چارچوبی برای انتقال از مدل های جریان کاری مفهومی به منطقی

Both conceptual and logical workflow models are needed to support business process automation via workflow systems. Conceptual models are normally used to document the generic business process requirements in the company. Logical models are generally used for defining technology specific requirements, where software modules as well as their behavioral patterns should be clearly specified. However, the transformation from conceptual models to logical models can be a tedious task, often causing errors in the resulting logical model. In this paper, we propose a formal approach that can be used to support efficient and accurate model transformation. First, we develop a procedure for transforming a conceptual workflow model into its corresponding logical workflow model. Business requirement analysis, dependency mapping, and workflow pattern-based model transformation are the major components of this transformation procedure. Second, we create a validation procedure that can validate whether the derived logical model is consistent with its original conceptual model. Business process ontologies are employed in our approach to describe both conceptual and logical models. We also implement a prototype system and conduct a demonstrative case study to show the feasibility of our approach.

Business process modeling and workflow technologies have become essential when developing enterprise information systems. There are various process modeling languages to describe processes [34]. While much attention has been paid to the logical correctness of these models [4] and [41], developing a workflow application that can fulfill given business requirements is also very important [16] and [38]. The terms “business process model” and “workflow model” are both found in the literature. The business process model is often used when communicating with managers, while the workflow model is commonly used at the system level. In this paper, we will use both terms interchangeably. This research is motivated by the need to resolve a real-world problem in the context of the Kuali project [21]. Kuali is a community source project to develop a comprehensive suite of administrative software that meets the needs of all Carnegie Class institutions. There are currently more than twenty development partners in the Kuali project. In this context, we need to develop a workflow model to support software change management based on a conceptual business process model. Further, we need to validate that the workflow model we develop is consistent with the given conceptual business process model. However, we could not find an existing approach for systematically transforming a conceptual business process model to a physical workflow model. Over the past twenty years, a lot of work has been done in the area of business process modeling. Much research has been devoted to model expressiveness [32] and [42], and some research has focused on business process model verification [4], [34], [41] and [46]. However, these approaches stop at logical correctness. Only a few approaches [16], [23] and [38] in the literature explicitly capture business requirements in the workflow design process even though doing so was suggested ten years ago [10]. Further, for formal verification, some workflow models are very difficult for managers to understand, which often results in a gap between managerial users and technical developers of workflow applications. For example, in order to add a new task to a Petri net-based workflow model, one must manipulate the model in terms of transitions, places, arcs and tokens, which can be done correctly and efficiently only by someone well-versed in Petri nets, a skill not normally possessed by ordinary managers. Designing a workflow model is a knowledge-intensive endeavor because creating a typical workflow model requires detailed understanding of various process components, such as business process logic, the organizational chart, and the information systems accessed by the workflow. The whole design process may require collaboration between an enterprise's functional and technical departments. More importantly, the model is subject to frequent modification due to changes in the process components. As has been done in the database field, dividing the design process into three phases, namely conceptual, logical, and physical design, should enhance the efficiency of modeling as well as the quality of the design output. A conceptual, logical, or physical business process model is the output of each design phase respectively. The conceptual business process model has a higher level of abstraction than the other two types of models. The transformation from conceptual business process model to logical business process model and then to physical business process model is very important in terms of mapping business requirements to system implementation. The terms of “conceptual model”, “logical model” and “physical model” are used to represent the three models in the rest of this paper. In this paper, we present a detailed transformation procedure from conceptual to logical models. We choose Dependency Network Diagrams (DND) [37] as the conceptual model because of its simplicity and expressiveness. Further, Petri nets are chosen as the sample logical modeling language because of the availability of abundant verification techniques [40]. Here, the conceptual model is mainly used to capture the business requirements in enterprises; the logical model is used for information system (e.g. workflow) design purposes; and the physical model is only used for system execution. In particular, the key challenges in this research are how to derive a logical model from a given conceptual model and how to validate that the derived logical model is consistent with the given conceptual model. To provide a flexible modeling framework, model designers can derive logical models iteratively based on conceptual models in the logical design process. That is, logical design of a workflow model can involve multiple logical models. The main contributions of this paper are three-fold. First, we propose a three-layer modeling approach that differentiates among conceptual, logical and physical models. Second, we develop a methodology for transforming a conceptual model into a logical model. Third, we create an approach for validating whether the derived logical model is consistent with its corresponding conceptual model. The rest of this paper is organized as follows. In Section 2, we review the related areas of this research. In Section 3, we define conceptual, logical and physical models in detail and we address relationships among the three types of models. Section 4 gives an example conceptual model in DND. Section 5 presents the transformation procedure for deriving a logical model from a conceptual model and introduces the validation procedure. Section 6 validates our approach by a case study and prototype system development. We conclude in Section 7 with a discussion of the results and limitations of our research

