A multi-methodological collaborative simulation for inter-organizational supply chain networks

Accepted Manuscript

A multi-methodological collaborative simulation for inter-organizational supply chain networks Qingqi Long PII: DOI: Reference:

S0950-7051(16)00002-2 10.1016/j.knosys.2015.12.026 KNOSYS 3372

To appear in:

Knowledge-Based Systems

Received date: Revised date: Accepted date:

3 June 2015 30 December 2015 31 December 2015

Please cite this article as: Qingqi Long , A multi-methodological collaborative simulation for inter-organizational supply chain networks, Knowledge-Based Systems (2016), doi: 10.1016/j.knosys.2015.12.026

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

A multi-methodological collaborative simulation for inter-organizational supply chain networks Qingqi Long* School of Information, Zhejiang University of Finance & Economics, Hangzhou, Zhejiang 310018, China

Abstract:

CR IP T

Inter-organizational collaborative simulation requires covering the knowledge of agent, flow and process to qualifiedly represent the supply chain network operation. This paper proposes a multi-methodological collaborative simulation framework for inter-organizational supply chain networks. This framework integrates the agent-based, flow-centric and process-oriented methodologies. This framework establishes a collaborative knowledge representation approach for simulation modeling. In the approach, a multi-agent system is adopted to represent the inter-organizational structure of a supply chain network; the three flows of material, information and time are enabled to

AN US

represent the operational mechanisms; and the processes are used to represent the micro behaviors of agents. This approach integrates the knowledge of agent and process with that of flow and solves the problems regarding the integration of the agent-based, flow-centric and process-oriented methodologies. To implement the inter-organizational collaborative simulation, a collaborative framework is proposed. This framework integrates multiple simulation formalisms, such as time series increments, event scheduling, policy control, process interaction and activity scanning; it promotes the unification of the

M

three different methodologies. A case of a five-level manufacturing supply chain network is studied using the proposed framework. The findings indicate that the proposed framework is particularly qualified in the knowledge representation of a supply chain network and is effective in implementing

ED

the inter-organizational collaborative simulation in a decentralized manner; in addition, it is well contributive to collaborative decision making through KPI analysis.

PT

Key words: Collaborative simulation; Inter-organizational supply chain network; Multi-agent system;

CE

Multi-dimensional flows; Supply chain network process

1. Introduction

A supply chain network is a complex adaptive system composed of several enterprises with a

AC

certain structure. The network is an integrated process wherein raw materials are manufactured into final products, then delivered to customers [3]; it is no longer a single chain but a network intertwined with a few chains [9, 18, 26, 40]. Competition and cooperation coexist among enterprises in the network. The networks should cross their boundaries to implement inter-organizational collaboration through such activities as planning, production, inventory and delivery to strengthen the competitiveness of the network [6]. The inter-organizational supply chain network has the following characteristics: dynamic reconfigurable structure, complex collaborative processes and adaptive collaborative decision making. These characteristics are doomed to pose challenges to both its theoretical research and applications. Particularly, it is a great challenge to connect decision making *Corresponding author. E-mail address: [email protected] (Q. Long). 1

ACCEPTED MANUSCRIPT

activities of enterprises into a unified systematic framework. Simulation is an effective methodology for complex systems; it supports the emergence analysis from micro activities to macro phenomena. In addition to such methodologies as operational research and control theory [43], it is another effective tool for quantitative research on the supply chain network. In view of the inter-organizational characteristics, enterprises in decentralized environments should be included in a unified collaborative simulation framework. Inter-organizational collaborative simulation is a framework that models the structure of a supply chain network and its micro inter-organizational operational mechanisms. This collaboration establishes the relation between the enterprises’ micro activities and the macro network phenomena and supports the what-if analysis for

CR IP T

the design and optimization of a supply chain network by means of a variety of virtual reproductions.

Inter-organizational collaborative simulation solves difficulties in modeling the complex collaborative processes and matching between macro phenomena and micro activities. An increasing volume of literature concerns inter-organizational collaborative simulation for supply chain networks. This literature generally adopts two methodologies: process-oriented simulation methodology and agent-based simulation methodology. A process-oriented simulation uses processes as the basic

AN US

modeling units and simulates the process sequence (process flow) in a centralized manner, for example, the studies of Jakkhupan, Arch-int and Li [13]; Longo and Mirabelli [24]; Windisch et al. [37]; Mohammadi, Mukhtar and Peikari [27]; and Windisch et al. [38]. These studies exploit the intrinsic attribute, a process, in a supply chain network operation, aiming to represent the entire operation by means of micro processes. However, these studies are weak in supporting the representation of the enterprises’ decision making capabilities, partial information sharing context and geographical decentralized environment. This weakness results in a discount of the simulation fidelity and

M

effectiveness. A multi-agent system promotes the studies on collaborative simulation for supply chain networks. In contrast to the process-oriented methodology, agent-based simulation utilizes a complex system research perspective. On the one hand, it uses an agent as the basic modeling unit to represent

ED

the micro entities and their interactions; on the other hand, it uses a multi-agent system to simulate the supply chain network operation for the emergence of macro phenomena. A large volume of literature focuses on this topic, for example, the studies of Santa-Eulalia, D’Amours and Frayret [32]; Labarthe

PT

et al. [15]; Bahroun et al. [2]; Li, Sheng and Liu [16]; and Long, Lin and Sun [22]. Considering the agent attributes of enterprises, these studies take advantage of distributed computing capability of

CE

multi-agent system to describe geographical distribution and support collaborative simulation under the context of partial information sharing. These are consistent with real supply chain network, with higher credibility and utility in the conclusions. However, agent-based simulation has several shortcomings. It

AC

is weak in representing the information flow [5], material flow and time flow [19, 21]. The combination of process-oriented and agent-based methodologies can contribute to an effective solution [5, 11, 20]. Although this combination has obvious advantages, defects remain. First, there is no framework that is established for inter-organizational collaborative simulation of a supply chain network; in particular, no knowledge representation regarding its inter-organizational collaboration is involved. Second, the knowledge related to the agent and the process in the supply chain network has been described in current studies. However, these studies failed to integrate the knowledge of the agent and process with that of the flow and address the contradiction of the agent-based, flow-centric and process-oriented methodologies in their integration. Third, these studies ignored the abovementioned knowledge in their frameworks to achieve a more realistic, effective and credible simulation to support effective decision making. All of these challenges provide great opportunities and motivations for the 2

ACCEPTED MANUSCRIPT

research in this paper. This paper proposes

a multi-methodological

collaborative

simulation framework for

inter-organizational supply chain networks. This framework develops a knowledge representation of inter-organizational supply chain network and its formal description for collaborative simulation. In the framework, a multi-agent system is adopted to develop the inter-organizational network structure; the three-dimensional flows represent the operational mechanisms; and the processes instantiate agents’ behaviors. Integration of a multi-agent system, the three-dimensional flows and operational processes helps to support the effective emergence of supply chain network. To implement the inter-organizational collaborative simulation, a collaborative framework is proposed. This framework

CR IP T

integrates multiple simulation formalisms, such as time series increments, event scheduling, policy control, process interaction and activity scanning; it promotes the unification of the agent-based, flow-centric and process-oriented methodologies. A case of a five-level manufacturing supply chain network is studied using the proposed framework. The findings indicate that the proposed framework is particularly qualified in the knowledge representation of supply chain network and is effective in implementing the inter-organizational collaborative simulation in a decentralized manner; in addition, it

AN US

is well contributive to collaborative decision making through KPI analysis.

