analysis of functional architectures for discrete event ... · 518 timothy sprock and leon f....

Procedia Computer Science 44 ( 2015 ) 517 – 526

Available online at www.sciencedirect.com

1877-0509 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).Peer-review under responsibility of the Stevens Institute of Technology. doi: 10.1016/j.procs.2015.03.022

ScienceDirect

2015 Conference on Systems Engineering Research

Analysis of Functional Architectures for Discrete Event Logistics Systems (DELS)

Timothy Sprocka*, Leon F. McGinnisa H. Milton Stewart School of Industrial and Systems Engineering, The Georgia Institute of Technology, Atlanta, GA

Abstract

The functional design of discrete event logistics systems (DELS) specifies the systems’ transformative functions independently from the technology required to implement those functions. Creating a functional architecture for a new system is as much an art form as a science and often relies on the tacit knowledge of the systems architect. In this paper, we present a method to integrate analyses to support the refinement of the functional architecture. This method relies on integrating the description of the functional architecture with a formal network description and using that formal network description to provide access to analysis generation technologies. The result is automated access to analysis tools, such as discrete event simulations, which provide critical feedback on the performance of a particular design early in the design process. © 2015 The Authors. Published by Elsevier B.V. Peer-review under responsibility of the scientific committee of Stevens Institute of Technology.

Keywords: functional architecture; discrete event logistics systems (DELS); SysML

1. Introduction

Discrete event logistics systems (DELS) are a class of dynamic systems that are defined by the transformation of discrete flows through a network of interconnected subsystems1. The DELS domain includes systems such as supply chains, manufacturing systems, transportation networks, warehouses, health care delivery systems, etc. DELS are inherently complex systems due to the large scale of the networks, the dynamic nature of interactions between actors, and the randomness of both the external and internal environments. This makes any decision making process difficult and implies the need for a wide range of modeling and analysis methodologies and tools to guide decision making.

* Corresponding author. Tel.: +1-856-371-4343.

E-mail address: [email protected]

© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).Peer-review under responsibility of the Stevens Institute of Technology.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.procs.2015.03.022&domain=pdf

http://crossmark.crossref.org/dialog/?doi=10.1016/j.procs.2015.03.022&domain=pdf

518 Timothy Sprock and Leon F. McGinnis / Procedia Computer Science 44 ( 2015 ) 517 – 526

Industrial engineering and operations research offer many tools and methods for analyzing DELS, but widely accepted generic design methods—and in particular, tool chains—for designing DELS have remained elusive. While methodologies for designing other complex systems can be adapted to the DELS domain, the supporting tools have been primarily focused on designing the physical architecture, e.g. resource selection, policy configuration, and layout. Therefore, there remains an explicit gap in the literature of designing and implementing useful analyses to support the earlier stages of design, specifically the functional design of DELS.

This paper addresses the problem of translating design methods into tools chains to support the design of DELS. Specifically, we focus on applying current research on formal domain modeling for DELS and creating model-to-model (M2M) transformations to provide designers with automated access to design-specific analysis tools. The new technical idea proposed here is the DELS functional requirements network (DFRN), which is developed by augmenting functional architectures with network semantics, and then integrating network flow based analytics to support the design of these architectures.

2. Background

Material handling systems, such as warehouses, have very few design constraints, which makes them difficult to design. Therefore, they are good candidates for exploring and demonstrating design methodologies and provide the examples threaded throughout this research. McGinnis2 explores the fundamental changes that Model Driven Architecture® (MDA)3 methodologies would bring to the engineering of material handling systems and discrete event logistics systems (DELS) in general. This section introduces an MBSE design and analysis methodology for DELS. First, we introduce a more general design methodology from the systems engineering domain, and then present current research on the design of DELS. This design process is supported by formal domain modeling using SysML®, which provides a pathway to automate the generation of analyses relevant to the design process, including discrete event simulation models.

2.1. Design Methodologies and the Functional Architecture Specification (FAS) Method

In the systems engineering domain, design methodologies4 help engineers cope with the complexity of systems through several lifecycle development models that include high-level design stages such as analysis, specification, design and verification. These high level design stages are further refined into stages that are more manageable for the systems engineer. For example in OOSEM5, the ‘Specify and Design System’ process is further refined into ‘Analyze Stakeholder Needs’, ‘Analyze System Requirements’, ‘Define Logical Architecture’, and ‘Synthesize Candidate Physical Architectures’. The functional design process focuses on systematically deriving logical architectures from use cases and requirements. This process is supported by the functional architecture for systems (FAS) method6, which establishes a process for deriving the functional (logical) architecture from use cases. The FAS method uses heuristics to group functions, which are then allocated to functional blocks, the subsystems of the target system.

