Knowledge induction from uncertain information systems

Rough set theory is an important method to deal with imprecise and vague knowledge. However, one of the difficult problems of rule induction is that the classic rough set theory cannot extract rules from those information systems which include uncertain continuous attribute values. In this paper, a new rough set approach that integrates fuzzy set theory is presented to induce knowledge in this kind of information system. Fuzzy similarity relationis used as the base of similarity classification of each object in a universe and the definition of upper and lower approximation of an object set X⊆U is proposed. A decision table includes uncertain continuous data and is divided into two subtables for rules induction, one being a consistent decision subtable and the other a completely inconsistent decision subtable; two different new methods for the two subtables are proposed to induce decision rules. At the end of the paper, an example is given for further illustration.     

DIAGNOSTIC RULE EXTRACTION FROM TRAINED FEEDFORWARD NEURAL NETWORKS

This paper describes a method of extracting diagnostic rules from trained diagnostic feedforward neural nets that are constructed to recognise different mechanical faults using automated weight and structure learning algorithms. The rule extracting method is based on an interpretation that considers hidden neurons as partitions in the input space. An initial set of rules is then generated from the training data and the subspaces defined by the partitions. A procedure consisting of a number of algorithms is then used to simplify and reduce the set of initial rules step by step. To demonstrate and evaluate the rule extraction method, diagnostic rules for detecting a high-pressure air compressor's (HPAC) suction and discharge valve faults were extracted from static measurements including temperatures and pressures of various stages of the compressor.     

整合规则下的结构拓扑自组织进化设计

结构拓扑自组织进化规则的选择对结构拓扑的进化过程有决定性影响。与采用单一进化规则体系的结构进化过程相比,本文提出采用基于应力场的局部间接力学规则和基于拓扑结构质量分布的BP神经网络的局部直接规则相结合的方法,在此基础上的结构拓扑自组织进化过程可以大幅节省运算时间,同时收敛性较好。通过对一典型悬臂深梁模型的实际拓扑进化模拟,表明了该方法的可行性和有效性。     

Vehicle Travel Time Models for AGV Systems under Various Dispatching Rules

The design and evaluation of AGV-based material handling systems are highly complex because of the randomness and the large number of variables involved. Vehicle travel time is a fundamental parameter for solving various flexible manufacturing system (FMS) design problems. This article presents stochastic vehicle travel time models for AGV-based material handling systems with emphasis on the empty travel times of vehicles. Various vehicle dispatching rules examined here include the nearest vehicle selection rule and the longest idle vehicle selection rule. A simulation experiment is used to evaluate and demonstrate the presented models.     

Evaluation of dispatching rules for cellular manufacturing

In this paper, the effects of various dispatching rules on the operation and performance of cellular manufacturing systems (CMS) are evaluated. When the study of a CMS considers the automated material handling, it is crucial to reduce the gridlock probability (i.e., the probability of an unsuccessful load transfer attempt occurring in the interface point between the intercell and intracell handling system). Preventing an unsuccessful load transfer is critical for the operation of the entire system as a blockage between the automated guided vehicle (AGV) and the overloaded cell results in a total system shutdown. The gridlock probability is influenced by the dispatching rule used to schedule the load transfers in the system. Therefore, in order to reduce this probability it is necessary to use a dispatching rule that will decrease the number of waiting loads in the transfer spurs. The main objective of the paper presented herein is to identify a dispatching rule that maintains the system operational at all times. A group of dispatching rules, including the first come first served (FCFS), shortest imminent operation (SI), longest imminent operation (LI), most remaining operations (MRO), shortest processing time (SPT), shortest remaining process time (SR), and a newly developed rule proposed by the authors, loads with the minimum number of processing first (MNP), are tested and evaluated with respect to whether the capacity of the transfer spurs of the cells is exceeded. This paper presents a simulation model of a cellular manufacturing system, which is used to further explore the effects of the dispatching rules on the system performance. The results show superior performance of the newly proposed MNP rule.     