The remainder of this paper is organized as follows: Section 2 presents a series of related work. Section 3 proposes a multi-methodological collaborative simulation framework for inter-organizational supply chain networks. Section 4 puts forth a multi-formalism collaborative framework for simulation implementation. Section 5 conducts a case study to verify the proposed collaborative simulation

M

framework. Section 6 concludes.

2. Related work

According to the different modeling methodologies, simulation for a supply chain network can be

ED

divided into two categories: process-oriented simulation and agent-based simulation. Process-oriented simulation uses micro processes as the basic modeling units; thus, it has obvious advantages in the abstract description of a supply chain network operation in simulation modeling.

PT

Tako and Robinson [33] studied the application of discrete event simulation and system dynamics in the logistics processes and in a supply chain context. Jakkhupan, Arch-int and Li [13] studied business

CE

process analysis and simulation for the RFID and EPCglobal Network-enabled supply chain. Cannella et al. [4] presented a simulation-based study of a coordinated, decentralized linear supply chain system. Longo and Mirabelli [24] proposed an advanced modeling approach and a simulation model to support

AC

supply chain management using a process-based simulation package, the eM-Plant. Karagiannaki, Doukidis and Pramatari [14] proposed a framework for mapping the RFID-enabled supply chain process redesign in a simulation model. Windisch et al. [37] applied a methodological framework to investigate two supply chains in different operational environments to identify the business processes and stakeholders that comprise the supply chains, using a business process mapping methodology. Mohammadi, Mukhtar and Peikari [27] proposed a methodology combining two approaches (i.e., grammar-based business process modeling and simulation) to facilitate process thinking. Reiner [31] described how process improvements can be dynamically evaluated under consideration of customer orientation and supported by an integrated usage of discrete-event simulations models and system dynamics models. Fröhling et al. [10] developed closed-loop supply chain planning systems by elaborating and implementing an operational planning approach for integrated planning of 3

ACCEPTED MANUSCRIPT

transportation and recycling for multiple plants based on process simulation. Windisch et al. [38] investigated an information-based raw material allocation process for increasing the efficiency of an energy wood supply chain using discrete-event simulation. Groznik and Maslaric [12] conducted a case study to investigate the impact of information sharing in a two-level supply chain using business process modeling and simulations. Wang et al. [36] evaluated the value of collaboration in a supply chain through business process simulation. With the power of business process simulation, the researchers have the capability to evaluate the benefits of collaboration considering the details of the operations and the stochastic characters. Persson [29] developed a simulation tool for a supply chain simulation based on the second version of the SCOR template that contains all major processes.

CR IP T

In contrast to a process-oriented simulation, agent-based simulation uses an agent as the basic unit to model the entities in a supply chain network. This simulation aims at the emergence of macro performance by means of agents’ decision making and behaviors as well as their interactions in multi-dimensional flows. This methodology has advantages in observing the macro phenomena of supply chain network that emerges from its micro activities.

Santa-Eulalia, D’Amours and Frayret [32] presented a novel methodological framework called

AN US

FAMASS (FORAC Architecture for Modelling Agent-based Simulation for Supply chain planning), which provides a uniform representation of distributed advanced supply chain planning and scheduling systems using agent technology. Manataki, Chen-Burger and Rovatsos [25] presented a multi-agent-based framework for simulating supply chain operation and re-configuration. Labarthe et al. [15] proposed a methodological framework for agent-based modeling and the simulation of supply chains in a mass customization context to facilitate their management. Bahroun et al. [2] proposed, based on multi-agent systems, a generic software agent model to model supply chains to simulate and

M

evaluate replenishment policies within these chains. Amini et al. [1] applied agent-based modeling and simulation methodology to analyze the impact of alternative production–sales policies on the diffusion of a new generic product and the generated NPV of profit. Chinh et al. [7] presented an agent-based

ED

simulation to model supply chains and evaluate the bullwhip effect under the stochastic demand and lead time. Ding et al. [8] developed a multi-agent simulation model applying a contract net protocol to analyze the cooperation and competition in flexible supply chain networks. Tan, Chai and Liu [34]

PT

proposed a message-driving formalism for the simulation of multi-agent supply chain systems to enhance the computational efficiency and retain the simulation credibility. Ni and Wang [28] developed

CE

an agent-based collaborative production system based on a real-time order-driven approach and simulated it using a Java Agent Development Frameworks platform. Li, Sheng and Liu [16] described a multi-agent simulation model to analyze the dominant player’s behavior of supply chains. Zhao, Qiu

AC

and Zhang [42] presented an agent behavior based simulation model of a pricing coordination process. Wu et al. [39] investigated retail stockouts through the development of an agent-based simulation model to develop a better understanding of the effect of different stockout lengths for different products on both the retailer and the manufacturer of the product. Zhao and Qiu [41] studied agent-based simulation for an order selection strategy in a supply chain collaboration process. Lin and Long [17] and Long, Lin and Sun [22] developed a multi-agent-based simulation system for a supply chain network emergence analysis. Long and Zhang [23] further proposed an integrated framework for agent based inventory–production–transportation modeling and the distributed simulation of supply chains. The two categories of simulation methodologies have their own advantages in the modeling and representation of a supply chain network; however, they have deficiencies. Process-oriented simulation can describe the process flows on the basis of centralized simulation, but fails to model the micro 4

ACCEPTED MANUSCRIPT

proactive entities as well as their inter-organizational operation realities, partial information sharing and geographical heterogeneous distribution and to support a distributed simulation implementation. Agent-based simulation can fill most deficiencies of process-oriented simulation; however, it is inadequate in representing the operational processes of a supply chain network. Therefore, it is urgent to integrate the two. Vieira, Barbosa-Povoa and Martinho [35] presented a multi-agent supply chain system model that integrates different supply chain processes. Chatfield, Hayya and Harrison [5] presented a conceptual architecture that combines simulation formalisms, which allows an agent representation of the supply-chain infrastructure while enabling a process-oriented approach to representing orders. Govindu

CR IP T

and Chinnam [11] proposed a generic process-centered methodological framework, the Multi-Agent Supply Chain Framework (MASCF), to simplify the MAS development for supply chain applications. MASCF introduces the notion of the process-centered organization metaphor; in addition, it creatively adopts the Supply Chain Operations Reference (SCOR) model into a well-structured generic MAS analysis and design methodology, called Gaia, for multi-agent supply chain system development. Long [20] proposed a methodology for distributed supply chain network modelling and simulation by means

AN US

of the integration of agent-based distributed simulation and an improved SCOR model. The methodology contains two components: a hierarchical framework for modelling a supply chain network based on the improved SCOR model and agent building blocks that integrate the standard processes from the SCOR model.