2.2. Design of Discrete Event Logistics Systems

In recent years, DELS themselves have become more integrated and more distributed at the same time. However, techniques and tools to design these increasingly complex systems have not kept pace. Traditionally, DELS design methodologies7,8 specify the functional architecture implicitly and instead focus on the physical architecture that should be utilized to accomplish a specific set of transformations (manufacturing, assembly, distribution, etc.). In the DELS domain, McGinnis et al.9 propose a systems engineering methodology for designing warehouses, with a focus on introducing and developing functional architectures for the domain. In research on reverse production systems, which at the time was a new domain that contained little expert knowledge on optimal design, Realff et al.10 focus on specifying the functional design and allocating the functions to geographic locations (systems). Govil and Proth11 present a framework for designing supply chains at the strategic level which produces a coarse functional architecture from a minimal set of supply chain activities. The collection of activities presented at this level parallel those contained in the SCOR model12.


2.3. Formal Domain Modeling with SysML

Formal domain modeling has been used in software engineering for designing and developing systems, and relies on domain specific languages (DSLs) and visualization to enable a systems engineer to focus on abstract modeling of the target domain. SysML13 has proven to be a powerful modeling language for system applications in a diverse range of engineering domains, such as electrical, mechanical, and industrial14–16. To tailor the modeling language to a specific domain, SysML provides two mechanisms for creating DSLs: model libraries and profiling17. Prior research has established the application of this methodology for creating a DSL in SysML, describing the system of interest using the DSL, and finally transforming a description of a system model specified using the DSL into a target analysis language, such as discrete event simulation18–21.

2.4. Automated Generation of DES and Relevant Analyses

Due to the complexity of DELS, mathematical models are often intractable or rely on a simplified model of the “real-world” behavior of the target systems. When model fidelity is important, discrete event simulation (DES) is often the tool of choice for modeling and analyzing DELS. Traditionally, automated generation of DELS analysis models has been approached in an ad-hoc manner dependent on the analysis type and target language. Several improvements have been made on the ad-hoc process including developing custom tools that model the supply chain infrastructure with a library of carefully designed objects with well-defined interactions22 providing a mapping between well-defined analysis objects and their corresponding schema23, and generic and automatic simulation generation techniques 24.

In research motivated by MDA’s approach to automated code generation, the model-to-model transformation paradigm from software engineering has been adapted to support the generation of DES analyses25. This approach relies on a formal network definition to provide a bridging abstraction between the system model and analysis models, which is a key requirement to integrating analysis tools to support the design of logical architectures. The transformation is executed in two steps from SysML to an IDE (MATLAB®) and then from the IDE to analysis tools, such as discrete event simulation in SimEvents or optimization with CPLEX®.

3. Functional Requirements Networks

During the ‘Analyze System Requirements’ specification phase, the system context is specified and consists of the interfaces to and interactions with the external ecosystem. In this specification phase of designing a DELS, the inbound flows (raw materials, orders, etc.) and outbound flows (finished goods, shipments, etc.) are captured independently of the transformation process that takes place in between. However, the exact transformation process lacks a single universal specification, and therefore the functional design stage is critical to properly specifying the exact nature of this transformation process. Previous research by McGinnis et al.9 and Govindaraj et al.26 has proposed capturing the structure of the transformation process as a network, a functional requirements network (FRN). In this section, we formalize the definition of a DELS functional requirements network (DFRN) and integrate this definition with existing research on designing functional architectures. A warehouse design use case is threaded through this section as an example of usage.

A formal definition of the DFRN is captured and presented as a reference architecture, which consists of three components specified using SysML: 1) a profile for token flow networks, a subclass of the classical network definition; 2) a definition of the node and edge components of the functional requirements network; and 3) a pattern for the specification of flow throughout the DFRN.