Estimation of Part Waiting Time and Fleet Sizing in AGV Systems

As manufacturing systems have grown in size and complexity, material flow control has become one of the key issues for system efficiency, and determination of the number of vehicles required is an important issue in the design of the AGV (automatic guided vehicle) systems for automated material flow control. In an AGV system, a part issues a delivery request for its transportation, and then an empty vehicle is assigned based on a pre-determined vehicle selection rule and provides delivery service. This research presents a fleet sizing procedure for an AGV system with multiple pickup and delivery stations. A queueing model is used to estimate part waiting times. The fleet sizing procedure estimates the minimum number of vehicles needed to ensure a predefined part waiting time limit. While most stochastic models assume first-come-first-served or random vehicle selection rules for the selection of an empty vehicle, this model considers such additional rules as the nearest vehicle selection rule, which is the most popular among all vehicle selection rules. The performance of the proposed model is examined through computational experiments. This work was supported by Korea Research Foundation Grant (KRF-2003-042-D20136).     

Evaluation of dispatching rules for cellular manufacturing

In this paper, the effects of various dispatching rules on the operation and performance of cellular manufacturing systems (CMS) are evaluated. When the study of a CMS considers the automated material handling, it is crucial to reduce the gridlock probability (i.e., the probability of an unsuccessful load transfer attempt occurring in the interface point between the intercell and intracell handling system). Preventing an unsuccessful load transfer is critical for the operation of the entire system as a blockage between the automated guided vehicle (AGV) and the overloaded cell results in a total system shutdown. The gridlock probability is influenced by the dispatching rule used to schedule the load transfers in the system. Therefore, in order to reduce this probability it is necessary to use a dispatching rule that will decrease the number of waiting loads in the transfer spurs. The main objective of the paper presented herein is to identify a dispatching rule that maintains the system operational at all times. A group of dispatching rules, including the first come first served (FCFS), shortest imminent operation (SI), longest imminent operation (LI), most remaining operations (MRO), shortest processing time (SPT), shortest remaining process time (SR), and a newly developed rule proposed by the authors, loads with the minimum number of processing first (MNP), are tested and evaluated with respect to whether the capacity of the transfer spurs of the cells is exceeded. This paper presents a simulation model of a cellular manufacturing system, which is used to further explore the effects of the dispatching rules on the system performance. The results show superior performance of the newly proposed MNP rule.     

基于双参数方法的诊断规则不确定性问题研究

提出规则蕴含的因果关系、充分性和必要性等不确定性规则的相关概念以及需要解决的规则不确定性问题，在此基础上介绍基于双参数的不确定性诊断规则的表示形式，进一步说明了规则条件参数计算方法和规则不确定参数的获取方法。与已有重要的不确定性知识的数值处理方法进行比较表明，该方法能够表示具有复杂不确定性加权关系的诊断规则，并且能够根据规则的约束条件通过数值计算获取规则的不确定性相关参数。实际应用结果表明，基于双参数的不确定性诊断规则表示方法是行之有效的。     

Developing an optimal valve closing rule curve for real-time pressure control in pipes

Sudden valve closure in pipeline systems can cause high pressures that may lead to serious damages. Using an optimal valve closing rule can play an important role in managing extreme pressures in sudden valve closure. In this paper, an optimal closing rule curve is developed using a multi-objective optimization model and Bayesian networks (BNs) for controlling water pressure in valve closure instead of traditional step functions or single linear functions. The method of characteristics is used to simulate transient flow caused by valve closure. Non-dominated sorting genetic algorithms-II is also used to develop a Pareto front among three objectives related to maximum and minimum water pressures, and the amount of water passes through the valve during the valve-closing process. Simulation and optimization processes are usually time-consuming, thus results of the optimization model are used for training the BN. The trained BN is capable of determining optimal real-time closing rules without running costly simulation and optimization models. To demonstrate its efficiency, the proposed methodology is applied to a reservoir-pipe-valve system and the optimal closing rule curve is calculated for the valve. The results of the linear and BN-based valve closure rules show that the latter can significantly reduce the range of variations in water hammer pressures.     

基于产品特征表达的集成化制造系统的研究

提出了一种并行工程环境下基于特征表达的产品设计和制造的集成化方法。该系统提供了用于工艺规划和可制造性评估的工艺知识库,可以被产品设计的各个阶段所共享,能够确定加工特征组合和加工顺序安排。还提出了一种确定多加工方向的加工特征组合方法,用于加方向的合理选择和特征集合的最少化,获得最优化的加工效率。另外还探讨了特征干涉对加工效率和加工质量的影响,提出了新的特征干涉分类、解决方法和特征干涉条件下加工效率的优化算法。