The integration of process-oriented and agent-based simulation provides a sufficient and efficient approach to represent the supply chain network realities and to implement distributed simulation. However, the current studies have deficiencies. First, an effective inter-organizational collaborative

M

simulation framework should be developed based on the collaborative knowledge representation of a supply chain network. Second, multi-dimensional flows, as the driving forces for a supply chain network operation, should be included in the process-oriented and agent-based methodologies.

ED

Problems regarding the integration of the agent-based, flow-centric and process-oriented methodologies should be resolved. Third, to support simulation model implementation in a decentralized manner and in a partial information sharing context, it is indispensable to establish a

PT

collaborative framework that includes all inter-organizational collaborative knowledge and that considers the integration of the three methodologies. Thus, to fill the research gap, this paper addresses

CE

the research on inter-organizational collaboration operation for supply chain networks and establishes a multi-methodological collaborative simulation framework to support collaborative decision making.

AC

3. Inter-organizational collaborative simulation 3.1 Integrated collaborative simulation framework To simulate inter-organizational collaborative operation, an integrated collaborative simulation

framework is proposed, as shown in Fig. 1. This framework consists of three layers: a collaborative simulation platform, a knowledge framework for collaborative simulation and supply chain network emergence. The platform is the basis, whereas the collaborative simulation knowledge is the core in supporting supply chain network emergence by means of virtual reproductions. Simulation platform layer: it provides related simulation software packages to implement simulation models. Models established in the knowledge framework layer are essentially multi-agent systems. The models can run in the corresponding agent-based simulation software packages. The 5

ACCEPTED MANUSCRIPT models’ simulation processes are observed, and the data are collected to support what-if analysis of supply chain network emergence. MASS_SCN [17, 22] is an agent-based simulation system for a supply chain network that matches the modeling approach in the knowledge framework layer. Other agent-based simulation software packages can also be adopted in the integrated framework. Knowledge framework layer: this layer shows the knowledge and its modeling method for an inter-organizational collaborative simulation model. This framework integrates several methodologies so that it can support knowledge representation for inter-organizational collaboration. Because a supply chain network is composed of enterprises with autonomous behaviors and decision making capabilities, its network structure is represented by a multi-agent system. This system is viewed as a network

CR IP T

composed of agent nodes and their edges because agents have similar characteristics to enterprises. In a supply chain network, the material flow is driven by the information flow (i.e., orders) in addition to the time flow [21]. It is difficult to represent the nature of collaborative operation of the flow-centric supply chain network with a multi-agent system. In addition, the inter-organizational collaboration for a supply chain network relies on smooth information flow, efficient material flow and lean time flow. These three-dimensional flows alone can cross the enterprises’ boundaries to achieve nearly the same manner

as that of

AN US

performance of inter-organizational collaboration in a decentralized

intra-organizational collaboration in a centralized manner. From the perspective of a complex system, the supply chain network performance emerges through micro agents’ actions and their interactions. The operational process can be regarded as the sequence of agents’ actions. For better representation of supply chain network operation, micro operational processes can be explored to enrich the knowledge of agents. The proposed knowledge framework integrates three types of knowledge of multi-agent systems, three-dimensional flows and operational processes to develop inter-organizational

M

collaborative simulation models for a supply chain network.

Supply chain network emergence layer: data collected during the execution of simulation models in the platform supports’ emergence analysis are in this layer. This layer is responsible for collecting

ED

simulation data and analyzing operational performance on the basis of KPI to optimize a supply chain network. An evaluation of this layer will provide guidance for adjusting knowledge representations in

AC

CE

PT

the simulation models and conducting further experimental analysis.

6

ACCEPTED MANUSCRIPT

Supply chain network emergence

Knowledge framework for collaborative simulation Time flow

M

AN US

Multi-agent system

Information flow

CR IP T

Material flow

AC

CE

PT

ED

Operational processes

MASS-SCN

Fig. 1. Integrated collaborative simulation framework

3.2 Knowledge framework for collaborative simulation The knowledge framework for collaborative simulation depicts the knowledge for simulation models. As shown in Fig. 1, the proposed knowledge framework displays a knowledge system for a supply chain network representation in the three-dimensional views of multi-agent systems, flows and processes. In the framework: a multi-agent system is adopted to develop the inter-organizational network structure; the three-dimensional flows represent the operational mechanisms; and the 7

ACCEPTED MANUSCRIPT

processes instantiate agents' behaviors. All of these are integrated in the inter-organizational collaborative simulation model. 3.2.1 Supply chain network structure In general, a supply chain network can be considered as the network composed of nodes and their relationships. This network can be described as G( V , E ) by graph theory. V refers to the nodes that can be located in different layers of a supply chain network to represent enterprises, departments or processing units. E means the arc between two nodes, representing their relationships, which can be treated as 1 or 0. The value 1 shows a relationship among two nodes exists, whereas 0 implies no

CR IP T

relationship. In agent-based modeling, nodes are replaced by agents, whereas arcs are replaced by agent’s relationships. Thus, nodes and their relationship are transformed into a multi-agent system. Considering the hierarchy of a supply chain network and the flexibility of multi-agent system modeling, a supply chain network can be modeled at different levels to obtain a different-scaled multi-agent system. An agent in the upper level can be decomposed into a multi-agent system that is composed of multiple agents and their relationships in the lower level, as shown in Fig. 2. Therefore, a supply chain

SCN

modeled

in

the

i

level

can

be

described

as

AN US

network

Structure  ( Agentsleveli , Agents' Re lationshipsleveli ) .

