Distributed Dynamic Integration of Open Architecture Systems

This paper proposes a new paradigm called ’distributed dynamic integration’ that attempts to dynamically implement system integration according to different application stages, system modules, and computing modes wherein each process stage possesses several characteristics. The concept, modelling, structure, and strategies of distributed dynamic integration are discussed. To incorporate the concept of distributed dynamic integration, two new methodologies, namely ’activity model’ and ’multiple-concept object-oriented method’, are presented. Based on an appropriate combination of the advantages of both the activity model and the object-oriented methods, activity-characterised objects and object-characterized activities are developed and applied in the construction of reusable sub-systems employed in the distributed dynamic integration. A machining simulation/monitoring application system that complies with the distributed dynamic integration strategies is designed to illustrate the effectiveness of such a paradigm. INTRODUCTION Over the past several years, extensive research has been carried out in the area of flexible automation and intelligent manufacturing. Different techniques have been employed in the development of integrated intelligent systems with the aim of improving the flexibility of manufacturing systems. The primary focus is on applications that are flexible in specific domains. For example, open system architecture for controls within automation systems (OSACA) concentrates on the open architecture research [1,2]. Re-configurable manufacturing systems, component software environment, software integrated chips (SIC) for CNC systems, and virtual instrumentation (VI), which essentially aim to modularize and encapsulate sub-function systems, have been proposed in order to address the dynamic and variable market requirements [1,3,4,5]. Integrated artificial intelligent (AI) decision support also plays an important role in the performance of flexible automation and manufacturing systems [6,7]. In addition, tele-manufacturing has also been proposed to incorporate the capability of network communication and remote control in a flexible system [8]. As manufacturing activities become increasingly concurrent and distributed, domain-specific flexible systems have limited applications, such is the case for shop floor level distributed monitoring. The basic reason is that in concurrent and distributed manufacturing, many factors are involved, such as network-based resource sharing and remote control. In addition, many applications embrace non-numeric, dynamic, and uncertain problems, and need to employ different or hybrid tools at different stages. In order to cope with such a situation, an overall multi-level support and consistent flexibility-enabling paradigm is needed and should encompass major elements like open architecture, sub-function system modularization, intelligent decision support, and distributed remote control. The goal of such a paradigm would be to use different elements of flexibility and dynamically integrate the required objects that include architecture components, AI techniques, client/server mode etc. This is essential as both the level of integration and the degree of intelligence contribute to system agility and automation robustness. This paper proposes such a paradigm called ’distributed dynamic integration’ that attempts to address the aforementioned objectives. Presented herein are two new methodologies, namely the ’activity model’ and the ’multiple-concept object-oriented method’, which incorporate the concept of distributed dynamic integration, with the ultimate aim of enhancing system agility and robustness. A machining simulation/monitoring system based on the distributed dynamic integration paradigm is designed to illustrate the effectiveness of such a system. INTEGRATION PERSPECTIVES FOR SYSTEM AGILITY An agile system that can be dynamically integrated should possess the following features: open architecture with modular sub-function systems, adaptive intelligent decision support, and distributed remote control. Accordingly, the implementation of system integration could be evaluated based on the four aspects of system architecture, entities, intelligence, and control mode. Open Architecture and Integration (Architectural Criteria) As is well understood, a complex intelligent manufacturing system finally needs to integrate different methods, tools, and subsystems together so as to enhance system power and at the same time, improve system flexibility. Therefore, the system architecture should aim to establish an environment that facilitates integration. Open architecture has been proven to be an effective environment for this purpose, which is indicated by five characteristics: portability, extendablility, interchangeability, scalability, and interoperability [1,2]. Sub-function System Modularization and Encapsulation (Entity Criteria) Based on an open architecture platform, application systems are established by embedding application-oriented sub-functions. A modular and encapsulated sub-system not only enables easy integration, but also greatly contributes to system reuse, reconfiguration and plug-play capability, aspects that embody system flexibility. The construction of architecture-compatible and domainoriented reusable entities is fundamental to agile integration. Hybrid AI Decision Support (Intelligence Criteria) Artificial intelligence has been successfully applied to process monitoring, troubleshooting, equipment maintenance [7], and many approaches have also been proposed to apply AI methods, techniques and paradigms to the solution of manufacturing problems [6]. Intelligence enhances flexibility and is one of the important criteria to evaluate flexibility. However, each AI tool can only solve some specific problems. In other words, they can exhibit excellent performance only in some specific areas or in a certain situation. For example, as far as condition monitoring is considered, knowledge-based systems are ideally suited for use in precise reasoning and explanation. These systems can provide extensive domain knowledge, including cases and rules. On the contrary, fuzzy logic and neural network may work better in those situations where exact information or expression is not available. The hybrid application of AI tools has thus proven to be more effective for decision support; consequently, an important criterion of versatile system integration is to facilitate dynamic loading/unloading of AI tools. Distributed Remote Control (Control Criteria) As factory operations get to be more geographically distributed, the recent trend in many manufacturing related systems, such as shop floor control, process monitoring and simulation, and