This paper first introduces the project of intelligent manufacturing in China and the research state of the IIMRC (Intelligent and Integrated Manufacturing Research Centre) of HUST (Huazhong University of Science and Technology), then reviews the recent advances in object- oriented and distributed artificial intelligence and puts forth the view that these advances open up the prospect of systems that will enable the true integration of enterprises. In an attempt to identify domain requirements and match them with research achievements, the paper examines the current literature and distinguishes 14 features that are common. It argues that effective enterprise-wide support could be greatly facilitated by the existence of intelligent software entities with autonomous processing capabilities, that possess coordination and negotiation facilities and are organized in distributed hierarchical states.

The paper reviews issues and trends in the field of embedded intelligent systems considering questions such as: What is intelligence? Why do we need intelligent machines and systems? What design methods are available for designing these systems? A framework for designing intelligent systems developed by the author is outlined. Three types of external behavior of machines are contrasted: programed behavior (automation), proto-intelligent behavior (self- regulation) and intelligent behavior (the capability for coping with a poorly structured and changing working environment, learning from operators, learning from own experience, self- maintenance, self-repair, and resolving conflicting demands by negotiation with other parties). The paper considers the concept of intelligent system architecture and discusses the selection of an appropriate architecture with a view to obtaining desired external behavior. In particular, the discussion centers on architectures which facilitate reliability, self-diagnosing, self- reconfiguring, safety, and incremental development. Trends towards multi-agent systems are reviewed and assessed in some detail.

The purpose of this paper is to introduce a modeling approach to define, simulate, animate, and control business processes. The intent is to introduce the undergoing methodology to build tools for designing and managing business processes. Similar to computer aided design (CAD) for mechanical parts, CAD tools are needed for designing business processes. It emphasizes the dynamic behavior of business process. The proposed modeling technique consists of a definition of each individual activity, the network of activities, a control mechanism that describes coordination of these activities, and events that will flow through these activities. Based on the formalism introduced in this modeling technique, users will be able to define business process with minimum ambiguity, take snap shots of particular events in the process, describe the accountability of participants, and view a replay of event streams in the process flow. This modeling approach, mapped on top of a commercial software, has been tested by using examples from real life business process. The examples and testing helped us to identify some of the strengths and weaknesses of this proposed approach.

This paper describes the development of a compute-mediated collaborative product development environment (COMCOPE) within the context of the construction industry, a collaborative project between Anglia Polytechnic University, the University of Strathclyde, and some industrial partners. The most prominent feature about COMCOPE lies in the particular emphasis on supporting human-human interaction across time and space through computer mediation within a distributed and networked environment. Based on a review of related research areas, traditional computer integrated construction concepts have been extended within the framework of computer supported cooperative work, which results in the COMCOPE conceptual architecture. As the core of COMCOPE architecture an interaction model has been developed to help articulate collaborative activities. Implementation issues have been addressed, and a prototype system, based on an augmented client-server model, has been outlined.

This paper introduces a plan of an intelligent information integrating platform for intelligent manufacturing. On the basis of the analysis of characteristics of knowledge and model processing in the intelligent manufacturing and structure of the CIM open system architecture and its integrating infrastructure, the structure, organization, and detailed design of servers in I3P is proposed.

In this paper, the problems of model selection and sequencing in distributed decision support systems (DDSS) are discussed. A kind of evaluation function is defined and a kind of two layer blackboard architecture is designed, such architecture is used to model selection and sequencing, underlying the distributed environment. An approach of optimal model selection is proposed.

This paper focuses on the large scale petroleum refinery manufacturing industry and has analyzed the characteristics and functional requirements of CIMS in continuous process industries. Then it compares the continuous manufacturing industry with the discrete manufacturing industry on CIMS conceptual model, and presents the functional model frame and key technologies of CIPS. The paper also proposes the optimization model and solution strategy for the CIMS in continuous industry.

This paper presents a new artificial intelligence programing language called ROOP which is machine independent and has the ability to deal with logical programming in general. ROOP is a multiple paradigm language which combines rule-based, procedure, logic and object-oriented programming techniques.

Many branches of European industry have had to recognize that their lead in the world market has been caught up with, particularly through Asian competition. In many cases a deficit of up to 30% in costs and productivity already exists. The reasons are rigid, Tayloristic company structures. The companies are not in a position to react flexibly to constantly changing environmental conditions. This article illustrates the methods of the `fractal company' which are necessary to solve the structure crisis. The fractal company distinguishes itself through its dynamics and its vitality, as well as its independent reaction to the changing circumstances. The developed methods, procedures, and framework conditions such as company structuring, human networking, hierarchy formation, and models for renumeration and working time are explained. They are based on practical examples from IPA's work with the automobile industry, their suppliers, and the engineering industry.

Recent research in fixture computer-aided design systems generally emphasizes the automation of the fixture design process, but a complete system is not yet available. Some theoretical problems for fixture intelligent computer aided design (FICAD) are discussed in this paper. The focus of the contribution is the system structure and modeling method. The system structure mainly includes production FICAD, variable FICAD, and design bases. The modeling method adopts parametric feature modeling. Based on parametric feature modeling, the feature structure and content are discussed and a feature base is set up.

The design information provided by feature based modeling facilitates bridging the gap between engineering design and manufacturing. This paper, based on the feature modeling system ZD_MCADII, introduces the function requirements and general architecture of a modern CAD system, and proposes a scheme of combining geometric modeling with feature modeling, and combining feature based design with feature recognition. It also offers an approach of implementing feature edition and feature library based on spatial constraints.

In this paper, a distributed scheduling approach of a flexible manufacturing system (FMS) is presented. A new class of Petri nets called networked time Petri nets (NTPN) for system modeling of networking environment is proposed. The distributed intelligent scheduling is implemented by three schedulers which combine NTPN models with expert system techniques. The simulation results are shown.

In this paper the scheduling problem on a single machine in a job shop production system has been considered. A general scheduling model of multiobjective on a single machine has been proposed and established. The simulated experiment is performed using simulated annealing algorithm.

In this paper, scheduling problems, including description, solving, and rescheduling, are presented in detail. The scheduling flexibility is an important factor to affect the system flexibility.

Simulation has become a very powerful tool because it can, in many cases, minimize the risk of FMS. This paper, on the basis of discussing methods of modeling, presents an approach to the hierarchical simulation, and sets up an FMS simulator which can be used for system simulation, cell simulation, and simulation of machining process according to described data structure and animation technique.

Genetic algorithm (GA) is a heuristic and random search technique mimicking nature. This paper first presents the basic principle of GA, the definition and the function of the genetic operators, and the principal character of GA. On the basis of these, the paper proposes using GA as a new solution method of the job-shop group scheduling problem, discusses the coded representation method of the feasible solution, and the particular limitation to the genetic operators.

There are several algorithms proposed in the literature for planning the sequence of tasks performed during the assembly of a mechanical part. All these algorithms are functional in the sense that they concentrate on the functionalities to be implemented. In this paper, an object- oriented approach is used in developing such an algorithm. The relational (functional) AND- OR graph generation algorithm by Homem-de-Mello and Sanderson is taken as the basis for the new approach. The equivalence of the object-oriented model to the functional model is informally justified by showing that the AND-OR graph generated is the same in both the approaches. Advantages of using object-oriented approach are also discussed.

