You are on page 1of 12

Computers in Industry 51 (2003) 257268

Towards a truly integrated enterprise modeling and analysis environment


Dursun Delena,*, Perakath C. Benjaminb
Department of Management Science and Information Systems, Oklahoma State University, 700 North Greenwood Avenue, Tulsa, OK 74106, USA b Knowledge Based Systems Inc., One KBSI Place, 1408 University Drive East, College Station, TX 77840, USA Received 3 May 2002; accepted 24 March 2003
a

Abstract Ever increasing complexity of systems has stimulated the development of sophisticated methods and tools for modeling and analysis of todays enterprises. Recent advances in information technology along with signicant progress in analytical and computational techniques have facilitated the use of such methods and tools in industry. However, enterprise modeling and analysis methods are yet to make a signicant impact on the decision-making processes of most companies and organizations. In this paper, we provide a detailed analysis of the major roadblocks to a broader use of enterprise modeling and analysis methods along with our methodical approach and a software implementation that addresses them. # 2003 Elsevier Science B.V. All rights reserved.
Keywords: Enterprise modeling; Conceptual modeling; Business processes; Enterprise model set; Model integration

1. Introduction Surviving in todays highly competitive and ever expanding worldwide economy requires a skillful management capable of monitoring and controlling highly complex problem situations and systems involving a growing number of interdependent parameters and variables. This phenomenon can be witnessed in a wide variety of organizations, institutions, and industries ranging from traditional manufacturing to software companies, medical facilities to government agencies. Increasing complexity of enterprise systems has stimulated the development of sophisticated methods
Corresponding author. Tel.: 1-918-594-8283; fax: 1-918-594-8281. E-mail address: delen@okstate.edu (D. Delen).
*

and tools for enterprise modeling and analysis. Information systems that integrate these tools and techniques are commonly called Decision Support Systems (DSS), Enterprise Information Systems (EIS), and more recently, Business Intelligence Systems (BIS) [1]. Fueled by tangible benets in many domains and by the synergy between academia and industry, many industries have increasingly accepted enterprise modeling and analysis methods such as optimization, simulation, Petri nets and cost analysis. Two key factors, which are fueled by a large number of research efforts carries out in the US, Europe and Japan, have accelerated the use of such methods: (1) advances in information technology (increased efciency in the collection and storage of information); and (2) signicant progress in analytical and computational techniques. Nevertheless, enterprise modeling and analysis methods remain largely unharnessed, and

0166-3615/$ see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0166-3615(03)00063-0

258

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

advances in modeling and analysis theories have yet to lter into the mainstream of managerial decisionmaking [2]. The primary reason for the limited success of enterprise modeling and analysis methods on a large industrial base is that these methods are generally very elaborate and require acute expertise to be used effectively. They operate on very intricate models of the enterprise being analyzed. Such models require specic formats and use technical jargon hardly comprehensible to the non-initiated. In addition, the dichotomy between the models created for analysis and the actual enterprises they represent has promoted the impression that enterprise analysis is complex, time consuming, and prohibitively expensive. This perception is reinforced by the following characteristics of todays modeling and analysis efforts [24].  Enterprise analysis efforts are analyst-dependent. To produce executable models, most enterprise analysis methods rely heavily on a group of experts who has considerable knowledge and experience in the domain-specific modeling and analysis tools and methods.  Enterprise analysis involves time- and communication-intensive activities. The communication between the domain experts who possess in-depth knowledge of the enterprise to be analyzed and the analysts who are experts in their particular modeling and analysis methods is probably the most critical part of the enterprise analyses effort.  A significant amount of the effort spent is not reusable. The knowledge that is transferred from domain experts to an analyst is mostly an ad hoc one and is seldom possible to reuse in other analysis efforts of different nature.  Decision-makers are not in control of the enterprise analysis effort. The prevailing approach used to model and analyze a particular problem of the enterprise depends on the knowledge and experience of the analyst and is not easily understandable to the decision-maker. These four characteristics are often viewed by decision-makers as signicant, if not insurmountable, obstacles that are far too costly to overcome. Therefore, a major challenge to increase the use of enterprise analysis methods in businesses and organizations is to provide the tools and methods that will address

those obstacles and render analysis activities more attractive to all participants. In the remainder of this paper, we provide a detail analysis of some of the major roadblocks to a broader use of enterprise modeling and analysis methods. Next, we describe a methodical approach that addresses each of those roadblocks. We then provide an overview of a truly integrated enterprise modeling and analysis software environment that implements our approach.