The token flow network (TFN) was created to serve two purposes: (1) to capture the essential nature of resources and flows in DELS; and (2) to be a unifying definition of the networks that are woven into DELS analysis models27. Using the UML profiling mechanism, the TFN expands the expressiveness of the typical network definition by elaborating the capabilities and responsibilities provided by nodes/edges and flow nodes/edges (Figure 1). Following from the SysML semantics, the edges between nodes represent relationships or associations, such as service contracts between a client and server. The flow node defines what is consumed and produced by a node, or a transformation process in a functional architecture, and the flow edge provides semantics for what, orders, parts, etc.; is flowing between flow nodes. The edge and flow edge are intended to be applied at the BDD and IBD modeling level respectively.

The second component of the DFRN definition is establishing the DELS domain specific functions (bottom half of Figure 2). In addition to the domain specific functions, we have integrated the DFRN with the information model presented for the FAS method6 and applied the TFN definition to this model. The TFN provides a layer of network semantics that clarifies the separation between the logical and physical architectures (flow nodes and nodes respectively). Moreover, it refines the semantics of flow between functions and how they relate to the associations between functional elements.

The last component of the reference architecture for the DFRN is an implementation pattern for creating a DFRN (Figure 3). This pattern captures several principles that are important to the specification of the DFRN and that support the warehouse design process. Throughout the design process, the DFRN is constantly altered and refined by adding, removing, and combining functional nodes. It provides a pattern for creating the flows that are incident to each function.

Figure 1 Basic Structure of the Token Flow Network24


The network abstraction is common throughout the DELS domain and provides access to a well-understood analysis framework and the associated tools. A formal definition of the network abstraction, such as the token flow network27, enables automation. If the abstraction is stable and broadly applicable, a model transformation engine based on that abstraction can be written once and reused across several applications and domains. This transformation engine25 is a key tool for integrating desired analyses into the design of DFRNs.

4. Analysis of Functional Requirements Networks

Throughout the functional design process, the systems architect incorporates feedback on how the system should perform to refine the functional architecture. Some methodologies rely on tacit knowledge, i.e. what has worked before, or heuristics6; e.g. the density of calls between the functions suggest good groupings, a heuristic supported by node centrality measures from graph theory27. One goal of this research is to integrate appropriate analytics to support the functional design stage: If automated access to analysis tools was available on demand, what analyses would be useful in the design process? How would you incorporate feedback from these analyses into the design process?

Figure 2 Functional Architecture Extended to Support Warehouse Functional Design

Figure 3 Implementation Pattern for Warehouse Functional Requirements Network


One unique feature of DELS is the physical flow of items through the system, which require physical resources to

transform, transport, and store. In the early stages of the design process, the capability and capacity requirements may only be derived from an analysis of the physical flow through the system. In these systems, the characteristics of the flow will factor heavily into how the systems are organized and which resources are selected; for example, moving a pallet requires a forklift, while an aircraft assembly may require an overhead crane. Therefore, a fundamental analysis for the DFRN would be to determine the capacity requirement of each functional node by computing the flows through the node (Table 1).

While this analysis is trivial to do programmatically or with simulation, it is difficult to propagate the flow types

through a general form network, a fact which prevents closed form solutions and easily manipulated spreadsheet analytics. These characteristics make it a good candidate for automation, which we implement by constructing a DES and capturing flow statistics. Furthermore, the feedback can be integrated into the design process by using the capacity requirements to determine a range of feasible technologies28. The feedback is useful for eliminating or reconsidering infeasible or undesirable solutions when the rework costs of preparing a new design are low. For example, an initial plan (Figure 4) may suggest using an automated storage and retrieval system (AS/RS) for every carton SKU in a warehouse. After determining that the estimated capital cost would exceed the specified requirement, Figure 5 shows the new design that divides the storage area into separate AS/RS and picker to goods regions based on flow characteristics such as SKU weight and pick frequency.

Figure 4 Initial Functional Requirements Network Specification

Table 1 Output Results of Flow Analysis

INFlow/OUTFlow of each UoH (Column Header) to/from each flow node (Row Header)


The second layer of analysis that our research has incorporated into the design process is based on process networks

and black-box analyses derived from queueing theory approximations29. Process networks are a natural formalism for modeling the sequence of conversions that occur to manufacture a product (product-centric view) or within a DELS (system-centric view). To create a formal model for the analysis domain, the process network is modeled as an extension of the TFN and incorporates attributes and methods derived from queueing theory (Figure 6).