Scheduling of operational sequences in manufacturing processes is one of the important problems in automation. Methods of applying Petri nets to model and analyze the problem with constraints on precedence relations, multiple resources allocation, etc. have been available in literature. Searching for an optimum schedule can be implemented by combining the branch-and-bound technique with the execution of the timed Petri net. The process usually produces a large Petri net which is practically not manageable. This disadvantage, however, can be handled by a truncation technique which divides the original large Petri net into several smaller size subnets. The complexity involved in the analysis of each subnet individually is greatly reduced. However, when the locally optimum schedules of the resulting subnets are combined together, it may not yield an overall optimum schedule for the original Petri net. To circumvent this problem, algorithms are developed based on the concepts of Petri net execution and modified branch-and-bound process. The developed technique is applied to a multi-robot task scheduling problem of the manufacturing work cell.

This paper is concerned with the development of a dynamic scheduling system for job scheduling, named DYNAS3. Though many scheduling criteria for job scheduling have been proposed, most of them are impractical for use in the low-volume/high-variety manufacturing environment. DYNAS3 is a scheduling system for job scheduling in a low-volume/high- variety manufacturing environment, developed by using client/server computing technology. The system provides us with a practical facility for job scheduling which takes into account the influence of many factors such as machine setup times, cell changes, replacement machines, and load balancing between machines. Numerical testing examples are taken from a real- manufacturing factory of Japan, and high-quality results are efficiently generated.

In order to select appropriate algorithm(s) intelligently, the strategy to recognize the model characteristics is investigated and the recognition method is presented for the characteristics about the variation tendency of objective function and shape of feasible area. Finally, authors demonstrate the technical feasibility of recognizing the model characteristics in the intelligent optimization system through two examples.

In manufacturing process design, one of the key problems is fixturing planning. This paper deals with automated fixturing planning for inspection process on coordinate measuring machine (CMM) in an automatic manufacturing system environment. A planning strategy is presented with analysis about difficulties in fixturing planning. The newly developed configurable fixture system was designed which is suitable for not only measuring process but also automated fixturing planning. The planning system determines the part orientation at the first. Then it analyzes the support surface and its four sides to make a decision about the fixing style and the distribution of the fixing points. The detailed fixture structure parameters and the part position are also generated. In application, the system performed fixturing planning for complicated regular non-rotational parts.

There is a lack of information flow from job shop environment to the CAPP (computer aided process planning) system, which prevents a CAPP system from being more practical. THCAPP-G (CAPP for garment, developed by Tsinghua University in 1994) is a two-stage, nonlinear, closed-loop and dynamic process planning system. It generates process plans taking advantage of the flexibility of the manufacturing process and job shop environment, according to the dynamic status of the job shop environment. Techniques of ES (expert system), CPA (critical path analysis), and heuristic method were utilized comprehensively. An information flow between job shop environment and CAPP system was realized in this system. The developed ideas and system structure are discussed mainly in this paper.

In this paper we present a nonlinear optimization method for solving engineering optimal design problems. In addition to maintaining main advantages of typical recursive quadratic methods, our algorithm uses an interior point quadratic programming (QP) subroutine as its QP solver. An implementation of the algorithm proposed in the paper has been applied to standard test problems and real engineering design problems. As the provided computational results demonstrate, our algorithm is very reliable and efficient for both standard and real optimization problems.

The 3-dimensional (3D) graphics realtime simulation is one of the difficult techniques in computer numerical control (CNC) machining. This paper presents the data structure representing the geometric volume. Based on the specification of the machining processes and the property of orthographic projection, we develop the local graphics display method for the dynamic graphics, adopting the front-to-back visible-surface determination algorithm. As a result, we realize the realtime 3D simulation in the industry controlled computer.

Probe measurement is one way to get some information about workpiece and machining status on-line. In this paper, the principle and feature of the probe system used in on-line measurement is discussed, and the algorithm for automatic measurement planning according to the information from the CAD system is presented.

In this paper, we consider some stochastic scheduling problems with a set of stochastic jobs on a manufacturing system with a single machine that is subject to multiple breakdowns and repairs. When the machine processing a job fails, the job processing must restart some time later when the machine is repaired. For this typical manufacturing system, we find the optimal policies that minimize the following objective functions: (1) the weighed sum of the completion times; (2) the weighed number of late jobs having constant due dates; (3) the weighted number of late jobs having random due dates exponentially distributed, which generalize some previous results.

A new process plan data model based on the STEP, called an object-oriented alternate process plan data model, is proposed in this paper. Many factors such as the dynamical features in a manufacturing environment and some optimization objectives in the process of product manufacturing are taken into account in the model so that the outcome model can not only depict multiple selective process plans but also dynamically suit the changing environment of manufacturing and various optimization objectives which are implemented through the technique that combines some decision-making knowledge with the model.

The object of this study is to discuss an intelligent solution to the problem of task-allocation in shop floor scheduling. For this purpose, the technique of distributed artificial intelligence (DAI) is applied. Intelligent agents (IAs) are used to realize decentralized cooperation, and negotiation is realized by using message passing based on the contract net model. Multiple agents, such as manager agents, workcell agents, and workstation agents, make game-like decisions based on multiple criteria evaluations. This procedure of decentralized cooperative problem solving makes local scheduling possible. And by integrating such multiple local schedules, dynamic flexible scheduling for the whole shop floor production can be realized.

Improving productivity through computer integrated manufacturing systems (CIMS) and concurrent engineering requires that the islands of automation in an enterprise be completely integrated. The first step in this direction is to integrate design, process planning, and scheduling. This can be achieved through a bidding-based process planning approach. The product is represented in a STEP model with detailed design and administrative information including design specifications, batch size, and due dates. Upon arrival at the manufacturing facility, the product registered in the shop floor manager which is essentially a coordinating agent. The shop floor manager broadcasts the product's requirements to the machines. The shop contains autonomous machines that have knowledge about their functionality, capabilities, tooling, and schedule. Each machine has its own process planner and responds to the product's request in a different way that is consistent with its capabilities and capacities. When more than one machine offers certain process(es) for the same requirements, they enter into negotiation. Based on processing time, due date, and cost, one of the machines wins the contract. The successful machine updates its schedule and advises the product to request raw material for processing. The concept was implemented using a multi-agent system with the task decomposition and planning achieved through contract nets. The examples are included to illustrate the approach.

This paper describes an application of simulation modeling to enhance system performance in the semi-automated lamps manufacturing industry with a recently developed multiattribute decision analysis. Multiattribute decision techniques seem to provide an easily understood, yet comprehensive, set of quantitative and qualitative approaches to justify advanced manufacturing systems. Much has been written about the use of multiattribute decision application models to evaluate manufacturing technologies. Simulation modeling approach is developed for planning and scheduling of a semi-automated manufacturing system, hence the qualitative and quantitative multiattribute factors of the manufacturing system are evaluated. A new proposal is developed to achieve a high level of productivity of plant and minimize the makespan and cost of production. An experimental comparison is performed on the effectiveness of the proposed system through a real life case study.

Simulation, as a management information tool, has been applied to engineering manufacture and assembly operations. The application of the principles to textile manufacturing (fiber to fabric) is discussed. The particular problems and solutions in applying the simulation software package to the yarn production processes are discussed with an indication of how the software achieves the production schedule. The system appears to have application in planning, scheduling, and quality assurance. The latter being a result of the traceability possibilities through a process involving mixing and splitting of material.

Neural networks have potential advantages such as real-time operation and robustness based on their parallel structure, self-organization, fuzziness, and particularly their adaptive learning ability. A single neural network is useful for identification of objects. To carry out identifying complex objects, however, it is necessary to consider hybrid architectures of two or more networks, which offer some degrees of improvement in performances. In this paper, neural learning techniques, the self-organizing feature mapping (SOFM), and learning vector quantization (LVQ2) have been applied to the automatic target recognition problem in the presence of a satellite object with high level noises. SOFM, unsupervised learning captures the homogeneity within-class characteristics; whereas LVQ2, supervised learning captures the heterogeneity of between-class.