2. Conceptual models as intermediate enterprise representations In order to overcome the above listed obstacles, practitioners and researchers have followed an approach that consists of providing intermediate representations between a practitioners perception of an enterprise and a typical executable model of that enterprise. These approaches are commonly cited in the literature as enterprise engineering, enterprise modeling, and enterprise integration modeling [2,5,6]. Some of these research efforts resulted in well-known enterprise modeling frameworks such as CIMOSA [7], PERA [8], and ARIS [9]. Most of these modeling frameworks and reference architectures are developed for Computer Integrated Manufacturing (CIM) enterprises in supporting their enterprise integration efforts. In this context, enterprise modeling is dened as the process of creating abstractions of business functions and processes in order to conduct what-if scenarios for the purpose of validating various enterprise integration strategies [10]. Our denition of enterprise modeling is somewhat different: we do not limit the purpose of modeling to only enterprise integration efforts, or to a specic industrial segment such as CIM. We dene enterprise modeling as the process of capturing the richest possible abstraction of the enterprise from different perspectives (which we call enterprise model set) so that these enterprise models can be used for a wide variety of scenarios in business analysis and managerial decision support. In enterprise analysis, representations that lie between a decision-makers perception and an analysts model of the same enterprise are often called conceptual models in recognition of their basic conceptual nature [11]. These special-purpose representations describe the various aspects of an enterprise

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

259

with the goal of supporting a business (usually decision-making) activity. Conceptual models are formal or informal abstractions of a system that are expressed using special-purpose modeling constructs. Typically part of a modeling languages syntax, these constructs include simple graphical elements such as circle, boxes and arrows. These graphical elements are combined into easy-to-understand diagrams that can generally be augmented using annotations. Note that the use of diagrams or graphical means to represent an enterprise does not imply a lack of formality. Some of the IDEF methods [1215] and Petri net models [16] for example, have precise syntax and semantics. 2.1. Characteristics of enterprise models We use the term enterprise model set to refer to a group of conceptual models built to obtain a coherent and comprehensive picture of an enterprise. This set includes models of various types, and each type of model denes a perspective or viewpoint from which the system is considered for a given purpose, concentrating on some aspects and hiding irrelevant ones to reduce complexity [17]. An enterprise model set can include various activity, process, organization, information, and behavioral models. This diversity of model types is based on an important insight. A typical enterprise contains many different information types arising from different aspects of that enterprise: the relatively static information that might be stored in an employee database, the dynamic information involved in planning or processing, the complex array of information found in a detailed product design, etc. Consequently, a particular model type is tailored to a given kind of information, and models tailored to one type of information may be quite unsuitable with regard to another. Enterprise model sets have three critical characteristics. First, as stated above, each type of model in a set is different in nature from any other model type, that is, a model of a given type does not simply provide a view of the information known about an enterprise, but also captures information that is different from the information captured in other model types. The central way in which one model type differs from another is not in the amount of information or the characteristics of the enterprise that it describes. Rather, the difference lies in its semantic categories, the kinds of things

that are taken as primitive (processes, activities, classes, attributes, etc.) and the logical relations those categories can maintain with one another. Zachman [18] states that a signicant observation regarding these architectural representations is that each is of different in nature than the others. They are not merely a set of representations, each of which is an increasing level of detail in certain dimension than the previous one. Level of detail is an independent variable, varying within each architectural representation. The second critical characteristic of these models is that all model types are equally important in describing an enterprise. Each model type is necessary to capture different aspects of the enterprise and, ideally, all types of models should be developed to provide a comprehensive and coherent description of the enterprise. Finally, the third major characteristic of an enterprise model set is that the models constituting the set are not independent from one another. Each model describes some aspect of the enterprise that depends upon and is constrained by aspects of the enterprise described in other models. For example, the information captured in a data model may limit the execution of tasks described in a process model. The dependencies and relationships across models ultimately enable the projection of a comprehensive, consistent, and coherent enterprise view. 2.2. Benets and pitfalls of todays enterprise modeling and analysis methods The use of methods and tools based on the conceptual models benets enterprises in several important ways. First, conceptual models can be used to transfer enterprise-specic knowledge among domain experts, system analysts and other stakeholders in three steps. In the rst step, domain experts record their knowledge of the enterprise in an enterprise model set. The system analyst then studies this set to gain a good understanding of the enterprise and its characteristics. Finally, the two parties meet to discuss missing pieces of information and ambiguities in these models. Thus, the time and associated cost of knowledge transfer activities is signicantly reduced in two ways. First, the interview process, formerly an activity in which success depended largely on the analysts interviewing and the domain experts description

