Scholarly article on topic 'A Multi-agent Based Decision –Making Approach for Field Service Delivery of IPS2'

A Multi-agent Based Decision –Making Approach for Field Service Delivery of IPS2 Academic research paper on "Civil engineering"

Share paper
Academic journal
Procedia CIRP
OECD Field of science
{"Industrial Product-Service Systems" / "Field service delivery" / "Multi-agent based simulation"}

Abstract of research paper on Civil engineering, author of scientific article — Rui Zhou, Yaoguang Hu, Shasha Xiao, Jingqian Wen

Abstract There is a trend increasing number of manufacturing companies offer field service to customers rather than only provide products for higher revenue and stronger competitiveness. However, these companies are experiencing a lack of proper and efficient decision support in maintenance policy selection and service resource planning of service delivery. In this paper, a multi-agent based approach to service decision-making of IPS2 is presented and the performance evaluation result is given. Multi-agent based approach is utilized to model the service delivery network and different agents are generated to present the elements of field service network. Each agent has a distinct state transition and processing mechanism which can be highly coincidence with the actual situation owing to the flexibility and operability of multi-agent based simulation model. Furthermore, negotiation mechanism among the agents is introduced into the multi-agent based simulation model to map the dynamic characteristics in the actual service delivery context. A case study scenario of agricultural machinery maintenance is presented which clearly demonstrated that the proposed approach is efficient and can provide reasonable suggestions to the IPS2 providers.

Academic research paper on topic "A Multi-agent Based Decision –Making Approach for Field Service Delivery of IPS2"

Available online at

¿IfjgP1 ScienceDirect

ELSEVIER Procedia CIRP 47 (2016) 228 - 233

Product-Service Systems across Life Cycle

A multi-agent based decision -making approach for field service delivery

of IPS2

Rui Zhou,Yaoguang Hu*,Shasha Xiao,Jingqian Wen

Beijing Institute of Technology,No.5 Zhongguancun South Street, Beijing, 100081, China * Corresponding author. Tel.: +86-010-68917880. E-mail address:


There is a trend increasing number of manufacturing companies offer field service to customers rather than only provide products for higher revenue and stronger competitiveness. However, these companies are experiencing a lack of proper and efficient decision support in maintenance policy selection and service resource planning of service delivery. In this paper, a multi-agent based approach to service decision-making of IPS2 is presented and the performance evaluation result is given. Multi-agent based approach is utilized to model the service delivery network and different agents are generated to present the elements of field service network. Each agent has a distinct state transition and processing mechanism which can be highly coincidence with the actual situation owing to the flexibility and operability of multi-agent based simulation model. Furthermore, negotiation mechanism among the agents is introduced into the multi-agent based simulation model to map the dynamic characteristics in the actual service delivery context. A case study scenario of agricultural machinery maintenance is presented which clearly demonstrated that the proposed approach is efficient and can provide reasonable suggestions to the IPS2 providers.

© 2016PublishedbyElsevierB.V This isanopen access article under the CC BY-NC-ND license (

Peer-review under responsibility of the scientific committee of the 8th Product-Service Systems across Life Cycle Keywords: Industrial Product-Service Systems; Field service delivery; Multi-agent based simulation

1. Introduction

Providing field services to customers has become increasing important for equipment manufacturing companies to obtain competitive advantages and profit level. It is as well a delivery way of Industrial Product-Service System (IPS2) for higher sustainability and maintainability. Specially, the following items have been contributing to the growing importance and motivation for offering field services in recent years. First of all, manufacturers can get detailed maintenance or repair data in providing of field service which will in turn feed back to the design and manufacturing process and ultimately improve the quality of products. Secondly, there is financial motivation that field service improves profitability by getting continuous revenue from service contracts. Finally, customers' personalized service demands which influence the purchase intention will be satisfied by the field service and can be a source for stronger market competitiveness of the manufacturer.

However, when an equipment manufacturing company decides to offer field services to its customers, several problems need to be tackled carefully and successfully in advance. The first problem a manufacturer needs to face is the maintenance policy selection. A maintenance policy that dictates which parameter triggers a maintenance and is usually critical for the effective service delivery at a lower cost [1]. Another particular challenge after the maintenance policy selection is the capacity planning and scheduling of required resources for service delivery [2]. In our case, customers of the manufacturer are geographically distributed and service requests are vast in number and have different types, such as installation, maintenance, repair or replacement. Consequently, the delivery of field service for IPS2 is increasing complex so as to higher request in the flexibility and robustness for decision-making of service providers. In order to guarantee a proper service capability and ensure a timely response to customer's uncertain service demand, a comprehensive capability planning for service stations is

2212-8271 © 2016 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 scientific committee of the 8th Product-Service Systems across Life Cycle doi: 10.1016/j.procir.2016.03.070

needed. Moreover, capacity planning should be made in a long term perspective aiming at stability of service capabilities and lower resource allocation costs. Subsequently, during the operative field service, delivery processes need to be specified and assigned to available resources. The main purpose of this part is to balance the limited resources with the service requests from different customers.