In this paper a new coordination approach for decentralized job shop scheduling rules is presented and analyzed in a simulation study. The coordination is based on look ahead information and contains a mechanism for demanding and supplying jobs. The simulation experiment shows that the performance of conventional scheduling rules is significantly improved using the coordination mechanism.

This paper describes a technique for automatically generating NC machine programs from CAD images of a rough work piece and an optimally positioned component. The paper briefly compares the generative and variant methods of automatic machine program development and then presents a technique based on the variant method where a reference machine program is transformed to machine the optimized component. The transformed machine program is examined to remove any redundant cutter motions and correct any invalid cutter motions. The research is part of a larger project on intelligent manufacturing systems and is being conducted at the CIM Centre, Swinburne University of Technology, Hawthorn, Australia.

The management and the operation of a plant require a huge engagement of human and economic resources, as well as a continuous improvement of safety to achieve zero accident target. Revamping of existing plants is essential for these following objectives: (1) higher overall efficiency to reduce operating costs; (2) improvement of operations (easy maintenance); and (3) increasing safety by using more reliable equipment. For this purpose the authors carried out a study of heat recovery systems for petrochemical and refinery heaters, which allows the check of existing plant sections and the increasing of performances by keeping unchanged as far as possible the plant topology by adding only proper and simple packages. The procedure of calculations uses group technology techniques, settled by the authors on the basis of previous activities, in order to obtain the maximum standardization of the new designed sections (packages). The main steps are: (1) thermodynamic check of the existing plant; (2) input of the new design data and of the new total efficiency to be achieved; (3) design of the heat recovery system and auxiliary equipment; and (4) selection of items which can be manufactured on the local market with local manpower and resources.

From the application point of view of CIM or CAD/CAPP/CAM, there should be a unified part model through which the different application system can share the part information. Using feature as the basic information unit, form feature and its binary tree as the skeleton, a new practical part model regarding the rotational components is presented in this paper. The data structure of the model is discussed in detail. An application example of the model is shown in this article. The model has been found to be flexible and effective in part creation, representation, automatic process planning, and CAM.

In this paper we present a feature-based part representation scheme for explicitly describing geometric features, dimensions, tolerances and other attributes in a CIM environment. In the scheme features are constructed by the feature semantic specification, the nominal geometry, the measured entities, and attributes. Parts are then described as a graph structure in which nodes represent the feature instances, arcs describe the relationships between features. The scheme can be used to capture the designer's intent and drive the downstream activities in a CIM environment efficiently.

This paper focuses on the fundamental issues related to the development of intelligent CAD/CAM systems for progressive dies which are used for manufacturing sheet metal parts in mass production. The CAD/CAM functions for sheet metal product design and those for design and manufacture of progressive dies are defined by studying the sheet-metal-product life cycle. The procedures of industrial design and manufacture of progressive dies are reviewed, and the need for intelligent CAD/CAM systems identified. The platforms, environment, and key technologies for the systems design and implementation are discussed. Finally, a prototype CAD/CAM system for progressive dies is introduced.

This paper describes the construction of an intelligent break-even analysis system for comparison of manufacturing alternatives. Such a system integrates expert system with existing PC software tools, for example, spreadsheet, database and conventional programming packages, etc. and is used to help managers to make the decision in production process. The system is developed by using KnowledgePro, a PC-based intelligent system development environment. The advantages offered by the system are: (1) their ability to deal with both quantitative computations and qualitative reasoning; (2) their suitability for small and middle size enterprise.

Feature extraction is highly desirable in bridging CAD and CAM systems. Based on automatic examination of spatial position relationships among geometric elements, and relationships among primitive objects, a number of knowledge-based rules are defined in this paper to identify common types of features. These rules can be applied to extracting form features from a designated CSG modeling output. The application of such rules are applied to an example in developing a spatial model of a building.

Feature recognition from geometric modeling systems plays an important role in the attempt of integrating CAD and CAM systems. The non-unique tree representations of a CSG solid modeler, however, can make a recognition process very complex. In this paper, three major reasons leading to the non-uniqueness are examined, and the most commonly occurring one is dealt with in detail. Based on the achievement of position relationships between orthogonal primitive objects, an approach for solving the non-uniqueness stemming from multiple definitions for an identical object is presented.

A feature-based concurrent engineering system covering most application domains in production, i.e., design domain, manufacturing domain, assembly domain, and inspection domain, is to be developed. Efficient information representation and manipulation are the basic criteria. In this paper, a hierarchical structure with generalized feature definition is proposed as the solution. It contains the vocabulary of the semantic feature-based language for information description and introduces the concept of feature transformation which forms the grammar of the language for information deriving and exchange. Its applications in generative process planning are demonstrated.

A fuzzy logic control (FLC) implementation of a ball balancing system (BBS) is presented in this paper. A ball balancing system is built. Fuzzification of the input and output is accomplished. A set of the rules among measurements and control output are established. A multi-region method is proposed to provide a nonlinear compensation output and obtain better control results. A comparison with LQR control is made.

In this paper, the significance and characteristics of multi-parameter diagnosis of a gray fault information system of rolling bearing are analyzed. Based on gray relational analysis, a new concept of non-isobaric threshold relational grade and a new data preprocessing method for non-isobaric normalization are put forward which have been validated by experimental results.

With the rapid growth of industrial application of ANN, an intelligent diagnostic system for a large rotor system based on probabilistic neural network is developed and introduced into practice. Due to its intelligence and rapid response without human interference, it is called rapid response neural net (RRN). In this paper, the principles of construction, net architecture, and feature selection are discussed. Minimum information loss in preprocessing net and correct architecture selection are emphasized in constructing a PNN of high performance. In order to reduce the amount of real training data, the counterexamples of real data are adopted. Some training and testing results of RRN are given. The practical effects in two chemical complexes are analyzed. Both of them indicate that RRN possesses good function.

A new generation instrument was developed for condition monitoring of rotating machinery. By detecting four channels of acoustic emission (AE) signal and four channels of vibration signal, the instrument can monitor the vibration and rubbing conditions of rotating machinery on-line and locate the rubbing fault approximately. It contributes to a correct diagnosis for fault reasons by analyzing the vibration and AE signals.

This paper presents the fault diagnosis system of machining center (FDSMC). The emphasis is put on its architecture, functions, and characteristics. A tentative discussion of the applications of the theory of fuzzy sets is made in the complementation of the FDSMC.

Based on the features of flexible machining devices and processes, and the requirements of monitoring and diagnosis, a multi-parameter describing and discriminating model of machining operating mode and process condition has been presented to satisfy the real-time monitoring and diagnosis under the actual flexible machining environments. It mainly contains three parts: the analysis of feature parameters and the normalized processing, the tracing and analysis of information of the rotating speed of motors, and the identification of the change of condition caused by the workpiece outline and the operating time-series. A practical system run successfully indicates the efficiency of the model.

Based upon research on sound and vibration characteristics of tool condition, we explore the multigrade monitoring system which takes single-chip microcomputers as the core hardware. By using the specially designed pickup true signal devices, we can more effectively do the intelligent multigrade monitoring and forecasting, and furthermore, we can build the tool condition models adaptively. This is the key problem in FMS, CIMS, and even the IMS.

An effective means of designing a fault-tolerant semiconductor memory subsystem using error-correcting codes is obtained. One such code is the 1-UbEC code, the other is the 2B- UbEC code which is used to correct two unidirectional burst bytes errors. Finally, as an example of the application, the on-chip realization of 2B-UbEC code is discussed; it shows that the application of the error correcting codes for fault-tolerance is effective and convenient.