260

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

Fig. 1. A conceptual framework for enterprise modeling and analysis.

skills, is now replaced by the structured best-practice guidelines and procedures provided by the modeling methods. Second, the amount of time required for meetings between the two parties is dramatically reduced. Another way in which enterprise modeling positively impacts analysis efforts is reuse. Because enterprise models are not committed to a low-level representation language (such as a particular simulation language), they provide the foundation from which a variety of analysis models can be built to satisfy various goals (Delen, 2001). Enterprise models created by a domain expert can be reused by a number of analysis method specialists to build a variety of analysis models. These points are illustrated in Fig. 1. Even as the enterprise changes over time, the enterprise models can evolve to reect these changes and, hence, be reused in future analysis efforts. Finally, because analysis models are built from an explicitly represented set of conceptual models, decisionmakers and domain experts can more easily relate the results of analysis efforts to the enterprise being analyzed. These conceptual models enable them to exercise better control over these efforts and to participate more fully in the design and evaluation of alternatives. The benets of using conceptual models

therefore increase domain experts acceptance and motivations to using analysis methods. Not surprisingly, these important advantages have prompted great interest and expectations in enterprise modeling. These expectations, however, have yet to be fullled. Although enterprise models are gaining popularity in industries and organizations and are at the center of a number of success stories, particularly in business process reengineering and in enterprise integration efforts, enterprise modeling technology has yet to realize its full potential in commercial settings. Four major challenges stand in the way [2,6]: 1. Challenge of heterogeneous modeling methods and tools: This challenge originates in the critical characteristics of enterprise model sets, namely the differential and complementary nature of the models that comprise the set. Recall that the various conceptual enterprise models (data, activity, process, organizational models, etc.) each focus on a specic type of information that differs in nature from that dealt with by any other model types. Each is necessary to describe a particular aspect of an enterprise, and only a set that includes all of the models can provide a comprehensive representation of an enterprises operations.

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

261

Consequently, the use of enterprise models for controlling and monitoring an enterprise requires the generation of a number of different model types. Today, this model generation can typically be accomplished only by developing each model separately, using independent software applications with proprietary storage mechanisms and formats. Hence, it is at best cumbersome and at worst impossible for a domain expert to reuse, in some automated fashion, the information captured in a model of one type in a model of another type. This situation has made the development of a complete enterprise model set a timeconsuming and effort-intensive endeavor that requires the use of a number of disjointed software applications. 2. Challenge of model correlation: The second challenge facing enterprise modeling technology stems from the third major characteristic of enterprise modeling sets, namely the interdependency of the conceptual models that constitute the set. Because all models of an enterprise capture some aspect of the same enterprise, there will be a number of relationships and dependencies among the concepts represented in the various models. Consequently, to develop a comprehensive and coherent picture of the enterprise, the various models must be correlated to permit understanding of the relationships and constraints between the elements represented in the various models. Only such correlation can make it possible for decision-makers to detect conicts and inconsistencies between models, identify missing information, and calculate the impact of changes in one aspect of the enterprise on other aspects. However, in todays modeling development environments, each type of model is generally captured, represented, and stored using a stand-alone application. Hence, it is impossible, for all intents and purposes, to explicitly correlate enterprise models and use the correlations effectively to: (1) obtain a coherent view on the enterprise; and (2) evolve the models over time while maintaining consistency between them. 3. Challenge of representation extensibility: As discussed earlier, each modeling method focuses on a particular aspect of an enterprise. For example, a process modeling method focuses on

