On building multidimensional workflow models for complex systems modelling

From well-known, classical workflows such as Petri nets to one of the recent developments of modelling frameworks such as the Business Process Modelling Notation, the development of system representations has long been established and improved through the years. The common goal of such frameworks is to produce traceable, effective, and well-understood functional and nonfunctional specifications of business and scientific systems. This goal addresses issues from workflow design, verification, control and monitoring, and continual improvement as systems evolve. Through the years, these frameworks had been constructed, enabled, deployed and used under a singular or dual perspective of modelling and verification relating to workflow dimensions, i.e. process, resource, case. In literature, there is a huge gap for the support and enactment of all three dimensions into one model for system representation and verification. That is, these undertakings are mainly either process- or resource-centric. In terms of modelling with all three dimensions in place, some support is observed in the Robustness Diagram(RD) of the ICONIX framework. Because of its notational backbone, it was posed to serve as a bridge for requirements traceability when using other workflows that focus solely on either structure or behavior of system representations. It has potential in providing support from requirements capture to testing to redesign of models. However, this diagram has been underdeveloped with respect to these potentials in modelling and verification especially for complex systems. In this research, the Robustness Diagram with Loop and Time Controls(RDLT) was introduced to support modelling and verification of complex systems. It is an extension based from RDs. Building on RDs, we propose formalizations of RDLTs with consideration for the use and representation of all three workflow dimensions in one model. Additionally, by accounting the requirements of volatility, persistence, multi-state configuration, and hierarchical structures and relationships present in complex systems, the concept of a reset-bound subsystem(RBS) in RDLTs was also formulated. RBS enacts capabilities of cancellation regions in well-known workflows in literature. However, RBS enforces topological and behavioral requirements to perform resets in values in models. Additionally, this research addressed the problems of explicitness and effectiveness of representing data, control flow patterns(e.g. sequential, parallel, splits, n-out-of-k joins, iteration, cancellation regions, etc.), and multi-level and multi-participant interactions. These problems were also dealt under the specifications of all three workflow dimensions, persistent and volatile structures and behavior in models. In particular, this research introduced attribute-driven typing of vertices, arcs, and substructures to enforce structural and functional specifications in RDLTs. The values of the attributes and the resulting types of RDLT components directly influence their usage and groupings for the execution of activities in models. Furthermore, they also influence mechanisms for encapsulation of data and control flows in RDLTs. We proposed the concept of Points-of-Interests(POIs) in RDLTs that also rely on these information. POIs are then used to establish special regions in the model. We formally defined these regions and characterize their neighborhood structures. We established encapsulation rules for RBS and add its information to the characterized neighborhoods to determine metrics for the analysis of RDLTs. In this research, the metrics that were developed mainly focus on topological- and type-bound reachability, delays that cause bottlenecks, task synchronicity, and activity completion. They are computed with consideration of the presence of maximal substructures in RDLTs. Each substructure supports the execution of an activity profile for the completion of a case in a multi-activity RDLTs. Among the model properties in literature for workflows, this research adopted and introduced soundness and free-choice for RDLTs. (We focus on these two properties because many other properties can be implied from them.) Noting the types of components and substructures and the presence of RBS in them, a behavioral profile of RDLTs that satisfy the first property is initially provided. This profile is produced by the use of the results from our proposed algorithm for activity extraction. However, this research devised two independent approaches to prove this property through structural views that account the types and the presence of RBS in RDLTs. The first approach was constraint-driven. RLDT attributes that pertain to the enforcement of splits and joins of type-alike components were used. We developed the concepts of an extended RDLT and a vertex-simplified RDLT to support this approach. Meanwhile, the second approach was component use-driven. Attributes that pertain to repeatability of task execution were checked in type-alike/mixed-type components and hierarchical structures and interactions in RDLTs. This approach used our proposed encapsulation rules for RBS. Collectively, both approaches proved soundness based on statically-verifiable information in RDLTs. Meanwhile, the free-choice property for RDLTs was proposed and built from using thePOI-driven metrics for reachability, boundedness, and synchronicity. In literature, this property focused on place-task relationships in workflows. In this research, it is extended by including combinations of topology-, constraint-, and type-driven free-choiceness in models. There are two RDLT configurations that were proposed for these combinations. One adopts the well-known view of free-choiceness while the latter its extension. However, both configurations consider all the workflow dimensions in modelling RDLTs. In addition, this research provides efficient verification schemes for soundness, freechoiceness, and other proposed model properties for RDLTs. Real-world examples of RDLT models, including that of a complex energy system, was provided to illustrate how their structural and behavioral profiles were captured where these properties are checked. Finally, this research proved the relationships and hierarchies of RDLTs based on the properties that can be verified from them.