SCN Structure

Time

M

Status Inputs Outputs

ED

Demands Events Policies Processes

Arcs

Agents/ MAS

Agents’ relationships

PT

Actions

Nodes

Sub-Nodes

Sub-Arcs

Agents/ MAS

Agents’ relationships

Policies generation Actions generation

AC

CE

Others

Time Status InitialAgents EndAgents Strategic relationship Logistics delivery Information communication

Sub-Sub-Nodes

Sub-Sub-Arcs

Agents/ MAS

Agents’ relationships

Others communication cost

Fig. 2. Supply chain network structure

In Fig. 2, the agent nodes represent micro entities’ that participate in inter-organizational collaborative operation. In simulation models, these entities can be described as a vector space 8

ACCEPTED MANUSCRIPT

composed of such elements as logical time, status, inputs, outputs, demands, events, policies, actions, processes, policy generation functions and action generation functions. This description can be expanded according to the realistic needs. The description is shown below: Agent  T , S , I ,O, D, E , P, A, PR, f P , f A ,Others 

(1)

The agent is an active entity. In view of the features of an agent and supply chain network, agent internal mechanisms should be developed in simulation models. An agent’s event can be determined by both its inputs and demands or solely by its demands at a specific logical time, as shown: (T  I  D ) (T  D )  E

(2)

to generate production events once production inputs approach.

CR IP T

For example, under the guidance of demands from their customers, enterprises will be motivated Policies can be generated at a specific logical time by the status, inputs and demands of agents or generated by their status and demands. Therefore, policy generation functions can be described as follows:

f p : (T  S  I  D ) (T  S  D )  P

(3)

AN US

For example, enterprises’ production policies, inventory policies or delivery policies are established according to their current production, inventory or delivery status on the basis of their demands plus the triggers of inputs at a specific logical time. Occasionally, enterprises are likely to develop corresponding policies on the basis of their current status and demands, although there is no trigger of inputs.

According to Formula (2) and Formula (3), policies can also be generated by status and events at a specific logical time, shown as follows:

M

fp :T S  E  P

(4)

In addition, agents’ actions are generated at a specific logical time under the guidance of their

ED

policies, as shown in the following formula: fA :T  P  A

(5)

For example, at a specific time, enterprises will take production actions based on the established

PT

production policies and achieve their expected outcomes. Correspondingly, agents’ actions have influences on their status, outputs or internal inputs. The

CE

action’s effects can be shown as follows: T  A ( S O  I )

(6)

A production action will change the enterprise’s status of production, inventory and delivery.

AC

Furthermore, the action can affect both its upstream and downstream enterprises by its outputs. Enterprises can make adjustments to their production actions in view of the action effects. On the basis of the analysis noted above, the essential operational feature of a supply chain

network, the agent’s process, can be referred to as the sequence of actions or events; this is shown as follows:

PR  A,Q    E ,Q 

(7)

The arc of an inter-organizational network structure can be modeled as a vector space composed of such elements as logical time, status, initialAgents, endAgents, strategic relationship types, logistics delivery types and information communicative methods, shown as follows: Arc  T , S , IA, EA, SR, LD, IC ,Others 

(8)

Other attributes related to a supply chain network (i.e., agent capabilities) and cost can be added 9

ACCEPTED MANUSCRIPT

into corresponding vectors according to the actual needs. By means of formal descriptions noted above, the structure, the basic attributes and the behavior characteristics of a multi-agent system are defined. 3.2.2 Supply chain network flows Various flows exist in a supply chain network, among which material flow, information flow and capital flow are comparatively important. Rai, Patnayakuni and Seth [30] studied IT-enabled supply chain process integration of physical flows, information flow and financial flow. Inter-organizational collaboration of supply chain network is realized by these flows. Collaborative simulation in this paper

CR IP T

attaches importance to the research on operations and does not consider the capital flow. In addition, the time flow is not only another important factor for representing the operational mechanisms but also a key means for supporting the distributed implementation of a virtual collaborative simulation. Therefore, material, information and time flows will be considered in the collaborative simulation models. The three flows describe the operational mechanisms for inter-organizational collaboration. These flows can be abstracted into a vector space composed of logical time, flow path, flow inputs and

AN US

outputs, components and the transformation function. All the three flows can be referred to with the following formula:

Flow  Time, Paths, Inputs ,Outputs,Components ,Transformation 

(9)

In the formula, the flow path shows that the flow moves along with agents in the network. The inputs and outputs of the flows are the materials, information and time that move into and out of the path. The components are the transitional forms of the materials, information and time in the path. The transformation function provides the transformation rules from inputs to outputs.

M

A supply chain network is hierarchical; hence, the three flows also show hierarchy. A flow can be decomposed into detailed sub-flows. The hierarchy of the flows can be shown by the following formula:

 Sub _ Flow

ED

Flow 

i

i

 min( Timei ),





PT

i

Pathi ,

i

Input i ,



Outputi ,

i

 i

Component i ,



Transformationi 

(10)

i

The logical time of a flow uses the minimum of that in the sub-flows to maintain the consistency

CE

of the time advancement.

In a real supply chain network, the three flows can be coordinated by various means (e.g., prediction, lead time, information exchange, contract constraints and demands distribution) to support

AC

collaborative operation and evolution. Collaborative simulation for a supply chain network aims to simulate the three flows. The three flows must be consistent; their collaboration can be reflected in maintaining the consistency of their processes and results. To maintain the consistency, certain methods can be adopted, as shown in the formula: Flow Collaboration  Consistency( Methods , PRocesses , Re sults( MF , IF ,TF )

(11)

The method of collaborative framework will be discussed in details in Section 4, which ensures correct time advancement for collaborative simulation. There are interactions among the three flows in the simulation, as shown in Fig. 3.

10

ACCEPTED MANUSCRIPT

Time flow

Drive

Drive Feed back

Feed back

Feed back

Drive

CR IP T

Information flow

Material flow

Fig. 3. Interactions between material, information and time flows

Information flow drives the operation of material flow, whereas the material flow can provide

AN US

guides to adjust the information flow with its operational feedback. Both the material flow and the information flow together drive the evolution of the time flow. Correspondingly, the time flow can affect the material flow and the information flow with its operational feedback. The interactions of the three flows for three-dimensional-flow-based collaborative simulation reflect the operational essence of inter-organizational collaboration of supply chain network. 3.2.3 Supply chain network processes

M

The micro operation of a supply chain network is obviously process-oriented. The macro phenomena of a supply chain network emerge from the micro processes of the entities. The advancement of the material, information and time flows is also process-oriented at the micro level. In following formula:

ED

a supply chain network, the process is the sequence of actions or events, which can be described by the PR  Elements , Attributes,Characteristics, Functions ,Triggers, Effects 

(12)

PT

In the formula, elements are the specific actions or events in the process. A process has its attributes and characteristics and can be triggered by certain conditions to implement its function that

CE

brings certain effects.