the processes or tasks that are performed, on the partial ordering of these tasks with respect to time, and on the objects involved in each task. A data model, on the other hand, captures information about the types of data stored and managed in the enterprise, the characteristics of the data, and the relationships and constraints between them. While soundly motivated, this restriction on the types of information managed by each modeling method leads to the third challenge facing enterprise modeling technology. When representing their knowledge using a modeling tool, decisionmakers and domain experts will often feel the need to record some information elements whose type is not supported by that particular modeling method and tool. These information types can be critical for providing a good understanding of the enterprises operations and/or for generating analysis models of the enterprise. Usually, these types of information elements are not supported by the modeling methods because they are too specic to a particular domain or enterprise and fall outside the scope of each particular method. To capture and store such information elements, practitioners are therefore reduced to using ad hoc notes and annotations (when the modeling tool supports them). Consequently, the effectiveness with which this additional information can be used (and reused) is dependent on the communication skills of the domain expert, in effect going against one of the primary reasons for using modeling methods in the rst place. 4. Challenge of enterprise model compiling: The last challenge to effective use of enterprise model technology is concerned with the automatic generation of executable analysis models (such as simulation or optimization models) from conceptual models. The goal of compiling an enterprise model into an executable model is not to reduce the role of analysis experts in the analytical process. Given the complexity and the expertise needed to apply analysis methods effectively and to interpret the results of analysis effort, this exclusion would not be a realizable goal. Rather, the goal is to minimize the nonvalue-added activities that are involved in transferring knowledge from domain experts to system analysts. Once automatically generated, an initial

262

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

executable model can be completed, ne-tuned, validated, and nally run by an expert in the corresponding analysis method. Today only a handful of commercially available software applications exist that are capable of compiling enterprise models into executable models. The reason is that automatic generation of executable models is a difcult task that requires expertise in both elds: enterprise analysis and enterprise modeling. Another issue about enterprise model compilers is the quality of the executable models generated. Today, most commercially available compilers generate executable models from one type of model (e.g. simulation models are generated from process or activity models, but not both). This restriction limits the quality of the executable models being generated as they can only rely on the description of one focused aspect of the enterprise. More effective and better initial models generated by an integrated set of enterprise models would together capture various but interdependent aspects of the enterprise. In summary, then, enterprise modeling faces four signicant challenges: (1) heterogeneous modeling methods and tools; (2) model correlation; (3) representation extensibility; and (4) enterprise model compiling. Bordie et al. [11] wrote . . . We believe that further advances in conceptual modeling require the integration of the concepts, tools, and techniques that were developed for system description . . . Although some advances have been made in conceptual model integration since then, the four challenges described above remain major roadblocks to enterprise modeling technology on a large industrial base. For a longitudinal perspective to some of the signicant research efforts in this eld, the reader is referred to [4,6,7,1923]. In the following sections, we will describe the framework for a truly integrated enterprise modeling and analysis environment that addresses the challenges described above. Section 3 focuses on the underlying conceptual approach to addressing those challenges while Section 4 provides a detailed description of the resulting software implementation. The development of the approach and of the commercial software that implements it is the result of over 6 years of continuing research and development efforts in the area of enterprise engineering, enterprise

model integration, and enterprise model compiler technologies.

3. An integrated approach to enterprise modeling and analysis Our approach to overcome the four challenges described in the previous section is the development of a truly integrated modeling environment that supports the following elements: capture of the entire enterprise model set within a single application, model integration, and extensions to the information types managed by the various model types [24]. In the following subsections, we describe how our approach provides a solution to each of the identied challenges. 3.1. Overcoming the challenge of heterogeneous modeling methods and tools Our approach to overcome the challenge of heterogeneous modeling methods and tools involves two phases. The rst phase is to provide an environment that supports the capture, representation, and storage of a variety of model types within one application. This application is composed of a variety of modules, each providing the necessary functionality for capturing and manipulating the information types relevant to a particular model type and each responsible for enforcing the rules and procedures of the associated modeling method. All models in the enterprise model set can then be concurrently viewed and manipulated within the applications uniform user interface. The information captured in the various models that constitute the enterprise model set is stored in a single integrated information base. Our second phase is to provide a mechanism for reusing information captured in one model type to create different model types. This is done by identifying relationships and dependencies among the concepts (i.e. information types) supported by the various modeling methods. The identied relationships are used to dene rules for automatically generating a given perspective of an enterprise (i.e. a given model type) from one or more existing enterprise models of different types. In this manner, the creation of a variety of model types to describe an enterprise is partly