Researches have been done a lot about the problem above mentioned. However, previous researches are static optimization and mostly based on quantitative approaches and give identified solutions. Here quantitative methods are usually mathematical models and intelligent algorithms. In general , these models and algorithms can provide a satisfactory solution for the determined initial parameters. But it is obvious that the problem we need to tackle is highly dynamic with the external environment. In addition, subjecting to the limitation of expression by the mathematical model, there are simplified assumptions to limit the problem's complexity. The consequences of these simplifications will fail in capturing the most important uncertainties and dynamics that dominate the delivery of field service [3].

Due to these limitations, a comprehensive and rational decision-making method for field service delivery of IPS2 is still required. In this paper, we present a multi-agent based approach to solve service decision-making problem for the delivery management of IPS2 and the performance evaluation result is given. Multi-agent based approach is utilized to model the service delivery network and different agents are generated to present the elements of field service network. In addition, by fine-tuning parameters of the simulation model, the approach presented can provide effective mechanisms for the management of dynamic and complex operations in service delivery of IPS2.

The remainder of this paper is structured as follows. Section 2 provides a review of related studies. In section 3, we describe the problem and specify the process of different agents in our approach. Section 4 presents a simulation model for the field service in the process of service delivery. A case study is described and the experiment results and discussions are presented as well. The conclusions and discussions are stated in section 6.

2. Literature review

2.1. Field service delivery of IPS2

Numerous authors have been addressing Industrial Product-Service System for the huge commercial value and social benefits behind it. Product-Service Systems (PSS) are generally defined as integrated of product and service that allow firms to create new sources of adding value and competitiveness [4]. IPS2 are put forward on the basis of PSS and can be defined as customer life cycle-oriented combinations of products and services to provide higher customer value [5]. Moreover, IPS2 represent as a socio-technical system that consists of several different phases (planning, development, implementation, operation and closure) according to [6]. Delivery is the key aspects for implementation of IPS2 and featured of highly complexity and dynamic. During the delivery phase, IPS2 providers need to

tackle many problems like service network planning, capacity planning and schedule of tasks and resources. Field service has become a major business pillar for most western machine and equipment manufacturers [7]. It is also a main mode for manufacturers to provide services to customers for sustainable competitiveness. To guarantee higher customer value there should have some key performance indicators (KPIs) for evaluation of a PSS delivery process from different dimensions (e.g. process efficiency, customer's satisfaction and environmental sustainability) [8].

Different authors have contributed to the field service delivery regard to Industrial Product-Service Systems. As for equipment manufacturers, field service usually is maintenance service and maintenance policy selection appears to be a difficult decision. Here maintenance strategies can generally be categorized into Preventive Maintenance (PM), Condition Based Maintenance (CBM) and Corrective Maintenance (CM). Most researchers focused on the single maintenance policy and usually is Preventive Maintenance which accounting for 70% according to [9]. But in the actual situation, different strategies usually should be tested at the same field service context in order to explore the optimal policy. However, maintenance service delivery can't be separated from service network. According to [10], customer, asset, employees and some others are configured as core elements of the field service network. In addition, researchers have contributed a lot in the field service network planning from different planning level s(strategic planning , tactical planning, and operational planning) in different approaches[3,11,12]. It is found that the majority of those studies focused on the architecture or process design but lack of operational details in the delivery phase. Meanwhile, a large number of mathematical models and intelligent algorithms has been proposed to solve resource scheduling and task allocation problems in various service industries, for example, aircraft maintenance planning [13], and telecommunications networks [14], etc. Nonetheless, due to the dynamic and customer-oriented features of field service delivery, a more flexible and robust decision-making method for policy selection and capacity planning in the field service delivery of IPS2 is still urgently needed.

2.2. Multi-agent based simulation for field service delivery of IPS2

Recent years field service has drawn more attention from the global equipment manufacturers as it contributes to more than 25 percent of all revenues and up to 40-50 percent of all profits in many companies in this sector [15]. As for equipment manufacturers, field service activities mainly refer to replacement, maintenance and repair service. Although research on optimization for field service has been established decades ago [16], the approach of simulation-based field service in IPS2 area is still in its infancy [17].