A process-oriented methodology is consistent with the three flows discussed in the previous sections. A flow can be considered as the sequence of processes. A process-oriented methodology and

AC

an agent-based methodology are different. The agent is regarded as the implementing entity for processes in the agent-based collaborative simulation. Hence, the characteristics of processes can be better reflected in the multi-agent system. Processes are included in agents through Formula (7). That is, a process is the sequence of actions or events. Actions are the attributes of an agent, whereas events are the elements in an agent’s events queue. Therefore, actions, events, processes and flows are integrated into the structure of a multi-agent system. In the knowledge framework, the process can be divided into six categories: production process, inventory process, delivery process, production planning process, inventory planning process and delivery planning process. Table 1 shows their descriptions, process elements and the corresponding agent types. Table 1. Processes in a supply chain network

11

ACCEPTED MANUSCRIPT

Categories

Descriptions

Process elements

Agent types

Production

A sequence of production

Schedule production activities, Issue product, Produce and test, Package,

Production

process

actions or production

Stage finished product, Release finished product to deliver and Waste

Agent

events

disposal.

A sequence of inventory

Manage product inventory, Manage incoming product, Manage

actions or inventory events

import/export requirements, Manage finished goods inventory, Manage

Inventory process

Inventory Agent

in-process products (WIP) and Manage transportation (WIP). Delivery process

A sequence of delivery

Process inquiry and quote, Receive, configure, enter and validate order,

actions or delivery events

Reserve inventory and determine delivery date, Consolidate orders, Build

Delivery Agent

CR IP T

loads, Route shipments, Select carriers and rate Shipments, Receive

product from source or make, Pick product, Pack product, Load product

and generate shipping docs, Ship product, Receive and verify product by customer, Install product and Invoice. A sequence of production

Identify, prioritize and aggregate production requirements; Identify,

Production

planning process

planning actions or

assess and aggregate production resources; Balance production resources

Planning Agent

production planning events

with production requirements; and Establish production plans.

Inventory

A sequence of inventory

Identify, prioritize and aggregate inventory requirements; Identify, assess

Inventory

planning process

planning actions or

and aggregate inventory resources; Balance inventory resources with

Planning Agent

inventory planning events

production requirements; and Establish inventory plans.

Delivery planning

A sequence of delivery

Identify, prioritize and aggregate delivery requirements; Identify, assess

Delivery

process

planning actions or

and aggregate delivery resources; Balance delivery resources with

Planning Agent

delivery planning events

delivery requirements; and Establish delivery plans.

M

AN US

Production

Table 2 describes the relations between the processes and flows. All the three flows include a

ED

sequence of corresponding processes. Table 2. Processes of flows

Material flow

Microscopic processes

PT

Categories

Production process, Inventory process, Delivery process. Production planning process, Inventory planning process, Delivery planning process.

Time flow

Request time advancement process, Balance and authorize time advancement process.

CE

Information flow

AC

3.3 Supply chain network emergence To clarify the combination of knowledge regarding multi-agent systems, flows and processes,

their relations in a collaborative simulation framework are delineated in Fig. 4.

12

ACCEPTED MANUSCRIPT

Composition

Processes

Flows Decomposition

Enabling

Analysis Driver

MAS

CR IP T

Execution

Fig. 4. Relations between multi-agent systems, flows and processes

AN US

As shown in Fig. 4, multi-agent systems execute the processes, whereas the processes enable multi-agent systems with certain functions. Multi-agent systems provide guidance to the analysis of the flows, whereas the flows drive the operation of multi-agent systems. The processes can comprise the flows, and the flows can be divided into the processes.

Supply chain network emergence originates from its micro operations. Based on the previous analysis of a knowledge framework, a supply chain network can be abstracted into the aggregation

M

emergence of processes, flows and an inter-organizational structure. SCN  f ( PRocesses   MF , IF ,TF  SCN _ Structure )

(13)

f is the emergence function based on the processes, flows and structures. The function is non-transparent. The emergence results of inter-organizational collaboration are determined by the

ED

knowledge and the knowledge interaction mechanisms in a supply chain network.

PT

4. Simulation collaborative framework

In a distributed environment, all of the knowledge in simulation models need to be combined in a

CE

unified framework that can coordinate the entity’s behaviors, process interactions and the time sequence of events. A collaborative framework that integrates multiple simulation formalisms for inter-organizational supply chain network simulation is proposed in this paper.

AC

4.1 Multiple formalisms for simulation collaboration Supply chain network operation has the characteristics of obvious time series increments, event

scheduling, policy control, process interaction and activity scanning. The formalism used for describing the dynamic structure of a simulation model is referred to as world views, or formalisms [5]. The current studies on supply chain network simulation put forth three formalisms: event scheduling, activity scanning and process interaction [5]. In the event scheduling formalism, each entity in a supply chain network is defined as an event queue, in which each event is stamped with a logical time. In a simulation run, the event with the earliest logical time will be scheduled preferentially; its corresponding actions will be accorded priority to be executed. Furthermore, the global virtual time will advance to this logical time. Through event 13

ACCEPTED MANUSCRIPT

scheduling, the behaviors of all entities in the supply chain network are executed according to the conservative time increments. This behavior ensures the consistency between the virtual simulation world and the real world. In the activity scanning formalism, the status and conditions of the entities will be continuously scanned. When the status and conditions are satisfied, the events with the earliest logical time will be triggered in the event queue. Similar to the event scheduling formalism, the scanning formalism is also a time increment conservative collaboration method. Usually the two formalisms are combined to ensure the simulation processes. The process interaction formalism emphasizes the role of the entities’ process-based interactions

CR IP T

for time advancement during the simulation run. The process is a sequence of events or actions. The essence of this formalism is event scheduling or activity scanning; however, it shows a difference in the triggering mechanism and the life cycle view. The process interaction is more advantageous in reflecting the essence of inter-organizational collaborative operation, particularly in depicting the process flows in a supply chain network.

The three formalisms will be integrated in this paper with other characteristics of

AN US

inter-organizational collaborative operation. Based on the integration, an effective simulation collaborative framework is proposed. 4.2 Collaborative framework

The knowledge regarding time, events, actions, activities, processes, policies and status will be integrated into an integrated collaborative framework presented in this paper. This framework defines a mechanism that integrates time series increments, event scheduling, policy control, process interaction framework is shown in Fig. 5.

M

and activity scanning to support inter-organizational collaborative simulation. This collaborative In the collaborative framework, each agent has its own time queue, event scheduling queue, policy

ED

queue, action set, process set and activity set and local virtual time (LVT). Each agent is responsible for its simulation advancement in the simulation run. The advancing process is controlled by its LVT. In the inter-organizational collaboration for a supply chain network, agents implement process interaction,

PT

whose basic form is message communication. The time advancement requests reflected in each agent’s

AC

CE

LVT are synchronized by global virtual time (GVT), as shown in Fig. 5.