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

263

automated, and the potential for model and model element reuse is increased dramatically. 3.2. Overcoming the challenge of model correlation Recall that the model correlation challenge is concerned with the capture, representation, and storage of inter-model relationships and the use of these relationships to maintain consistency across the enterprise model set. Our approach to addressing this problem is to provide users with the means to identify and record relationships between model elements across model types. Because the integrated modeling environment enables users to view multiple models concurrently, the identication of relationships is simply a matter of selecting the model elements to be related and the relationship that relates them. Once identied, the relation instance is stored in the integrated information base and used to maintain consistency between the related elements. Our approach to using relation instances for maintaining inter-model consistency is based on a simple but powerful observation. Inconsistencies across models in an enterprise model set can be of three types. The rst type occurs when one or more properties of an element in a model (or of the model itself) conict with properties of one or more elements in another model. An example is the duration of a process in a process model differing from the duration of an activity representing the same real-world event in an activity model. The second type occurs when an association between two elements in one model conicts with associations between elements in another model. As an example, consider a process in a process model being associated to a particular type of resource through the use-as-input relation. Suppose that the real-world event represented by the process is recorded as an activity A and the type of resource as a concept C in an activity model of the same enterprise. Then, an inconsistency between the two models occurs if C and A are not related through the input relation (or some similar relation) in the activity model. Finally, the third type of inconsistency that can occur between two models of the same enterprise happens when one of the models lacks information the other contains. The detection of this type of inconsistency is harder to automate as it may simply be the result of a conscious decision on the part of the domain

expert to describe an aspect of the enterprise with more details than those used in another model. Given the nature of inter-model inconsistencies, it is possible to automate conict detection and consistency maintenance in the following manner. The relationships between elements across models are characterized by the way they constrain: (1) the properties of the elements they relate; and (2) the relationships that these elements have with other elements in their respective models. The impact of a change on an information element in an enterprise model can then be automatically assessed based on the characterization of the relationships it bears to elements in other models. This assessment is then used by the environment to detect conicts and propagate changes automatically. 3.3. Overcoming the challenge of representation extensibility An important characteristic of enterprise models is that each type of model focuses on a very specic aspect of an enterprise and therefore is limited with regard to the types of information that it can represent. As described in previous sections, although soundly motivated, this feature of enterprise models hinders the ability of domain experts to capture information that is highly domain-specic and hence may not fall within the types of information supported by the environment. Our approach to overcoming this challenge is to enable seamlessly integrated extensions to the information types supported by the environment. Such extensions are made possible by the exible underlying structure of the integrated knowledge base. The integrated information bases underlying structure is rooted in a exible knowledge representation system, which is called the Container Object System (COS) [25]. Briey stated, the COS representation scheme is based on the separation of existential knowledge from descriptive knowledge. A container object has a non-qualitative, unique property that distinguishes it from any other object. The existence and identity of the object is therefore determined by that property. The set of descriptive information elements for that object, including the objects qualitative properties and the relationships that it bears to other objects, forms a container that is simply attached to the object and can be easily modied over time.

264

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

The integrated information base is thus structured to represent the various information types needed by modeling methods as types of containers. In this manner, an object can have multiple sets of descriptive knowledge (or containers) that are used to describe it according to various perspectives. In effect, the containers capture the information known about the model elements that represent, in the various models, the realworld object corresponding to the container object. This approach enables and facilitates extensions to the properties of existing information types (an activity which corresponds to adding a property or relation to a type of container) and the creation of new information types (an activity which corresponds to the construction of new types of containers). Because relationships are explicitly represented in the information base, new information elements and types can easily be related to existing ones, and behaviors can be assigned to those relationships in the manner described in Section 3.2. Extensions made in such a manner are seamlessly integrated with existing information elements in the information base and can be manipulated as if they were an integral part of the enterprise model set. 3.4. Overcoming the challenge of enterprise model compiling Recall from Section 2 that the generation of executable models requires detailed information typically not found in more abstract conceptual models and that, traditionally, executable models have been generated from a single perspective (i.e. a single type of model). There are three ways in which our approach addresses these problems associated with the challenge of enterprise model compiling. The rst two ways, which also provide solutions to the other challenges to enterprise modeling technology, come in the form of the integrated information base and its extensibility. The extensible property of the information base allows information specic to analysis methods to be captured, represented, and stored in the information base as well as explicitly linked to the appropriate model elements. In this fashion, information captured for the purpose of generating a particular executable model can be reused for other analysis efforts of a different nature. The integrated information base also enables enterprise model compilers to generate executable models from the total set of enterprise models.