7.1. Discussions Important design science contributions create and evaluate IT artifacts intended to solve identified organizational problems [13]. In order to facilitate workflow model design, this research presented a formal approach to transformation from conceptual to logical workflow models. We first proposed a semi-automated procedure to add information to the conceptual model and transform it to a logical one. Further, the consistency between the conceptual model and its corresponding logical model is checked via an ontology-based approach. To the best of our knowledge, our study represents the first attempt to (a) formally define three layers of workflow models, (b) transform conceptual models to logical models, and (c) check consistency between conceptual and logical models. While this study has direct practical implications for workflow model designers, it may also have theoretical implications. Our research has theoretical implications for workflow modeling. We have shown that workflow modeling can be done successively from simple to complex in three steps in terms of information contents. At the conceptual level, we do not consider logical and physical constraints. By applying the theory of information hiding [25] from software engineering to model building, we can simplify the modeling tasks by focusing on the most important concepts without worrying about unnecessary details. In addition, changes to the workflow models can be started in any of the three layers first and then propagate the changes to other layers. Our model transformation approach also extends the theory of model-driven software development [33] to the workflow domain and requires further enhancement of the theory since the current IDEs do not explicitly distinguish the three levels of workflow models. These theoretical frameworks have guided our research efforts and help make our research results acceptable to other researchers and developers who might adopt our viewpoints. Our work is a step towards a theory of workflow models, similar to the theory of data models [3]. As a relatively young field in software engineering, workflow design in the industry has not adopted a standard process similar to what has been done in database design. This is unfortunate since the three-layer approach in database design has been widely adopted in the industry with great benefits. However, how to standardize the workflow design process remains an open question since it is not yet widely agreeable if workflow design should adopt a layered approach and how workflow design tasks should be layered as we suggested in this paper. We believe that we have laid the groundwork for extensive theoretical and empirical research into workflow model design. Some of the conjectures that can be derived from our research (e.g., the efficiency of layered workflow modeling approach, the relationship between different layers of workflow models, the effect of workflow modeling tools with transformation) call for further investigations. 7.2. Limitations and future work We identify two limitations of this study. First, we limited this study to popular process modeling patterns (e.g. five basic workflow patterns) and techniques (e.g. Petri nets and DND). We believe it is representative of the most popular techniques based on earlier studies. The smaller scope enables us to focus our work and to avoid too many extra constraints. Although the general model transformation and validation framework can be applied to other modeling languages, the detailed steps might need to be altered according to different modeling languages. Second, our approach is semi-automatic and relies heavily on user input of additional information. This could introduce potential errors and inaccuracies to the transformation process. More efforts may be needed to automatically extract additional information from other sources such as business policies and rules. In our future research, we intend to extend our work in two directions. First, while the correctness of the proposed model transformation approach is validated by a case study, user studies are needed to assess the effectiveness of such systems in practice. These studies must address a bevy of issues, including appropriate user interface design, methods for enhancing the perceived usefulness of the system, mechanisms for error alerting messages when interacting with users, etc. Second, we will explore the relationships between logical models and physical models that are system dependent and propose a design theory for layered workflow model design.