14

ACCEPTED MANUSCRIPT

GVT

LVT

Time queue

t1 t 2 t 3

AN US

Actions

Interactions

Processes

Processes

Policies

Activities

Activities

Actions

S

CR IP T

t 3 t 2 t1

Events

e3 e2 e1

LVT

Time queue

S

e1 e2 e3

Policies

M

Fig. 5. The collaborative framework

At a specific simulation logical time, each agent generates its events by the inputs and demands received during its process interactions according to Formula (2) and stores the events in its event

ED

queue. In the process of event scheduling, each agent determines its policies by combining its scheduled events and status at a specific logical time according to Formula (3) and (4). The agent generates its actions by the determined policies at a specific logical time according to Formula (5). As

PT

the elements of the activity and process, the actions enable the agents to participate in inter-organizational operation of the supply chain network. Each action creates a request for its logical

CE

time advancement. The authorization for the time advancement request relies on whether the requested logical time is smaller than or equal to the global virtual time. If yes, the agent will be allowed to advance to this requested logical time to schedule the events, implement corresponding actions and

AC

advance its local virtual time to this time. In the process, the agent interacts with other agents and sends outputs to other agents to stimulate other agents to generate events, determine policies, take actions and request time advancements. After this circle, the agent receives new inputs, rescans its activities and status and moves to the next time advancement request. In the collaborative framework, material, information and time flows reflected by agents’ events

and actions are correspondingly collaborated when the events and actions are collaborated. Collaboration is obviously process-oriented. The three-dimensional flows are composed of multiple sub-flows. The flows’ inputs and outputs and components’ transformations need time advancements according to Formula (9) and (10). According to Formula (11), material, the information and time flows are collaborated by the proposed collaborative framework. In the collaboration process, the three-dimensional flows display the interactions in Fig. 3. 15

Events

ACCEPTED MANUSCRIPT

In the collaborative framework, each agent interacts with others and advances the inter-organizational collaborative operation of the supply chain network based on its processes. In the process interaction, agents’ processes are collaborated through actions or events according to Formula (7) on the basis of the process definition in Formula (12). According to Formula (13), the supply chain network structure, flows and processes collaborate with each other in the simulation run, emerging their macro phenomena with the methodology of complex system research. In conclusion, the framework in this paper integrates several simulation formalisms, namely, time series increments, event scheduling, policy control, process interaction and activity scanning, to

CR IP T

support more effective and comprehensive collaboration; thus, it can achieve what-if analysis of supply chain network performance through the knowledge representation of structures, flows and processes.

5. Case study

This paper presented an approach for inter-organizational collaborative simulation. This approach

AN US

is a type of logical approach to support the development of simulation models. The model covers knowledge regarding agents, processes, flows and integrates several methodologies; however, it is essentially a multi-agent system. The model can be implemented in any agent-based simulation platform. In this paper, the MASS-SCN will be utilized to implement the multi-agent system. A five-level manufacturing supply chain network is studied using the proposed framework. As shown in Fig. 6, this network is composed of five levels: the first level is composed of three suppliers; the second by three manufacturers; the third by two manufacturers; the fourth by two distributors; and

M

the fifth by three customers. This network produces three types of products for three customers, respectively. Competition and cooperation coexist among these enterprises. These enterprises compete with each other in the vertical direction but cooperate with each other in different degrees in the

AC

Supplier 3

Customer 1 Manufacturer 4

Distributor 1

Manufacturer 2

CE

Supplier 2

Manufacturer 1

PT

Supplier 1

ED

horizontal direction.

Customer 2 Manufacturer 5

Distributor 2

Manufacturer 3

Customer 3

Fig. 6. A five-level manufacturing supply chain network

In this case, a collaborative simulation model is developed based on the proposed knowledge framework. The knowledge framework focuses on agent, flow and process modeling using multiple methodologies. First, the model is developed at the operational level; its network structure consists of 35 agents. A supplier is modeled as an inventory agent and a delivery agent; a manufacturer is modeled as a material inventory agent, a production agent, a product inventory agent and a delivery agent. A distributor is modeled as an inventory agent and a delivery agent; a customer is modeled as an 16

ACCEPTED MANUSCRIPT

inventory agent. In addition, a production planning agent and a delivery planning agent are modeled to coordinate the agents’ production and delivery operations. The agent’s relations are determined according to the enterprises’ strategic relationship types, logistics delivery types and information communicative methodologies. Second, according to the network structure and Formula (10), material, information and time flows are defined. The flows’ definitions include logical time, flow path, flow inputs and outputs, components and transformation function. Third, according to Tables 1 and 2, the processes related to the agents and the flows are abstracted to support the exploration of the supply chain network operations. Finally, the model is instantiated as a multi-agent system and implemented using the proposed simulation collaborative framework by the MASS-SCN. The modeling process and

CR IP T

model implementation of the case indicate that the proposed framework is feasible.

The data gathered during and after the simulation run is conducted using KPI analysis for decision making. The KPI analysis verifies the validity of the proposed framework. In this case, several KPIs are selected to support the analysis and decision making of the supply chain network. The order fulfillment is an important indicator for a supply chain network performance evaluation. This indicator focuses on material, information and time flows of the agents of customers in product delivery

AN US

processes. Fig. 7 is the time series analysis of order fulfillments of three customers. Customer 1 has priority over customer 2 from distributor 1; customer 3 has priority over customer 2 from distributor 2. Thus, the orders of customers 1 and 3 are preferentially fulfilled. As shown in Fig. 7, customer 1 ranks first in the degree of satisfaction, followed by customer 3 in second and customer 2 in third. At the logical time 9940 (it can be converted into a real product delivery cycle), customer 1 has an order

AC

CE

PT

ED

M

fulfillment rate of 100%; customer 3 has a rate of 50%, whereas customer 2 has a rate of 0%.

Fig. 7. Order fulfillments without policies adjustments

Appropriate decision making can be derived from the order fulfillments. Without reducing the degree of satisfaction of customers 1 and 3, adjustments can be made to the production and delivery policies related to product 2 to improve the customer 2’s satisfaction. Fig. 8 is the time series analysis of the order fulfillments of the three customers after delivery policies adjustments. These adjustments’ 17

ACCEPTED MANUSCRIPT

objective is to increase the delivery priority of the materials related to product 2 at the upstream of manufacturers 4 and 5 in the sub-network. Accordingly, the degree of satisfaction of customer 2 is improved. At the logical time 9940, the order fulfillment rate of customer 1 remains at 100%; the rate

AN US

CR IP T

of customer 3 increases to 67%, whereas the rate of customer 2 increases to 30%.