In this paper, we describe the logic function of the memory unit (MEU) board, a type of dynamic random access memory unit which includes a refresh circuit for data retention, and present a fault detection and location algorithm based on the fault models we describe. This method has been used to diagnose the MEU board of AXE-10 SPC digital telephone system.

Real-time fault diagnosis for large-scale mechatronic systems is a fairly complex problem and has not been solved up to now. Parallel processing and distributed artificial intelligence (DAI), as rapidly emerging and promising technologies, provide powerful tools for solving this hard problem. Based on the study of basic fault diagnosis process for large-scale mechatronic systems, the idea of parallel processing and DAI is introduced into the fault diagnosing field. The parallelism at four different levels in the fault diagnosis process is proposed. It lays down the theoretical basis for the development of the real-time parallel distributed fault diagnosing system.

The sweep differential equation (SDE) method is based upon analyzing and computing the boundary of the cutter swept volume in 5-axis NC milling. The points on the boundary of the cutter swept volume are shown to be a subset of the union of (1) the grazing points of the cutter for the entire sweep, (2) the ingress points of the cutter at the beginning of the sweep, and (3) the egress points of the cutter at the end of the sweep. The grazing points at each time instant are computed from a tangency function which describes the relationship between the sweep vector and the outward surface normals for all the points on the boundary of the cutter. An algorithm is developed and a software program is written for the computation and display of the tool swept volumes for a typical 5-axis NC milling machine. An example is given to demonstrate the algorithm in computing the swept volume of a ball-end cutter.

This paper presents rapid prototyping, manufacturing (RPM) and rapid tooling technology, and how it is used as a new enabling technology for concurrent engineering. Several RPM processes are introduced, then the effect of the rapid prototyping technology to concurrent engineering is discussed. With the RP technology we can get the physical model of the new developed product much more quickly. By integrating the new technology with other enabling software and concurrent engineering philosophy, the product development cycle can be reduced dramatically. Some new technology problems for further research are proposed and the conclusion is given.

Product design is the most important stage in the course of product development. All the important decision-making regarding manufacturing and assembly is made during this period. Concurrent engineering (CE) has been proposed to facilitate the whole development processes by using multi-disciplinary team-work. Under the concept of CE there is a need to provide the designers an expert system to investigate their design in the early stage of design. In this paper the construction of a product design simplification system has been explored and a prototype system has been developed. The system can be used to evaluate product design on the basis of design for assembly (DFA) principles. Design simplification can be accomplished by three approaches. The first uses the three criteria of Boothroyd and Dewhurst's Methodology, the second examines the functional implication of parts and the third considers if a part is secured after assembly. The system will give the designer advice according to the evaluation results. All the rules in the rule-base are derived from the expertise in DFA field. On the basis of the implementation of this prototype system further research directions are also suggested.

This article analyzes the NC machining method of the ruled surface in detail. On the basis of these analyses, we analyze the reasons why tolerances exist in the side milling of ruled surface and give all the equations needed in the tolerance calculation.

Computer integrated manufacturing system (CIMS), as an advanced methodology, has been applied in many industry fields. There is, however, little research on the application of CIMS in the garment industry, especially on the integrated approach to CAD, CAPP, and FMS in garment CIMS. In this paper, the current situations of CAD, CAPP, and FMS in the garment industry are discussed, and information requirements between them as well as the integrated approaches are also investigated. The representation of the garments' product data by the group technology coding is proposed. Based on the group technology, a shared data base as an integration element can be constructed, which leads to the integration of CAD/CAPP/FMS in garment CIMS.

A flexible manufacturing system (FMS) provides many benefits, such as, increased machine utilization, increased productivity, reduced labor, reduced lead time, consistent product quality, and so on, but the complexity of the manufacturing control system makes it unreliable. A multiprocessor system provides high performance and a very good environment for fault-tolerance and monitoring, so the reliability of the control system for FMS can be greatly improved. YH-MCS is a multiprocessor-based cell controller for FMS based on transputers and PCs implemented in China. This paper describes its characteristics and implementation issues on architecture, fault-tolerance, and fault location.

The paper describes a low-cost industrial inspection system (MAQVIS) developed by the authors and based on a personal computer equipped with a simple frame grabber and a digital input/output controller. Originally developed under a contract with an industrial partner for inspecting text and symbols printed on electronic components, MAQVIS can be applied to many inspection tasks since it is based on a generic template matching algorithm. The task of inspection is broken down into two steps; first, location of the symbols that must be verified in the image and second, verification of the symbols using criteria previously defined by the operator during the set-up phase. The first step is important in many applications where it is impossible, due to lack of mechanical precision, to guarantee that the object being inspected (or the symbols printed upon it) will always appear in the same position relative to the camera. The second step produces a pass/fail result for each symbol and finally for the component itself. MAQVIS can be programmed to display failed symbols on-screen and indicate the reason for the failure, or simply divert the offending component to the reject bin. Both the location and the verification steps are performed on gray-level images using algorithms specifically developed for this application. Statistical information about the number of failed components and the failure rate of individual symbols is generated and can be used to trace problems in the manufacturing process. The paper describes the system in general, and specifically the template matching algorithm used to locate symbols in the image and also the verification algorithm that validates each symbol as it is found. Results of tests performed using the system on the production line of an electronic components manufacturer are also presented.

Flexible fixturing is a necessary aspect of flexible manufacturing systems (FMS) and computer-integrated manufacturing systems (CIMS). Modular fixtures are mostly used in industry for job and small batch production. Computer-aided fixture design (CAFD) has become a research focus in implementing FMS and CIMS. Fixture configuration design is an important issue in the domain of CAFD. A review of the current research in CAFD indicates that one major problem impeding the automated generation of fixture configuration design is the negligence of study on fixture structures. This paper investigates fundamental structures of dowel-pin based modular fixtures and fixturing characteristics of commonly used modular fixture elements. A modular fixture element assembly relationship graph (MFEARG) is designed to represent combination relationships between fixture elements. Based on MFEARG, algorithms are developed to search all suitable fixturing unit candidates and mount them into appropriate positions on the baseplate with an interference check. A prototype system for automated design of dowel-pin modular fixture configurations is presented in this paper. Fixture design examples are given at the end of the paper.

This paper proposes a generic description for product and process. The objective is to define a strategy concerning the manufacturing choices. The product is structured by the mean of architecture and basic technical properties. The general process is structured at different levels concerning, respectively, function, phenomenon, and basic processes. Finally, we compare characteristics of the product with the application field of the basic process. This study is a pragmatic approach allowing the user to separate some processes and optimize the technical choice.

Neural network training requires a large quantity of samples and consumes a great deal of computing time. Despite this, we still do not know the generalization capacity of a trained neural network in a certain domain. In this paper, we propose an algorithm for training neural networks to approximate polynomials. This algorithm can work with a relatively small sample set and predict the generalization capacity of the learned neural network. Simulation results demonstrate the property of this algorithm.

This paper presents a Kohonen neural network application for on-line tool condition monitoring in a turning operation. A wavelet technique is used to decompose dynamic cutting force signal into different frequency bands in time domain. Two features are extracted from the decomposed signal for each frequency band. The two extracted features are mean values and variances of the local maxima of the absolute value of the composed signal. In addition, coherence coefficient in low frequency band is also selected as a signal feature. After scaling, these features are fed to a Kohonen neural network for diagnostic purposes. A framework of a tool condition monitoring system using the neural network is presented. The tool wear criterion and the parameters of the neural network are discussed. The experimental results show that the unsupervised neural network has great potential in an on-line tool condition monitoring system, especially when the number of input features is larger.