In our integrated modeling environment, because the various models that constitute the enterprise model set are tightly coupled through meaningful relationships and constraints, executable models can be generated from a coherent and comprehensive picture of the enterprise. The third way in which our approach addresses the enterprise model compiling challenge is by providing the means to extend the functionality of the environment and to integrate such extensions through a uniform user interface. This feature allows for the total integration of the various compilers with the original environment. With such extensions, the integrated modeling environment truly supports all phases of analysis efforts, from conceptual modeling to the generation of executable models, while maximizing reuse and minimizing non-value-added activities.

4. A truly integrated modeling environment Model Mosaic is a commercial software application under development that has its aim to implement the approach described in the previous section. In addition to addressing the major challenges of enterprise modeling and analysis, the design and development of the Model Mosaic has been motivated by the need for the effortless and rapid specialization and extension of current modeling tools. The goal of the Model Mosaic is therefore to provide an integrated suite of enterprise modeling tools and an environment that can be easily be extended to provide an ideal setting for performing enterprise analysis. The impetus for the system design was to support two fundamental audiences. One is the end users or model builders. They have the task of analyzing a realworld domain, of creating and abstracting out of that domain the various enterprise models that represent different aspects of the enterprise, and of generating initial executable analysis models from these enterprise models. The second audience is the developers of modeling tools and enterprise model compilers. They are responsible for building and supporting modeling activities by maintaining and extending the environment as necessary. Model Mosaics architecture supports the activities of both audiences by: (1) supplying a team-based integrated modeling environment; and (2) providing developers with the mechanisms to

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

265

Fig. 2. Conceptual software architecture of Model Mosaic.

rapidly and easily generate inter-operable extensions to Model Mosaic. Fig. 2 illustrates the system architecture that supports these objectives. The heart of the system, the core component, provides three main pieces of functionality, the user interface shell, the database access services, and the mechanisms for model integration. The user interface shell provides the foundation for the common user interface, which includes basic user gestures support, common menus, toolbars, etc. The database access services implement the basic functionality for storing and managing models and other types of information in the integrated database via a common database interface. Finally, the core component encapsulates a powerful inference engine that is responsible for providing model integration through inter-model consistency maintenance and automatic change propagation. The rst echelon of the architecture contains the various modeling modules, called modelers, which

operate within the environment. Each modeler supports the creation and management of a single model type. Each modeler is a separate, complete component that interfaces with the core through the common database and common user interface. Each modeler provides the core component with a description of the types of model elements that must be managed in the common database. The core component uses this description to create the necessary database tables and entries as requested by the modeler. This technique allows new modelers to be added to the environment without disturbing the database. The modelers also provide their own user interfaces that extend the cores user interface to enable the creation and management of models and model elements. At the second echelon, the plug-in applications provide another level of functionality and integration. They extend the environment even further by building on the core component and one or more of the modelers. In particular, this echelon is designed to

266

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

facilitate and support the development of enterprise model compilers to create bridges between the basic modelers and a variety of analysis tools. At the third and nal tier lie the custom extensions. Custom extensions are only a degree different from applications as they are melded with the environment in a similar way. The goal is to provide an easy point of access for end users to create their own extensions to the modeling environment. This echelon allows virtually unlimited extensibility to the environment while maintaining the integrity and consistency of the formal methodologies supported by the tool.

that addresses these challenges. Some of the most signicant of these challenges are as follows: 1. The proliferation of stand-alone enterprise modeling support applications that makes it difcult for domain experts to reuse existing models and to obtain a comprehensive view of the entire enterprise. 2. The lack of automated support to integrate in a meaningful way the various models that constitute an enterprise model set and, consequently, the lack of support for maintaining consistency and propagating changes across these models. 3. The lack of support for capturing information that falls outside the scope of enterprise modeling methods and for linking that information to knowledge captured in the various enterprise models. 4. The relative immaturity of enterprise model compiling technology, which renders using en-

5. Conclusion In this paper, we have provided an analysis of the challenges facing a broader use of enterprise modeling and analysis techniques and presented an approach

Fig. 3. A screen-shot from Model Mosaic toolkit.