M

Fig. 8. Order fulfillments after delivery policies adjustments

Fig. 9 is the time series analysis of manufacturers’ productions. This indicator considers material and time flows of the agents of manufacturers in production processes. Fig. 9 displays that the

ED

productions in the upstream sub-network are very stable because a stockout rarely occurs. However, in the downstream sub-network, certain materials are out of stock, which results in the production fluctuation of manufactures 4 and 5. On the basis of the analysis, the production and delivery policies

PT

can be optimized in the upstream sub-network to reduce the production fluctuation and enhance the

AC

CE

network overall performance.

18

CR IP T

ACCEPTED MANUSCRIPT

AN US

Fig. 9. Time series analysis of manufacturers’ productions

Fig. 10 is the time series analysis of material inventories of manufacturers 4 and 5. This indicator cares about material and time flows of the agents of manufacturers in inventory processes. The figure shows that manufacturer 5 has a higher material inventory level than manufacturer 4. The unit production time for product 3 is longer than that for product 1, which leads to the backlog of materials. The analysis can guide manufacturer 5 to suitably increase the production capacity or to coordinate

AC

CE

PT

ED

M

materials deliveries with its suppliers, thereby reducing its material inventory level.

Fig. 10. Time series analysis of material inventories of manufacturer 4 and 5

Therefore, this case study signifies that the proposed methodology has obvious advantages, particularly in efficient and comprehensive descriptions of supply chain network models. The attributes of agents, the essence of process operation, and operational mechanisms of flows are included to 19

ACCEPTED MANUSCRIPT

support more effective collaborative decision making. In addition, the effectiveness and rationality of a combination of multiple simulation formalisms, such as time series increments, event scheduling, policy control, process interaction and activity scanning, are verified by the case study.

6. Conclusion and further study Inter-organizational collaborative simulation has become an effective tool for analyzing complex supply chain network. A supply chain network covers the knowledge of agent, flow and process that should be extracted, modeled and instantiated in corresponding collaborative simulation models. This

CR IP T

paper presents a framework that integrates the knowledge representation of a multi-agent system, three-dimensional flows and operational processes for inter-organizational collaborative simulation. This framework solves the problems regarding integration of the agent-based, flow-centric and process-oriented methodologies. A knowledge framework for collaborative simulation is also established in this paper. The framework defines three types of knowledge with its formal descriptions and delineates the methodology with which the three types of knowledge enable the macro phenomena

AN US

of supply chain network to emerge. In addition, an inter-organizational collaborative method that integrates the multiple simulation formalisms, such as time series increments, event scheduling, policy control, process interaction and activity scanning, is proposed to back up the collaborative simulation advancement under the knowledge framework. The knowledge framework clarifies the logical modeling for inter-organizational operation, whereas the collaborative method solves the problem regarding the synchronization in implementing logical models. Blending the framework and method contributes to the establishment of a logical solution for inter-organizational operation analysis. The

M

solution has been verified in a case study. This proposed framework, for one thing, provides an effective knowledge representation for inter-organizational collaboration of supply chain networks; for another, it supports the achievement of a more realistic and efficient inter-organizational collaborative

ED

virtual operation in a decentralized manner with a unified collaborative framework. Both contribute to conducting KPI analysis for collaborative decision making. In practice, the collaborative simulation for inter-organizational supply chain networks often

PT

involves more factors, such as complex intelligence of agent and knowledge flow. Further research will focus on the methodology improvement and more real-world case studies of supply chain networks on

CE

the basis of this methodology.

AC

Acknowledgements

This work was supported by the National Natural Science Foundation, China (No. 71401153), the

Humanities and Social Science Youth Foundation of the Ministry of Education of the Republic of China (No. 14YJC630090) and Zhejiang Science & Technology Plan of China (No. 2015C33080).

References [1] M. Amini, T. Wakolbinger, M. Racer, M.G. Nejad, Alternative supply chain production-sales policies for new product diffusion: An agent-based modeling and simulation approach, European Journal of Operational Research, 216 (2) (2012) 301-311. [2] Z. Bahroun, M. Moalla, G. Baazaoui, J.P. Campagne, Multi-agent modelling for replenishment 20

ACCEPTED MANUSCRIPT

policies simulation in supply chains, European Journal of Industrial Engineering, 4 (4) (2010) 450-470. [3] B.M. Beamon, Measuring supply chain performance, International Journal of Operations & Production Management, 19 (3) (1999) 275-292. [4] S. Cannella, M. López-Campos, R. Dominguez, J. Ashayeri, P. Miranda, A simulation model of a coordinated decentralized supply chain, International Transactions in Operational Research, In press, 2015. [5] D.C. Chatfield, J.C. Hayya, T.P. Harrison, A multi-formalism architecture for agent-based, order-centric supply chain simulation, Simulation Modelling Practice and Theory, 15 (2) (2007) 153-174.

CR IP T

[6] J.H. Cheng, Inter-organizational relationships and information sharing in supply chains, International Journal of Information Management, 31 (4) (2011) 374-384.

[7] N.Q. Chinh, Z.P. Li, T.P. Siew, X.S. Chen, O.Y. Soon, An agent-based simulation to quantify and analyze bullwhip effects in supply chains, In: Proceedings of IEEE International Conference on Systems Man and Cybernetics Conference, Manchester, 2013, pp.4528-4532.

[8] P. Ding, H.S. Ding, K. He, J. Zhang, Analysis of cooperation and competition in flexible supply

AN US

chain network based on multi-agent simulation, In: Proceedings of IEEE International Conference on Intelligent Computing and Intelligent Systems, Shanghai, 2009, pp.908-912.

[9] M. Fasli, Y. Kovalchuk, Learning approaches for developing successful seller strategies in dynamic supply chain management, Information Sciences, 181 (16) (2011) 3411-3426.

[10] M. Frohling, F. Schwaderer, H. Bartusch, O. Rentz, Integrated planning of transportation and recycling for multiple plants based on process simulation, European Journal of Operational Research, 207 (2) (2010) 958-970.

M

[11] R. Govindu, R.B. Chinnam, MASCF: A generic process-centered methodological framework for analysis and design of multi-agent supply chain systems, Computers & Industrial Engineering, 53 (4) (2007) 584-609.

ED

[12] A. Groznik, M. Maslaric, Investigating the impact of information sharing in a twolevel supply chain using business process modeling and simulations: A case study, In: Proceedings of 23rd European Conference on Modelling and Simulation, Madrid, 2009, pp. 39-45.

PT

[13] W. Jakkhupan, S. Arch-int, Y.F. Li, Business process analysis and simulation for the RFID and EPCglobal Network enabled supply chain: A proof-of-concept approach, Journal of Network and

CE

Computer Applications, 34 (3) (2011) 949-957. [14] A. Karagiannaki, G. Doukidis, K. Pramatari, A framework for mapping the RFID-enabled process redesign in a simulation model, Journal of the Operational Research Society, 65 (11) (2014) 1700-1710.