Process planning has been recognized as an interface between computer-aided design and computer-aided manufacturing. Since the late 1960s, computer techniques have been used to automate process planning activities. AI-based techniques are designed for capturing, representing, organizing, and utilizing knowledge by computers, and are extremely useful for automated process planning. To date, most of the AI-based approaches used in automated process planning are some variations of knowledge-based expert systems. Due to their knowledge acquisition bottleneck, expert systems are not sufficient in solving process planning problems. Fortunately, AI has developed other techniques that are useful for knowledge acquisition, e.g., neural networks. Neural networks have several advantages over expert systems that are desired in today's manufacturing practice. However, very few neural network applications in process planning have been reported. We present this paper in order to stimulate the research on using neural networks for process planning. This paper also identifies the problems with neural networks and suggests some possible solutions, which will provide some guidelines for research and implementation.

Artificial neural networks have shown their prominence for pattern recognition, signal processing, and robot manipulation, etc., but the learning convergence procedure, generally, is long. Thus in many application fields, a more efficient learning algorithm is required. In this paper, we present an available open-loop learning algorithm for the generation of binary- to-binary mappings. This learning algorithm preserves the properties of open-loop algorithm, such as fast convergence procedure and simple design, etc. Since this open-loop algorithm is based on Gram-Schmidt Orthogonalization (GSO) algorithm, the neural network is termed as orthogonal projection binary neural networks (OPBNNs). Finally, examples are given to show the efficiency of OPBNNs.

In this paper, we present an optimal test method using the adaptive resonance technique (ART) neural network for bare printed circuit boards (PCB). In this method the electrical net on a board is taken as the basic test unit. By extracting electrical net features from the CAD data file, net characteristic vectors can be formed. Using the characteristic vector as the input to the ART network, the electrical nets can be clustered into different classes. Analyzing the long time memory matrix of the ART network, we can further determine the physical meaning of each class, and then make a proper test choice of the electrical nets.

The prognosis of vibration condition for a turbo-generator set using artificial neural networks is raised in this paper. Its effectiveness is compared with the traditional prediction method using auto regressive (AR) model. Moreover, a synthetical prediction method based on multi- factor inputs is presented. The results indicate that the accuracy of prognosis using neural networks is higher and more effective.

In this paper an optimum cutting parameters selection strategy based on neural network and artificial intelligence is proposed. It combines NN with AI and solves the problems of intelligent decision-making for cutting parameters during machining process. BP algorithm and inference engine design are discussed. Application examples of the strategy are simulated. The results show that the proposed strategy is very effective.

In this research, supervised and unsupervised neural network systems are used to detect tool breakage in turning operations. Before applying the neural network, the sensory signals of the cutting force are processed in time domain into more representative forms for the neural network to make the decision correctly. The back propagation (BP) network must be trained with samples of measurements taken at tool breakage. Utilizing feature mapping, the Kohonen's self-organizing network adapts the prototype values but cannot be used effectively on-line. Only relying on the normal category, the single category-based classifier (SCBC) adapts weight values on-line so as to continuously update the normal category. Extensive tests prove that the SCBC network correctly categorized 92% of the presented experimental data.

It is very important to determine the thermal susceptible points accurately in thermal error compensation of a machine tool. These points would be used to arrange temperature measurement points and compensation thermal sources. In this paper, based on the study of relationship between temperature change and thermal deform of a machine tool, a new method to identify the position of these points by neural network has been proposed. The method uses neural network to establish a model which represents the relationship between thermal displacement and temperature of measurement points according to experimental data, then uses the derivative of thermal error to temperature change of every measurement point obtained from that model to identify the position of these thermal susceptible points. The practical application proves this method to be very effective.

This paper describes the application of neural networks to the determination of orientation of parts for CNC machining. The paper introduces the theoretical basis of neural computing and emphasizes the back propagation paradigm. A geometry for training the network to recognize the orientation of an object is described together with the required input training data set. Some initial results are presented showing how the network was capable of recognizing the orientation of a part.

This paper first gives the definitions of task similarities (TS), fuzzy schedule set (FSS) and fuzzy schedule (FS). The application of two methodologies, artificial neural network (ANN) and fuzzy similarities setting, are discussed as applied to the operations scheduling associated with textile yarn manufacturing. The two methods are applied in the same stage of operations scheduling to recognize the tasks pattern and their similarities based on the production priority rules. The fuzzy scheduling setting is only used in the ANN learning stage to generate the required data which will be established and revised by local rescheduling and simulation. The concept of fuzzy distance between task similarities employed to evaluate the task-like level and grouping of tasks into fuzzy schedule sets is discussed.

An intelligent diagnostic model for mechanical wear state based on a fuzzy neural network is proposed. The model can handle input features presented in quantitative and/or linguistic form and classify patterns in terms of fuzzy class membership values. This allows efficient modeling of fuzzy diagnosis of mechanical wear state in terms of the Ferrogram Analysis Report Sheet. The application of the model to the diagnosis of wear state of diesel engines is demonstrated. The results are very satisfactory.

Kohonen networks are well known for clustering analysis. Classical Kohonen networks for hard c-means clustering (trained by winner-take-all learning) have some severe drawbacks. Fuzzy Kohonen networks (FKCNN) for fuzzy c-means clustering are trained by fuzzy competition learning, and can get better clustering results than the classical Kohonen networks. However, both winner-take-all and fuzzy competition learning algorithms are in essence local search techniques that search for the optimum by using a hill-climbing technique. Thus, they often fail in the search for the global optimum. In this paper we combine genetic algorithms (GAs) with fuzzy competition learning to train the FKCNN. Our experimental results show that the proposed GA/FC learning algorithm has much higher probabilities of finding the global optimal solutions than either the winner-take-all or the fuzzy competition learning.

This paper introduces the principle and method of thermal deformation compensation for a coordinate grinder with artificial heating sources using neural networks.

Radial basis function (RBF) neural networks have been used extensively in many applications for their simple architecture and fast learning. This paper principally discusses the training problem of RBF classifiers which can be used for classification. For RBF classifiers, how to correctly initialize the number of network hidden nodes and their parameters is very important. Genetic-based Gaussian kernel clustering method and fuzzy decision technique are explored to complete this work. Then the network is trained further with back propagation learning algorithm in order to attain optimal performance. Results from the typical experiments are used to illustrate the power and efficiency of the method.

This paper proposes an improvement of the approximate fuzzy c-means (AFCM) clustering algorithm, called BAFCM, that is based on a fast method for edge detection using fuzzy sets which is used to initialize the c clusters effectively. Our results show that the BAFCM requires about half as much computer time as the AFCM while yielding the same accuracy as the AFCM. One may use BAFCM to accelerate AFCM processing.

In this paper genetic algorithm (GA) is developed and applied to the optimizing or rationalizing permutation problems with FMS scheduling as an example. Special methods of genetic coding for permutations, fitness function definitions, crossover and mutation operations are designed and tested. A novel catastrophe technique is suggested to rescue the GA procedure from evolution halt. The effectiveness of GA for permutation problems is identified by computational experiments.

Advertising (ads) is an important design area. Though many interactive ad-design softwares have come into commercial use, none of them ever support the intelligent work -- automatic ad creation. The potential for this is enormous. This paper gives a description of our current work in research of an automatic advertising creation system (AACS). After careful analysis of the mental behavior of a human ad designer, we conclude that case-based approach is appropriate to its intelligent modeling. A model for AACS is given in the paper. A case in ads is described as two parts: the creation process and the configuration of the ads picture, with detailed data structures given in the paper. Along with the case representation, we put forward an algorithm. Some issues such as similarity measure computing, and case adaptation have also been discussed.