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268

267

terprise models to generate executable analysis models a cumbersome and time-consuming endeavor. Our proposed approach to overcoming these critical challenges is to provide an integrated modeling environment that supports: (1) the development of all types of models needed to capture the various aspects of an enterprise; (2) the seamless integration of these models and the use of inter-model relationships to automate consistency maintenance across models; and (3) information and functionality extensions to the environment. This approach has served as the foundation for the development of an integrated enterprise modeling and analysis environment called Model Mosaic. The current state of Model Mosaic toolkit implementation includes: (1) the fully developed modules to specify the enterprise model set; (2) the knowledge bases for integration rules; (3) modules to automatically generate optimization and simulation models via the use of the enterprise model set; and (4) a state-ofthe-art GUI that facilitates the creation and utilization of the enterprise model sets. Fig. 3 illustrates a screenshot of the software tool where a number of different perspectives of an enterprise are captured in their abstract forms using function, process and simulation modeling. Model Mosaic and its underlying framework satises two main objectives: (1) provide an intuitive, easyto-use, yet powerful environment to domain experts to create and manage enterprise model sets; and (2) facilitate information and functionality extensions to the environment via its plug-compatible open architecture implementation. The rst objective is attained by using state-of-the-art user interface techniques, automating the process of building and integrating models, and using an underlying integrated information base to provide a comprehensive and coherent view of the enterprise. The second objective is attained by providing a exible framework using a componentbased system development architecture, which enables developers to extend, enhance, and evolve the environment. The future plans for Model Mosaic includes adding capabilities to translate problem-specic information captured in the enterprise model set to Petri net and queuing network models for analysis. In addition, we think that a natural extension to our enterprise

modeling framework would be enhancing it with an intelligent, automated advisor. This automated advisor would aid users (decision-makers) in dening a structured problem from a given set of symptoms and identify the best tools to analyze the dened problem for the given situational parameters. Our initial thoughts, based on our previous experience, on the kind of technology to use in implementing such an intelligent advisor concentrate around a hybrid architecture that takes advantage of both expert systems and information fusion methodologies.

Acknowledgements This research effort partially funded by: (1) Virtual Enterprise Engineering Environment (VE3), DARPA, CN: AF33615-96-C-5601; (2) Information Integration for Concurrent Engineering, US Air Force, CN: AF33615-90-C-0012; and (3) Ontology Driven Integration, NASA CN: DAAH01-98-C-R124.

References
[1] E. Turban, J.E. Aronson, Decision Support Systems and Intelligent Systems, Prentice-Hall, Englewood Cliffs, NJ, 2002. [2] S.H. Lim, N. Juster, A. dePenninton, Enterprise modeling and integration: a taxonomy of seven key aspects, Computers in Industry 34 (1997) 339359. [3] D. Delen, D. Pratt, M. Kamath, Enabling multi-use, multitool models of manufacturing systems through an experimental frame expert system, Journal of Intelligent Manufacturing 12 (3) (2001) 247255. [4] J.H. Mize, H.C. Bhuskute, D.B. Pratt, M. Kamath, Modeling of integrated manufacturing systems using an object-oriented approach, Institute for Industrial Engineers Transactions 24 (3) (1992) 1426. [5] A. Gokhale, D.C. Schmidt, B. Natarajan, N. Wang, Applying model integrated computing to component middleware and enterprise applications, Communications of the ACM 45 (10) (2002) 6570. [6] U. Frank, Multi-Perspective Enterprise Modeling (MEMO) conceptual framework and modeling languages, in: Proceedings of the 35th Hawaii International Conference on System Sciences, IEEE Computer Society Press, 2002, pp. 110. [7] K. Kosanke, F. Vernadat, M. Zelm, CIMOSA: enterprise engineering and integration, Computers in Industry 40 (1999) 8397. [8] T.J. Williams, The Purdue enterprise reference architecture, Computers in Industry 24 (23) (1994) 141158.