Using the analysis generation method based on the network abstraction25, we can evaluate critical questions about the system design: e.g. “what is the capacity requirement of this system if this collection of functions is allocated to it?” or “given estimates of throughput capabilities, how many pickers does this system need to remain stable?”. As more refined information becomes available, the transformation engine can generate a DES of the process network to model more complex behaviors. Since the analyses are generated automatically, the system design and their supporting analyses can be continuously updated to support decision making in other disciplines, such as product design. This methodology allows the system engineer to incorporate crude information into early lifecycle decisions, such as automated vs manual technologies28, and refine both the functional and physical design quickly as more information

Figure 5 Incorporating Refinements to the Flow Specification into the FRN

Figure 6 Process Network Augmented with Factory Physics Analysis Attributes


becomes available. It is intended to support the transition from functional design to embodiment design by helping to allocate functions to systems and size those systems.

5. Future Work: A Complete Design Methodology and Supporting Toolchain

A complete methodology and a supporting toolchain for designing systems in the DELS domain have remained elusive despite significant progress on the individual analysis methodologies. Building on the results presented here, this research can be immediately extended to support the elaboration of a functional design into the specification of the physical artifact itself. As functional nodes are allocated to systems and resources and policies are selected and configured (Figure 7), it is useful to extend the methodology to continuously provide feedback on system performance. This requires integrating additional analysis methodologies and tools into the design process, which might be as simple cycle time or max throughput analyses, but should provide a basis for building a final detailed simulation to configure the physical architecture.

However, one of the most significant challenges to realizing a complete design methodology is lack of canonical models to describe DELS. In the status quo, design efforts and analysis tools rely on implicit models of the system based on information models and existing analysis models. A canonical reference architecture of this domain is critical to integrating design methodologies with an entire suite of tools to support each stage of the design process, from statistical tools to support functional requirements analysis to optimization and simulation tools to support laying out and configuring the physical system. With relevant analyses accessible throughout the design process, the methodology itself can be refined iteratively to incorporate new analysis results to produce incremental improvements to the system design (Figure 8).

Figure 7 Candidate Design Incorporating AS/RS and Material Handling Channels (MHC)


6. Conclusion

The focus of this research is an MBSE approach to functional design that separates the design and analysis of the system allowing the supply chain engineer to focus on designing the system (and incorporating feedback through on-demand analyses) rather than focusing on the implementation of the analyses themselves. In the context of functional design, there are few system details to incorporate into high-fidelity simulations or complex optimization models. Instead, we rely on intuitive analyses, such as flow analysis, or simple black-box analyses derived from the queuing networks framework. The functional requirements network and associated transformation capability provides a method to integrate analyses into the design of functional architectures and creates significant opportunity to evaluate and incorporate useful analyses to answer specific design questions.

References

1. Monch L, Lendermann P, McGinnis LF, Schirrmann A. Special Issue on Grand Challenges for Discrete Event Logistics Systems. Computers in. 2011;62(6):555–6.

2. McGinnis L. The Future of Modeling in Material Handling Systems. 11th International Material Handling Research Colloquium - 2010 Material Handling Industry of America [Internet]. 2010. Available from: http://www.mhi.org/downloads/learning/cicmhe/colloquium/2010/mcginnis.pdf

3. OMG MDA. OMG Model Driven Architecture (OMG MDA) Version 1.0.1 [Internet]. Object Management Group; 2003 [cited 2013 Feb 23]. Available from: http://www.omg.org/mda/

4. Estefan JA. Survey of Model-Based Systems Engineering (MBSE) Methodologies. Pasadena, California: Jet Propulsion Laboratory; 2007.

5. Friedenthal S, Moore A, Steiner R. A Practical Guide to SysML: The Systems Modeling Language. ELSEVIER Inc.; 2009. 6. Lamm JG, Weilkiens T. Method for Deriving Functional Architectures from Use Cases. Systems Engineering. 2014;17(2):225–36. 7. Beamon BM. Supply Chain Design and Analysis: Models and Methods. International Journal of Production Economics. 1998;55(3):281–

94. 8. Min H, Zhou G. Supply chain modeling: past, present and future. Computers & Industrial Engineering. 2002;43(1):231–49. 9. McGinnis L, Goetschalckx M, Sharp G. A Comprehensive Model of Traditional Warehouse Design. 9th International Material Handling

Research Colloquium—2006 Material Handling Industry of America. 2006. 10. Realff MJ, Ammons JC, Newton D. Strategic design of reverse production systems. Computers & Chemical Engineering.

