SVM-based pipeline leakage detection and pre-warning system is presented in this paper. In the system an optical cable is laid in parallel with a pipeline in the same ditch and three single mode optical fibers inside constitute the distributed vibration sensor. The sensor is based on Mach–Zehnder optical fiber interferometer and can detect the vibration signals along a pipeline in real time. Then the eigenvectors of vibration signals are extracted by “energy-pattern” method based on wavelet packet decomposition. Subsequently the vibration signals are recognized by support vector machine (SVM) through the features so that it can judge whether any abnormal event is taking place. If any abnormal event is found along a pipeline, the location is thus calculated. A series of trials in situ have been done, showing that the system is of good accuracy and real time performance both in recognition and locating.     

Scheduling in a two-machine flow-shop for earliness/tardiness under learning effect

In the literature, earliness/tardiness (E/T) problem was known as weighted absolute deviation problem, and both tardiness and earliness is very important performance criteria for scheduling problem. While total tardiness criteria provides adaptation for due date (ignoring results of earliness done jobs), it deals with only cost of tardiness. However this phenomenon has been started to change with just-in-time (JIT) production concept. On JIT production, earliness is as important as tardiness. The phenomenon of the learning effect has been extensively studied in many different areas of operational research. However, there have been a few studies in the general context of production scheduling such as flow-shop scheduling. This paper addresses the minimization of the total earliness/tardiness penalties under learning effects in a two-machine flow-shop scheduling problem. Jobs have a common due date. We present mathematical model to obtain an optimal schedule for a given job sequence. We also present heuristics that use genetic algorithm and tabu search, based on proposed properties. Furthermore, random search was used for showing the significance of the study by comparison purpose. A new set of benchmark problems is presented with the purpose of evaluating the heuristics. The experimental results show that the performance of proposed approach is quite well, especially for the instances of large size.     

Control theoretic model using PID controller for just-in-time production scheduling

Today, manufacturing companies are under the pressure to reduce the variations in product delivery to customers with minimized manufacturing costs. In this circumstance, the job completions are required to be as close to due dates, which can be the objective of just-in-time (JIT) production. Thus, this paper proposes an innovative approach using a feedback controller for JIT production scheduling. Specifically, a feedback control system using proportional, integral, and derivative (PID) controller is designed to control the trajectory of arrival times in a single machine configuration, based on which various schedules are generated. Importantly, control system gains (proportional, integral, and derivative) are synthesized to generate an optimal or close to the optimal schedule in terms of due-date deviations and system stability. More specifically, control system gains are trained using the problem sets with known optimal schedules, and these trained gains are tested. It is expected that the proposed PID controller can be effectively applied for scheduling and rescheduling under various disturbances that cause destabilization of the system.     

Developing manufacturing control software: A survey and critique

The complexity and diversity of manufacturing software and the need to adapt this software to the frequent changes in the production requirements necessitate the use of a systematic approach to developing this software. The software life-cycle model (Royce, 1970) that consists of specifying the requirements of a software system, designing, implementing, testing, and evolving this software can be followed when developing large portions of manufacturing software. However, the presence of hardware devices in these systems and the high costs of acquiring and operating hardware devices further complicate the manufacturing software development process and require that the functionality of this software be extended to incorporate simulation and prototyping.This paper reviews recent methods for planning, scheduling, simulating, and monitoring the operation of manufacturing systems. A synopsis of the approaches to designing and implementing the real-time control software of these systems is presented. It is concluded that current methodologies support, in a very restricted sense, these planning, scheduling, and monitoring activities, and that enhanced performance can be achieved via an integrated approach.