manufacturing resource planning, is towards distributed network-based applications. On the other hand, with the rapid development of the information highway, data sharing and industrial control via network take into consideration the huge distributed networked applications involved and thus become an important aspect in the development of the computer integrated manufacturing system (CIMS). Distributed information management and control adapts the system to the classification of complex tasks and the independence of modules, and essentially contributes to the system flexibility. In general, an agile and dynamic system should consist of three major elements: open architecture, distributed remote control, and intelligent decision support. It is essential to integrate them into a consistent system in order to achieve the best performance. However, such integration is difficult to achieve due to the different characteristics and application levels of these three elements. On account of this, most integrated systems currently have to limit their integration tasks to some specific and narrow domains. In the forth section of this paper, an effective integration method, referred to here as distributed dynamic integration complying with the aforementioned criteria, is proposed to address this problem. ACTIVITY MODEL & MULTIPLE-CONCEPT OBJECT-ORIENTED METHOD An effective approach to the planning and grouping of the information and functionality of modules, i.e. encapsulation and modularization, has significant influence on the reuse and agility of system integration. Activity model is proposed for this purpose, and aims to establish a basis for information and functionality planning. Besides, multiple-concept object-oriented method is also proposed to establish a universal object model. Activity Model System cohesion and independence. The purpose of open architecture is to make system integration easier and more convenient, while in an integrated system, efficient information interaction is indispensable. Integration and openness essentially require the system modules to possess excellent internal cohesion while maintaining independence from other modules. To balance the internal cohesion and external independence, a simple but efficient model — activity model is presented. Under such a model, an application, however complicated it may be, can be depicted in five universal elements: input, output, support, control, and functionality. All the objects of an application system could then be integrated together by object capsulation and sub-system modularization. Concept of Activity Model. Activity model is derived from IDEF0. IDEF is an acronym for ICAM DEFinition, which is an acronym for Integrated Computer Aided Manufacturing — an initiative conducted by the U.S. Air Force in the 1970s [9]. IDEF0 was proposed for function modelling, while activity model is proposed here for module encapsulation by re-organising and planning the functional model. In addition, an activity model is a form of fusion with the object-oriented method. In essence, an activity model is generated by extracting the common features from the constituents of a system to form a generic model. By using such a model, any complex system can be represented by five basic elements: input, output, support, control, and functionality, as shown in Figure 1(a). Input refers to what should be processed; output refers to the results produced by function; support provides the necessary environment based on the function that can be run; control indicates the directions of functionality implementation; and functionality is the core processing unit of the module. In other words, an activity model is a kind of re-organization and planning of the module attributes and methods in terms of their functional characteristics. It provides an effective means for system integration while at the same time balancing the internal cohesion and external independence of the system as well. The functionality of an activity model is similar to a black box that encapsulates individual methods while interface information is well-classified and wellcharacterized by open platform criteria. To combine the advantages of both the object-oriented method and the activity model, object classes are constructed complying with the modularized structure of the activity model. On the other hand, any modular sub-function is designed to consist of one or several well-encapsulated object classes. Such combinations are illustrated in Figure 1(b) and Figure 1(c). (a). Basic Model (b). Activity-Characterised Object (c). Object-Characterised Activity Figure 1. Activity Model All constructed objects and modules are categorized into various types of object sets, while the syntax, descriptions and relationships (derivation and inheritance) of objects are listed in an index tree. Based on object sets, distributed dynamic integration of application system is accomplished through rapid configuration and dynamic linking with the assistance of some search engines. Multiple-concept Object-oriented Method Multiple-concept object-oriented method is proposed as an extension of the object-oriented method in this paper. Special emphasis is put on the construction principles of different characterized objects with this method. Under the support of the multiple-concept object-oriented method, any method, tool, algorithm, module, even computing mode or system architecture, will be constructed and encapsulated as different types of objects by different supporting principles or methods. For a distributed integrated system, there are two major types of object methods corresponding to the aforementioned integration criteria, described as follows. Architecture object method — common API. The basis for any open control systems is established by a standardized system platform [2]. In the system architecture for open control systems of OSACA, the basic element faced by users is the architecture object. These architecture objects are mainly realized via common APIs [2]. As most of the common APIs are constructed upon hardware, operating system, and communication system, they are vendor-neutral and provide interface between higher applications and lower basic hardware management. On the other hand, an open architecture platform has the same phase and plays the same role with common APIs in the overall integrated system. Therefore, common APIs for integrated systems are suitable objects to construct the system open architecture platform. Sub-system object method — component model. The modularization of sub-systems is a prerequisite for system reuse and re-configuration. A highly interactive interface is a trend and a common characteristic of re-configurable system. Nowadays, the rapid development of Windows Output Input Support Functionality Control