The paper describes the decision making and learning strategies for determining tools and cutting conditions of an automatic decision-making and learning system embedded in the intelligent CNC system/workstation of machining centers (ADLS-CNC). The functions of the system include: selecting the correct cutting tool, determining the optimal size and number of cuts, specifying machining parameters (speed and feed) for each cut and adapting them to a specific workshop environment from learning on-line.

The shortcomings of conventional methods used for developing test sets on current automated printed circuit board (PCB) test machines consist of overlooking the information from CAD, historical test data, and the experts' knowledge. Thus, the generated test sets and proposed test sequence may be sub-optimal and inefficient. This paper presents a weighting bare PCB test method based on analysis and utilization of the CAD information. AI technique is applied for faults statistics and faults identification. Also, the generation of test sets and the planning of test procedure are discussed. A faster and more efficient test system is achieved.

Spatial layout design is a process of locating components in an available space while satisfying spatial relationships among the components, and the model of spatial layout design is used to determine the optimal locations of components in an assembly from spatial relationships between the components. Because of complicated constraints relations, the previous approaches cannot build an exact mathematical model for spatial layout design problems. This paper presents an approach that is constraint-based spatial layout design using a combination of solid modeling and optimization technology, and builds a model of constraint-based spatial layout design and describes the representation of constraints networks. This approach can overcome the above mentioned defeats. Moreover, the constraint-based model has been applied to the practical use of 'general layout design for heavy duty NC machine tools based on modular design technology' in 'Wuhan Heavy Duty Machine Tool Works (WHDMTW).'

Modular design has played an important role in design and manufacturing of machine tools. With the fast development of modern industry and the appearance of the computer integrated manufacturing system (CIMS), modular design has been developed aiming at the automatic and intelligent design. It is well known that intelligent layout becomes the bottleneck of intelligent design of mechanical products. In this paper, combined with the scheme design, an intelligent layout design methodology is proposed to solve the modular design of machine tools. The authors analyze the properties of modular design, and present the object-oriented representation of layout design knowledge. The techniques of expert system, redesign based on constraints reasoning, parametric solid modeling, and automatic assembly are studied, and the design model, problem-solving strategy, and system architecture are set up. An example is given for the applications.

Owing to the more enhanced networking, the designed data for manufacturing can be transmitted internationally. Therefore, the production of the potential product at one side of the world according to the design requirement at the other side of the world is possible and quicker than ever before. The shipment can be eliminated by producing the product at the potential market place. This paper demonstrates the development of a machining system showing the manufacture of a product based on this idea. Unix, Microsoft Windows, and NFS under the LAN and Internet environment are adopted for data communication and message passing in the devised international remote machining system. The personal computer is the server of the machining center. A C program is developed for the direct control of the machining center through DNC2 interface. The command of machining process is issued from Sun Sparc station to the personal computer through Internet. The devised system structure can also be extended to link with an automatic workpiece loading system, e.g., robot an AGV, to form an automatic machining cell for CIM. The requirements of achieving such an international remote machining cell that links machining center and robot are also discussed.

This paper presents an expert system as a solution approach to automatically determine the process routing and production of mechanical parts on a turning machine, and concerns with the realization of a CAPP system of generative type. The system is intended as an aid to the planner during the process planning. The automated solution of this problem was searched in the ambit of problem planning as search of domain dependent type, and in which the knowledge domain is used to limit and to structure the searching space. The realized model can generate the expert's procedure process routing and can be divided in the following two modules: form-features recognition and process planning. The form-features recognition module receives in input the part's geometrical description and provides in output the part's description through features tree-orientation; in such a tree the nodes represent the form- features which compose the part, while the edges connect the form-features which have this type of relation <contained in$RFAN. The results obtained with proved examples have borne out the goodness of the adopted solution. The choices taken by the operator have yielded good results, and reflect the same evaluations of the expert. Therefore, the following results are seen: (1) Good static evaluation criteria for machining reliability, and (2) Good dynamic evaluation roles for paths determination in the search space.

Hypermedia enables users to perceive an abstract level of concepts more clearly and easily than a plain text by providing powerful and maneuverable system interfaces. It has been applied to various fields including design, documentation, information retrieval, manufacturing, education, and training. However, it lacks intelligence which is most critical for intelligent systems. Expert systems (ES) have been very successful in many domain areas. However, ES also has several problems including knowledge transfer (user interface) and knowledge acquisition. Among the problems knowledge transfer can be alleviated by using a hypermedia technology, since hypermedia allows more dynamic user control of the information in the knowledge base. ES also can alleviate the inherent problem of hypermedia since it has a reasoning ability. This paper discusses an integration of ES with hypermedia, which provides the mutual benefits of each system to remedy parts of the above mentioned problems. Two architectures of integrating the two systems are presented. The LOCUS system, a prototype, has been developed to test one of the proposed architectures in this research. The integration of ES with hypermedia enhances usability, interactivity, learning ability, flexibility, adaptability, intelligence of the system and can be used in various fields (e.g., CAD/CAM design, simulation, intelligent manufacturing, training, education, etc.), since it offers several amenities as described in this paper.

Case based reasoning (CBR) is receiving increasing attention as a way of developing intelligent advisory systems. Early CBR systems were developed using AI languages. More recently, commercial tools, such as CBR Express, ReMind, and Induce-It have been developed to facilitate the rapid implementation of CBR applications. In this paper, we describe the development of a prototype plant explosion advisory system using Induce-It. We use this experience as a basis for suggesting the strengths and weaknesses of Induce-It and other commercial CBR tools.

An expert system has been developed which is capable of selecting appropriate cutting tools and determining required machining parameters for milling operations. One feature of this system which distinguishes it from conventional applications of expert systems is that it uses information stored in database files. This causes the existence of a limited umber of expertise rules stored in the knowledge base; bringing up the advantage of shorter running time for the system. Another advantage of this system is its ability to retrieve required information from computer-aided design (CAD) systems through an input file, run the program, and restore the resulting information in an output file without any user interaction. These advantages make it an ideal system to be used in computer-aided design and manufacturing (CAD/CAM) systems.

The increasing level of competition in the sheet-metal working industry requires measures to reduce costs. There is considerable potential in the simplification of process procedures. Automated handling and the further processing of bent semifinished products require a high level of manufacturing precision in bending processes. However deviations in angles of several degrees from the desired values can arise as a result of material variations within one batch. An adaptive control system is developed in order to avoid expensive further manufacturing steps. This calculates the parameters required for the correction of the process control from variations in process values measured online. These adjustment values must be communicated to the bending machine and set during the ongoing forming process. For larger series of parts the coefficients of the correction matrix are also continuously improved, so that an optimum adaptive correction can be achieved for the current scatter in the material properties. The adaptive correction procedure is particularly effective in combination with the three-point bending technology. The highest levels of angular precision can already be achieved after a short independent optimization phase. However, the current spectrum of parts for a certain manufacturing task and a defined quality requirement should provide the basis for decisions concerning the economy of this process arrangement.

This paper presents an autonomous calibration procedure for identifying robot geometric parameters using a wrist force sensor, which guides the robot end effector to track the section contour of an accurately cylindrical workpiece and to find its center. The information from the wrist sensor is needed to determine the motion direction for the end effector and to generate control strategy (hybrid control law of position and force), meanwhile the force vector is required to correct the deformation of the manipulator, which maps in turn into the joint differential vector. The system of constraint equation is in fact nonlinear, and can be linearized for the constraint surface of the cylinder. Simulation has been performed for a PUMA 760 robot and the result shows that the robot positioning accuracy can be improved to the level of the repeatability by the proposed calibration method.

A hybrid position/force controller is designed for the joint 2 and the joint 3 of the PUMA 560 robot. The hybrid controller includes a multilayered neural network, which can identify the dynamics of the contacted environment and can optimize the parameters of the PID controller. The experimental results show that after having been trained, the robot has both stable response to the training patterns and strong adaptive ability to the situation between the patterns.