Agent based design and application methods are newly emerged paradigms in computer science and artificial intelligence area [18]. In general, during the delivery phase, different decision-making roles like customers, service technicians and others are involved and various delivery processes need to be carried out. In addition, those involved roles are inherently interact on each other along with the

implementation of the field service. Due to those reasons, problems in this area usually too hard for analytical approaches. In this case, multi-agent based simulation as a powerful simulation methods which can model a system from perspective of individual to interactions among the components has received the attention of many scholars and been introduced to solve the capacity planning problem and performance evaluation. The most simulated elements of the field service delivery system in previous researches are service resources and equipment. Moreover, service policy usually can be an input parameter for optimum of the maintenance strategies [9]. Therefore, more efforts still needed in this area for highly case specific and effective simulation model of the field service delivery.

Our research is primarily motivated by those two limitations in the above literature. In this paper, we present an agent based decision-making method for the field service delivery and attempt to solve the maintenance policy selection and capacity planning problem during this phase in an IPS2 environment.

3. Problem statement

In this paper, we consider a real-world situation that a service station of the agricultural equipment manufacturer is responsible for offering field service to its customers. Considering the extensive geographical distribution of the equipment, a manufacturer will usually establish several intermediate service stations and assign them in charge of the specific service delivery process. The agricultural equipment manufacturer provides free components and maintenance service by intermediate service stations to customers those who have bought their equipment during the terms of service. Meanwhile, the service station selects its maintenance policy and settings capacity planning with considering some Key Performance Indicators (KPI) like revenue, cost, equipment availability, customer satisfaction, service technician utilization, etc.

Consider a fleet of agricultural equipment units that are distributed geographically within a certain area. Each equipment unit generates revenue while it is working. It, however, sometimes breaks down and has to be repaired or replaced for the key part. Then the corresponding customer will send service requests to the responsible service station. And maintenance also will be scheduled if the maintenance period is due at this time. After that a service station delivers service to its customers timely according to service requests.

In capacity planning of service stations, configuration of service technicians is the critical issue. In general, the service station consists of a number or service technicians that are all in the service station location. When a maintenance request is received by the service station, one of the technicians takes it, drives to the equipment and performs the required work. During the service work, it may turn out that the equipment cannot be repaired and then it will be replaced for the faulty parts. Having finished the work, the service technician may take another request and drive to the next unit location. The service technician can also replace aged part even if it is still working, subject to the maintenance policy. A service

technician has constant daily cost associated with him. Each operation (maintenance, repair, or replacement) has a different additional one-time cost.

4. Approach to field service delivery of IPS2

In this section, agent based modelling approach of field service delivery for IPS2 is illustrated. The goal of the model is to find the proper capacity and the maintenance policy that bring the maximum profit of the equipment fleet and customer satisfaction in the service delivery phase.

4.1. The multi-agent based simulation model

The objective of the simulation-based approach is to design an experimental environment, which is a good representation of actual service organization [1]. To achieve this goal, the simulation model need have the same input variables and reacts to the different inputs in line with the actual situation. Considering these features, multi-agent based model can better express the interactions among these involved roles and simulate the dynamic characteristics in the actual service delivery context compared with system dynamic modelling and discrete event modelling.

Therefore we adopted an agent-based simulation model here to solve the problem mentioned in the last section. To map the problem into model clearly, the agents here consist of equipment and service technician which are consistent with the actual delivery context. Moreover, service station as the dispatcher and other geographical information will be described at the top level of our simulation model.

The simulation model will be described in detail in the following sections. Our model is built in the simulation software Anylogic Professional 7.0.2.

4.2. Equipment agents

Equipment is the core physical entity of field service as well as the important carrier of a company to create value. Therefore we use a population agent represents a fleet of equipment. Narrowing our vision to the process of field service delivery, the behaviour of equipment will be working, failed, maintenance and so on. And in the process of field service, the state of the equipment changes constantly with the communication and interaction with external environment and other agents in the simulation model. Certainly, at the beginning of the building model, we give initial value of some necessary parameters obtained by cooperation with manufacturer and service stations, for example, failure rate, maintenance period, repair time and so on. And for confidentiality reasons, specific information about the case company will not be disclosed here.