2000;24(2):991–6. 11. Govil M, Proth J-M. Supply Chain Design and Management: strategic and tactical perspectives. Academic Press; 2002. 12. anon. SCOR: The Supply Chain Reference Model Version 11.0 [Internet]. Supply Chain Council, Inc; 2012. Available from:

www.supply-chain.org 13. OMG SysML. OMG Systems Modeling Language (OMG SysML) Version 1.3 [Internet]. Object Management Group; 2012 [cited 2013

Feb 23]. Available from: http://www.omg.org/spec/SysML/1.3/ 14. Johnson T, Kerzhner A, Paredis CJJ, Burkhart R. Integrating Models and Simulations of Continuous Dynamics into SysML. Journal of

Computing and Information Science in Engineering. 2012;12(1):11002. 15. Peak RS, Fulton RE, Nishigaki I, Okamoto N. Integrating Engineering Design and Analysis Using a Multi-representation Approach.

Engineering with Computers. 1998;14(2):93–114. 16. Shah AA. Combining Mathematical Programming and SysML for Component Sizing as Applied to Hydraulic Systems [Master of

Figure 8 A Warehouse Design Methodology and Accompanying Toolchain


Science - Mechanical Engineering]. [Atlanta, GA]: Georgia Institute of Technology; 2010. 17. Selic B. A Systematic Approach to Domain-specific Language Design using UML. Proceedings of the 10th IEEE International

Symposium on Object and Component-Oriented Real-Time Distributed Computing (ISORC’07). IEEE; 2007. p. 2–9. 18. Batarseh O, McGinnis LF. SysML to Discrete-event Simulation to Analyze Electronic Assembly Systems. Proceedings of the 2012

Symposium on Theory of Modeling and Simulation-DEVS Integrative M&S Symposium. 2012. p. 48. 19. Batarseh O, McGinnis LF. System Modeling in SysML and System Analysis in Arena. Proceedings of the 2012 Winter Simulation

Conference. 2012. p. 258. 20. McGinnis L, Huang E, Kwon KS, Ustun V. Ontologies and simulation: a practical approach. Journal of Simulation. 2011;5(3):190–201. 21. McGinnis L, Ustun V. A simple example of SysML-driven simulation. Proceedings of the 2009 Winter Simulation Conference (WSC).

IEEE; 2009. p. 1703–10. 22. Chatfield DC, Harrison TP, Hayya JC. SISCO: An object-oriented supply chain simulation system. Decision Support Systems. 2006

Oct;42(1):422–34. 23. Biswas S, Narahari Y. Object oriented modeling and decision support for supply chains. European Journal of Operational Research. 2004

Mar;153(3):704–26. 24. Cope D, Fayez MS, Mollaghasemi M, Kaylani A. Supply Chain Simulation Modeling Made Easy: An Innovative Approach. Proceedings

of the 2007 Winter Simulation Conference. 2007. p. 1887–96. 25. Sprock T, McGinnis LF. Simulation Model Generation of Discrete Event Logistics Systems (DELS) Using Software Patterns.

Proceedings of the 2014 Winter Simulation Conference. Savannah, GA; 2014. 26. Govindaraj T, Blanco EE, Bodner DA, Goetschalckx M, McGinnis LF, Sharp GP. Design of warehousing and distribution systems: an

object model of facilities, functions and information. 2000 IEEE International Conference on Systems, Man, and Cybernetics. IEEE; 2000. p. 1099–104.

27. Thiers G. A Model-Based Systems Engineering Methodology to Make Engineering Analysis of Discrete-Event Logistics Systems More Cost-Accessible. [Atlanta, GA]: Georgia Institute of Technology; 2014.

28. Apple JM, Meller RD, White JA. Empirically-based Warehouse Design: Can Academics Accept Such An Approach? 11th International Material Handling Research Colloquium - 2010 Material Handling Industry of America [Internet]. 2010 [cited 2014 Feb 13]. Available from: http://www.mhi.org/downloads/learning/cicmhe/colloquium/2010/apple.pdf

29. Cassandras CG, Lafortune S. Introduction to discrete event systems. Springer; 2008.

analysis of functional architectures for discrete event ... · 518 timothy sprock and leon f....

Documents