AC

[15] O. Labarthe,B. Espinasse, A. Ferrarini, B. Montreuil, Toward a methodological framework for agent-based modelling and simulation of supply chains in a mass customization context, Simulation Modelling Practice and Theory, 15 (2) (2007) 113-136. [16] J. Li, Z.H. Sheng, H.M. Liu, Multi-agent simulation for the dominant players' behavior in supply chains, Simulation Modelling Practice and Theory, 18 (6) (2010) 850-859. [17] J. Lin, Q.Q. Long, Development of a multi-agent-based distributed simulation platform for semiconductor manufacturing, Expert Systems with Applications, 38 (5) (2011) 5231-5239. [18] C.C. Lin, T.H. Wang, Build-to-order supply chain network design under supply and demand uncertainties, Transportation Research Part B: Methodological, 45 (8) (2011) 1162-1176. [19] Q.Q. Long, An agent-based distributed computational experiment framework for virtual supply chain network development, Expert Systems with Applications, 41 (9) (2014) 4094-4112. 21

ACCEPTED MANUSCRIPT

[20] Q.Q. Long, Distributed supply chain network modelling and simulation: Integration of agent-based distributed simulation and improved SCOR model, International Journal of Production Research, 52 (23) (2014) 6899-6917. [21] Q.Q. Long, Three-dimensional-flow model of agent-based computational experiment for complex supply network evolution, Expert Systems with Applications, 42 (5) (2015) 2525-2537. [22] Q.Q. Long, J. Lin, Z.X. Sun, Modeling and distributed simulation of supply chain with a multi-agent platform, International Journal of Advanced Manufacturing Technology, 55 (9-12) (2011) 1241-1252. [23]

Q.Q.

Long,

W.Y.

Zhang,

An

integrated

framework

for

agent

based

CR IP T

inventory-production-transportation modeling and distributed simulation of supply chains, Information Sciences, 277 (2014) 567-581.

[24] F. Longo, G. Mirabelli, An advanced supply chain management tool based on modeling and simulation, Computers & Industrial Engineering, 54 (3) (2008) 570-588.

[25] A. Manataki, Y.H. Chen-Burger, M. Rovatsos, Towards improving supply chain coordination through agent-based simulation, Advances in Intelligent and Soft Computing, 71 (2010) 217-224.

AN US

[26] K.J. Mizgier, S.M. Wagner, J.A. Holyst, Modeling defaults of companies in multi-stage supply chain networks, International Journal of Production Economics, 135 (1) (2012) 14-23. [23] M. Mohammadi, M.B. Mukhtar, H.R. Peikari, A grammar-based process modeling and simulation methodology for supply chain management, Lecture Notes in Computer Science, 7066 (1) (2011) 77-86.

[24] Y.T. Ni, Y. Wang, Development of an agent-based collaborative production system based on real-time order-driven approach, Arabian Journal for Science and Engineering, 40 (4) (2015)

M

1239-1253.

[25] F. Persson, SCOR template—A simulation based dynamic supply chain analysis tool, International Journal of Production Economics, 131 (1) (2011) 288-294.

ED

[30] A. Rai, R. Patnayakuni, N. Seth, Firm performance impacts of digitally enabled supply chain integration capabilities, MIS Quarterly, 30 (2) (2006) 225-246. [31] G. Reiner, Customer-oriented improvement and evaluation of supply chain processes supported by

PT

simulation models, International Journal of Production Economics, 96 (3) (2005) 381-395. [32] L.A. Santa-Eulalia, S. D'Amours, J.M. Frayret, Agent-based simulations for advanced supply

CE

chain planning and scheduling: The FAMASS methodological framework for requirements analysis, International Journal of Computer Integrated Manufacturing, 25 (10) (2012) 963-980. [33] A.A. Tako, S. Robinson, The application of discrete event simulation and system dynamics in the

AC

logistics and supply chain context, Decision Support Systems, 52 (4) (2012) 802-815. [34] W.Z.Tan, Y.T. Chai, Y. Liu, A message-driving formalism for modeling and simulation of multi-agent supply chain systems, Journal of Systems Science and Systems Engineering, 20 (4) (2011) 385-399.

[35] C.M. Vieira, A.P. Barbosa-Povoa, C. Martinho, An agent-based collaborative model for supply chain management simulation, In: Proceedings of 26th European Conference on Modelling and Simulation, Koblenz, 2012, pp.318-324. [36] W. Wang, Y.T. Chai, J. Dong, H.W. Ding, C.R. Ren, M.M. Qiu, Evaluating the value of collaboration in supply chain through business process simulation, In: Proceedings of 2011 IEEE International Conference on Service Operations, Logistics and Informatics, Beijing, 2011, pp. 307-312. [37] J. Windisch, D. Roser, B. Mola-Yudego, L. Sikanen, A. Asikainen, Business process mapping and 22

ACCEPTED MANUSCRIPT

discrete-event simulation of two forest biomass supply chains, Biomass & Bioenergy, 56 (2013) 370-381. [38] J. Windisch, K. Väätäinen, P. Anttila, M. Nivala, J. Laitila, A. Asikainen, L. Sikanen, Discrete-event simulation of an information-based raw material allocation process for increasing the efficiency of an energy wood supply chain, Applied Energy, 149 (2015) 315-325. [39] T. Wu, S.M. Huang, J. Blackhurst, X.L. Zhang, S.S. Wang, Supply chain risk management: An agent-based simulation to study the impact of retail stockouts, IEEE Transactions on Engineering Management, 60 (4) (2013) 676-686. [40] J.P. Xu, Q. Liu, R. Wang, A class of multi-objective supply chain networks optimal model under 178 (8) (2008) 2022-2043.

CR IP T

random fuzzy environment and its application to the industry of Chinese liquor, Information Sciences, [41] C.C. Zhao, J.J. Qiu, Agent-based simulation for order selection strategy in collaboration process of supply chain, In: Proceedings of the 2014 IEEE 18th International Conference on Computer Supported Cooperative Work in Design, Hsinchu, 2014, pp.18-22.

[42] C.C. Zhao, J.J. Qiu, X.D. Zhang, Agent behavior based modeling and simulation for pricing

AN US

coordination in supply chain, Journal of Convergence Information Technology, 7 (23) (2012) 484-493. [43] G.Q. Zhang, G.L. Zhang, Y. Gao, J. Lu, Competitive strategic bidding optimization in electricity markets using bilevel programming and swarm technique, IEEE Transactions on Industrial Electronics,

AC

CE

PT

ED

M

58 (6) (2011) 2138-2146.

23