Control of a class of single-link flexible robot arms with bounded unknown pay-load is addressed in this paper. The Lyapunov's direct method is used to prove the asymptotical stability of the closed-loop system under the proposed linear controller. The stability analysis is performed directly on the system's partial differential governing equation, so that no approximation is involved. Thus the 'spillover' problem, which is caused by using finite dimensional (truncated) models to approximate the infinite dimensional models of distributed parameter systems, is eliminated. It is theoretically shown that a simple linear controller with feedback of only joint variables is sufficient to stabilize asymptotically the complex system of a flexible arm with bounded unknown payload. A numerical example of a single-link flexible arm with bounded unknown payload is presented for illustration.

In this paper, based on the complex optimization theory, the gait planning of biped robots is studied theoretically and a method of biped gait planning is proposed. This method is developed by taking the general model of biped robots as the object of study, so it is suitable to all kinds of biped robots. It can be used not only in the gait plannings of static walking but in those of dynamic walking. In order that this method can be used correctly, some technicalities are explained and discussed yet in the paper.

This paper presents approaches to real-time motion planning for robot application in flexible manufacturing systems (FMS), which are based on an algorithm for fast mapping obstacles from a workspace (W-space) into a configuration space (C-space) by defining fundamental obstacles in W-space. The mapping time for a two-dimensional obstacle is about constant. To plan a collision-free path for a master robot in a multiple robot system, a free subspace is built by computing slice C-space obstacles, while swept volumes produced during master robot motion are computed as additional moving obstacles for a generation of a collision-free path for a slave robot. For robot motion planning with many degrees of freedom, several types of reactive behavior are defined to coordinate motion conflicts between different links and the approach to motion planning with a few degrees of freedom is adopted to quantitatively realize reactive behavior. In order to demonstrate the effectiveness of the proposed approaches, graphical simulation experiments are reported.

The work presented in this paper is related to the study of a system that uses optical sensors as directional antennas for locating laser sources. The use of large area optical transducers, which have a specific angular gain, allow us to design a system that is sensitive to angular displacements of a laser source within the sensor field of view. Through specific algorithms it is possible to obtain an estimate of the direction of the plane waves striking the sensor head. This can be achieved with a resolution better than 1 degree using only four discrete sensors and covering a 180 degree field of view. Since the estimate of the angular position of the laser source is done on a pulse per pulse basis, this system is not affected by any fluctuation or degradation of the signal during the time of exposure. Sensor modeling, radiometric analysis, and algorithms are presented. Specific trade-offs regarding resolution, angular coverage, and SNR are also discussed.

This paper presents a study of an omnidirectional locating sensor for guiding mobile platforms. Using a reference laser source, it is possible to use this sensor for attitude determination of a mobile platform. A specific sensor has been designed for this application and allows us to cover an azimuth of 360 degrees without any scanning or mechanical systems. The sensor head is composed of a set of large area silicon photodiodes, characterized by a specific angular gain and producing signals that are dependent on the angular position of the reference source. The locating technique produces an estimate of the angular position of the reference beam in terms of azimuth and elevation, with a resolution better than one degree, for a set of nine detectors. Providing proper positioning of the reference beam, it is possible to guide a platform avoiding any physical link along any trajectory and over great distances. The analysis of accuracy and system performance is also presented.

A parallel configuration robot is based on closed loop kinematic chains, these offer a number of advantages in terms of strength and rigidity, leading to improved accuracy in robotic assembly applications. This paper, in addition to presenting the basic design issues, discusses the control strategy for the robot. The control of the robot will be PC based, and the main component of the control circuit is an LM 628 motion control chip. A design for the control circuitry is presented in the paper. A full simulation of the design has been carried out using a commercial dynamic simulation package (CADSI DADS) and these results are presented. These include the mapping of the work envelope, the joint angles of the links to enable linear movement of the tool platform, and the torque requirements for the joint actuators.

A technique for calibrating stereo cameras to determine their relative rotations and translations without using absolute position information from external world reference points is proposed. The main idea is to calculate these rotations and translations using a single camera calibration method through a virtual world coordinate system. In this way the stereo camera calibration can be carried out using three reference points of known relative positions on the plane formed by the three points. In addition, we decompose the calibration equations into two groups. The first involves only relative rotational parameters, while the second deals only with translational parameters which can be solved analytically once the rotational parameters have been found. Extensive numerical simulations have shown that the proposed calibration technique is very robust.

The manual inspection and classification of soft fruit is a subjective activity, prone to human error. A fuzzy logic methodology has been developed to facilitate the use of automated inspection. The methodology is based on capturing and analyzing the digital images of soft fruit items. By identifying descriptive features for a specific type of fruit, and associating each feature with a set of possible membership values, a fuzzy logic classification scheme was applied. Preliminary tests were conducted on a sample of green apples, using a microcomputer-based, image processing system. The results indicate that the proposal is feasible and time efficient. Based on the average processing time for discrete items, a throughput rate of 2 items/second is predicted for a low-cost installation. Further testing and refinement of the fuzzy logic methodology is necessary before a commercial application can be developed. The use of direct color classification instead of the grayscale spectrum, is also investigated.

The paper introduces the research work on a Robot Language Conversion System. In the paper, the structure of the hardware development is discussed and two different development methods of software development have been described. Finally the paper summarizes the research achievement up to now. The system can convert the medium size structure VAL program into a PT600 program and convert the simple VAL program into different robot languages by an individual conversion system and universal conversion system, respectively.

In this paper, we propose a strategy for controlling a six DOF eye-in-hand robot manipulator to grasp a non-model-based polyhedron. A pair of suitable grasping components can be determined from the 2D images observed by a CCD camera. A grasping target point and an approach vector are calculated to direct the gripper approaches to the object. The Kalman filter techniques are used to estimate the relative position between the object and the gripper. This information is then used to determine the new target of the robot for the next control cycle. A heuristic is used to compute the joint 6 motion in order to get the best result for the Kalman filter estimation. The robot grasps the object when the gripper approaches the object close enough. Experimental results are also discussed.

The shape of non-uniform rational B-spline (NURBS) based free-form surface is determined by its control polygon. Shape modification is usually achieved by either individually changing the position or adjusting the weights at each of the control points. This paper proposed a unified method for shape modification of NURBS surface. The method provides a mathematical relationship between changes in the weight and location of control points to positional adjustment of surface points.

The recognition of an object in a complex background is one of the most difficult tasks in the machine vision application. A new algorithm for tank object identification in a complex background that features the perceptual organization is proposed in this paper. Experiment results have shown the algorithm can identify tank objects in a complicated scene effectively.

In this paper, we use the approach of fractal segmentation and improved fractal image coding, and match images by the decoded images which we use as the fractal features of original images, taking advantage of the stable features which the images of nature scenes have. Experiments show that much higher accuracy and probability of matching for images of nature scenes were obtained than using traditional methods such as mean absolute difference of graylevel or edge features.

Filtering and edge detection are widely used in image processing and computer vision. The following problems are important: computational complexity, precision, possibility to be realized in parallel, and detecting edges with a subpixel precision. By use of orthogonal polynomial theory, we propose in the present paper a new realization of image filters and their derivatives for edge detection, with a reduced and constant complexity and a good precision. In particular, we present the Hermite integration to realize Gaussian filters and the Laguerre integration for Shen-Castan filters. We show that the output of these filters can be calculated by the weighted sum of the signal values at positions determined by the roots of orthogonal polynomials. Generalization to M-D cases and to derivative calculation for edge detection, and the implementation in discrete cases are presented as well. Our method is implemented and tested for computer-generated and real images, the experimental results are satisfactory.