268

D. Delen, P.C. Benjamin / Computers in Industry 51 (2003) 257268 [22] G. Bruno, R. Agarwal, Modeling the enterprise engineering environment, IEEE Transactions on Engineering Management 44 (1) (1997) 2030. [23] G.R. Jardim, A.G. Steiger, Implicit multilevel modeling in exible business environments, Communications of the ACM 45 (10) (2002) 5357. [24] D. Delen, Development of an Expert System Based Experimental Frame for Modeling of Manufacturing Systems, Ph.D. Dissertation, Oklahoma State University, Stillwater, OK, 1997. [25] L.K. Sanders, D. Browne, C. Menzel, R.J. Mayer, Container objects: a description based knowledge representation scheme, in: Proceedings of Autfact 91, 1991, pp. 155163. Dursun Delen received his BSc in industrial engineering from Istanbul Technical University, Istanbul, Turkey, in 1986; his MSc in industrial engineering from Yildiz University, Istanbul, Turkey, in 1988; and his PhD in industrial engineering and management from Oklahoma State University, Stillwater, Oklahoma, in 1997. He was appointed as assistant professor of management science and information systems at Oklahoma State University in 2001. Prior to his appointment, he worked as a research scientist for Knowledge Based Systems Inc., for more than 4 years, during which he participated and led a number of funded research projects. His research interests include systems modeling, discrete event simulation, object-oriented modeling, knowledge representation, and applied articial intelligence. Perakath C. Benjamin received his MSc in industrial engineering from the National Institute for Training in Industrial Engineering in 1983 and his PhD in industrial engineering from Texas A&M University, College Station, Texas, in 1991. As vice president of innovation and engineering at Knowledge Based Systems Inc., he manages and directs a large number of research and development projects. He has over 15 years of professional experience in managing projects that involve in systems analysis, design, development, testing, documentation, deployment, and training. His research interests include business process modeling, discrete event simulation, knowledge acquisition and representation, data mining and applied articial intelligence.

[9] A.W. Scheer, ARISBusiness Process Modeling, second ed., Springer-Verlag, Berlin, 1999. [10] A. Ortiz, F. Lario, L. Ros, Enterprise integrationbusiness processes integrated management: a proposal for a methodology to develop enterprise integration programs, Computers in Industry 40 (1999) 155171. [11] M.L. Brodie, J. Mylopoulos, J.W. Schmidt, On Conceptual Modeling, Springer-Verlag, New York, 1984. [12] Knowledge Based Systems Inc. (KBSI), IDEF Function Modeling Report, Air Force Wright Aeronautical Laboratory Technical Report developed under Air Force contract AFWAL-TR-81-4023, available at http://www.idef.com. [13] Knowledge Based Systems Inc. (KBSI), IDEF1 Information Modeling Report, Air Force Wright Aeronautical Laboratory Technical Report developed under Air Force contract AF33615-80-C-5155, available at http:// www.idef.com. [14] Knowledge Based Systems Inc. (KBSI), IDEF5 Ontology Description Capture Method Report, developed as part of Information Integration for Concurrent Engineering (IICE) project, Air Force contract AF33615-90-C-8005, available at http://www.idef.com. [15] Knowledge Based Systems Inc. (KBSI), IDEF3 Process Description Capture Method Report, developed as part of Information Integration for Concurrent Engineering (IICE) project, Air Force contract AF33615-90-C-0012, available at http://www.idef.com. [16] J.M. Colom, Applications and Theory of Petri Nets, SpringerVerlag, Berlin, 2002. [17] F.B. Vernadat, Enterprise Modeling and Integration, Chapman & Hall, London, 1996. [18] J. Zachman, A Framework for Information Systems Architectures, Report No. G320-2785, IBM, Los Angeles Scientic Center, 1986. [19] C.J. Petrie (Ed.), Proceedings of the First International Conference on Enterprise Integration Modeling, MIT Press, Cambridge, MA, 1991. [20] M.S. Fox, The TOVE Project: towards a common-sense model of the enterprise, in: Proceedings of the 5th International Conference on Industrial and Engineering Applications of Articial Intelligence (IEA/AIE), Paderborn, Germany, 1992, pp. 2534. [21] F. Fillion, C. Menzel, R.J. Mayer, T. Blinn, An ontologybased environment for enterprise model integration, in: Proceedings of the Workshop on Basic Ontological Issues in Knowledge Sharing at IJCAI95, Montreal, Canada, 1995, pp. 3345.

You might also like