This paper describes an imaging technique for the determination of rough component envelopes of cast and forged components. The paper includes several image acquisition methods currently used in this area but concentrates in detail on the method known as the light stripe method. Results presented show the advantages of the light stripe method to obtain a fast and accurate 3D description of the cast and forged components. The research is part of a larger project on intelligent manufacturing systems and is being conducted at the CIM Centre, Swinburne University of Technology, Hawthorn, Victoria, Australia.

This paper presents a visual language for integrated circuit (IC) test. It abstracts the test flow as several classes which map to some forms. Test engineers can use the form-based environment which reduces the cost of developing high quality programs. Test classes are defined by AT&T C++ 2.0. Forms are implemented by X-Window. All the systems run on a SUN SPARC workstation.

A new approach for vectorization of line drawings is proposed. In this paper, contour vectorization is first applied to convert the scanned image of drawings from raster form to vector form. After the operation of contour matching, elements of skeletal vector and sub- shape are obtained. On the basis of the two types of elements, a graph G(V,E) is built up to help analyze the forked joint in line drawings, and the method of curve tracing and splitting is also discussed to fulfill the sub-optimal extraction of line and arc segments. In addition, experimental results are given showing better results than that of the traditional thinning approach.

A visual CAD platform for designing the positioning system of optical disk drives is described in this paper. The platform consists of the opto-mechanical assembly (OMA) and its simulator of an optical disk drive, a servo control board (SCB) with a high performance digital signal processor (DSP), and a personal computer (PC). The OMA and SCB form the complete digital positioning control system of an optical disk drive. The OMA simulator and SCB provide a real-time simulation environment under which a digital controller (i.e., the DSP program) is run. On a PC with the CAD software packages, the controller can be conveniently designed and loaded into SCB. Furthermore, the behavior of the disk drive and the status, parameters and structure of the controller are visualized on the PC's display without interrupting the controller. The visualization can be implemented on three levels: post-processing, tracking, and even steering.

In this paper, we generalize the Hough transform to match the edge segments in trinocular stereovision and to determine the parameters of the segments in 3-D space. We show that the corresponding segment triplet candidates can be detected by a Hough transform in the parameter plane ((theta) , (phi) ) which characterizes the 3-D segment orientation. These triplets can then be verified, and the position parameters of the 3-D segments can be detected by a Hough transform in the parameter plane (Y, Z). So the matching of geometric primitives in trinocular stereovision images can be found by the cascade of searchings in two 2-D parameter spaces only. Experimental results are satisfactory. Our method shows the following advantages: (1) Trinocular stereovision image matching is transformed into searching in 2-D parameter spaces, which much reduces the computational complexity. (2) Matching can be done completely in parallel. (3) No a priori similarity between images is needed, so very different views can be used, which improves the precision of 3-D reconstruction. (4) It is very efficient to solve false targets. (5) Our method gives good results even for partially hidden segments.

Along with the emergence of high-speed communications technologies and protocols, tools for multimedia applications are growing in terms of their variety and their performances, making multimedia data flows candidates to the integration in almost every computer system. We particularly focus on the case of the factory plant where sound and video equipments could soon take part in the control/monitoring of manufacturing processes. In this paper we propose a MMS-like multimedia application interface for such an environment.

The importance of the plunge grinding system parameter, i.e., the time constant, is discussed and the system for the grinding process is described. An in-process time constant estimation method that involves only a few mathematical operations is presented. The method is particularly useful for adaptive control of the plunge grinding process. Experimental testing of the method is explained. A comparison between the results obtained during a grinding test and that from an analysis shows that the proposed method is effective for adaptive control of the plunge grinding process.

This paper presents an intelligent PID controller based on a gradient descent learning algorithm. A BP neural network is needed to learn the characteristics of the dynamic systems. The possibility of using neural network models directly within a model-based predictive control strategy is also considered by making use of an on-line optimization routine to determine the future inputs that will minimize the deviation between the desired and predicted outputs. The controller's structure and the learning algorithm are very simple and easily realized. It can also replace the traditional PID controller, control the complex systems, and require neither process model nor more tuning parameters.

A notable characteristic of an electro-rheological fluid (ERF) is its apparent variation in viscosity with the application of an electric field. The applications of this characteristic to the performance of controllable devices on vibration control of manufacturing systems are proposed here. It can be shown experimentally that by changing the applied electric field on an ER fluid, the machining chatter can be reduced remarkably.

The accurate detection and automatic control of liquid level has a positive effect on the safety of production as well as the quality and cost of products. For a long time, however, this problem has not yet been solved satisfactorily, because of the restriction of the physical and chemical properties of the measured medium and the condition of circumstances. This paper presents a new method of detecting liquid levels on-line accurately and of following them up automatically, in terms of refractivity of liquid boundary. This method, solving the problem in question perfectly and having a series of advantages, therefore, is at present applied widely.

The problem of pole placement for linear stochastic systems is considered. The system, represented in state space form, is formulated into a singular pencilsystem model whose internal variables and parameters are estimated using an ordinary Kalman filter. The estimated internal variables and parameters are then used in the state feedback controller to place the poles of the closed loop system onto desired locations.

Velocity feedback based actin vibration control of axially moving continua, which include mechanical systems such as high-speed magnetic and paper tapes and band saws, is presented in this paper. Lyapunov direct method is used to prove the asymptotic stability of the closed- loop system under the proposed control law. The stability analysis is performed directly on the system's partial differential equation so that no approximation is involved. Thus the 'spillover' problem, which is caused by using finite dimensional (truncated) models to approximate the infinite dimensional models of distributed parameter systems, is eliminated. The proposed control law can be implemented in either distributed or discrete forms. The sufficient condition for stabilizing the system using non-collocated sensors and actuators is also given. A numerical example of an axially moving string is presented for illustration.

It is well known that point-to-point positioning can be well achieved by a conventional PID controller. However, precise tracking of preplanned motion trajectories requires more sophisticated model-based control algorithms. This paper presents a study on the design and implementation of precise tracking controllers for positioning tables during high-acceleration high-speed motion. A controller incorporating both acceleration feedforward and robust friction compensation is theoretically analyzed under various operating conditions and model uncertainties. An effective technique for tuning feedback gains was found, allowing the determination of the largest well-damped feedback gains. The control design was tested on an industrial belt-drive high-speed positioning table, and much better results were obtained than those of conventional controllers.

The objective of this paper is to present a new control design for precisely controlling overhead traveling cranes using microprocessors. Overhead cranes are widely used in manufacturing factories and are currently controlled by human operators. Automatic control can bring a number of important benefits, but existing algorithms tend to result in steady-state positioning errors in the presence of friction. In the proposed control design, a state feedback control algorithm is used to guarantee stability, a trajectory planner used to avoid saturation, and an additional integral term is used to eliminate steady-state error. Theoretical analysis and experimental implementation on a laboratory crane verifies the effectiveness of the approach.

Through the modeling of robots working concurrently and continuously, in this paper the MSS and the MPNS are introduced in succession, first heuristically and then formally. In order to unify the continuous and discrete behaviors of the system, the leap (flying) axioms of countably infinite cardinal and of continuum (preset GCH) cardinal on a strictly partial ordered (SPO) set of an occurring structure of an MSS are given.

In order to model and solve scheduling problems, a new Petri Nets Model, Reduced Petri Nets (RePN), is introduced. Compared with previous Petri Nets models of scheduling problems, the RePN model not only has strong ability of analysis but also reduces the number of places and transitions in a great amount. So the reachable set of the RePN model is much smaller, which makes it more convenient to search the optimal or near-optimal solution of scheduling problems.