On the basis of above analysis, we have established the state diagram of the equipment that can be observed in Fig. 1. Equipment can be in working or not working definitely. While the equipment is not working, it may be in failure and send service requests to the service station. Then the equipment will wait for service technicians to fix it and it may be repaired or replaced, or be on scheduled maintenance. To

be specific, when an equipment be repaired or replaced completely, it will get maintenance if the maintenance period is due. Therefore here we adopted different service types to the same simulation model which will make our model more appropriate to the actual maintenance policy as well as reduce the service cost by reducing the round trips of technicians.

Finish Rep lacement


FinîshMaintenance —©-s


^£TAr rived For M tee ( Maintenance



£T Arrived For Rep.

_ | FinishRepair 9 J" Repair



Fig. 1 state diagram of equipment

The transitions among those different states are triggered by different ways like time-driven, agent's arrival, or based on probability. For example, failure is clearly a stochastic event interval to the equipment, and both repair and replacement start upon the arrival of a service technician.

4.3. Service technician agents

Apart from the space shared by all agents, communications between the equipment and service technician agent should be defined and managed. The initial state of equipment is working and the service technician is idle. Then, an equipment unit fails and generates request to the service station. Here the service station as a dispatcher and notifies to service technicians and assigns tasks immediately. Otherwise, the service requests are assigned to service technicians in order when both two technicians are idle. In addition, communication between different agents is defined by messages in our model. For example, when a service technician arrives at an equipment unit for service, it send a message to equipment to inform its arrival. After that when the technician finishes his work he will send a message to equipment to live the working state as well.

Of particular note is the handling for different types of service requests. We made the repair service requests treated as high priorities because the failed equipment stops bring revenue for the manufacturer. Besides, the equipment may generate two requests in parallel and in order to avoid more than two technicians arrive at the same equipment, we will check if there was another request from the same unit when adding request to the queue. To be accurate, we have designed many functions for the control process in our simulation model. They are both used to implement agent-to-agent and agent-to-environment communication, and to coordinate diverse activities within a single agent.

5. Implementation of agent-based simulation: a case study

5.1. General model configuration

Fig. 2 state diagram of service technician

4.4. Space and agent communications

In our model, service technicians drive around the geographical space for delivery of field service. Therefore we define a space in the model as a service area. Agricultural equipment units are distributed in the service area randomly. Technicians move within the service area as a constant speed and offer service to the specific requests. It is worth mentioning that the technicians are located at service stations initially and return to service station after the requests completed.

To demonstrate the benefits of using the multi-agent based decision-making approach for field service delivery of IPS2, we will present an application. The following case is based on a real industrial case involving an agricultural equipment manufacturer which sells equipment and offer field service to its customers. Due to the confidentiality of sensitive data, involved maintenance parameters of the equipment maintenance are replaced by fictional data. The service delivery process has been configured as displayed in state paradigm of equipment and service technician. Service requests are generated from the failed equipment and sent to service station. The station at the top level of the simulated model accepts service requests and assigns to service technicians according to geographical locations, state and orders in the queue.

The model includes 80 equipment units distributed randomly in an area of 500km2 square (range of the model can also be adjusted depending on the actual demand). There is a service station located in the central of the simulation area. Number of field service technicians can be chosen in our model in order to find the best configuration of a service station which will be described in detail latter. The service requests generated from equipment may be repair or preventive maintenance and the former has a higher priority.

Then our designed model have mapped and implemented with Anylogic Professional 7.0.2.

5.2. Analysis of KPIs for field service delivery of IPS2

To evaluate the model output as well as provide scientific suggestions to the managers for capacity planning and maintenance policy development, we need to introduce some performance indicators to our model. Here KPIs are evaluated to assess the performance of IPS2 provider-service station thus effectiveness of service resources and service costs are what we concerned. Besides, customer satisfaction is worthy of attention in a PSS context and so equipment availability is also being listed as a KPI. Consequently, equipment availability, service technician utilization, cost and revenue of the service station are taken as performance indicators in our model.

Both equipment availability and service technician utilization are expressed as the percentage of working equipment or technician of all agents. Other relevant parameters such as maintenance time and repair time of the equipment, idle time and driving time of the technician are also calculated in our model and displayed. Fig. 3 shows the value of KPIs in the base experiment. It should be noted that in the base experiment case the number of technicians is initial value three, and the maintenance policy is replacement only when it can't be repaired. Cost in our model consists of two parts: one is the daily fixed cost of employing service technicians; the other is the per-operation cost of maintenance, repair or replacement.

Table 1. Value of KPIs with different technician number

2020 i-fGii

■ Iг,г On Maintins

Service crew utilization [annual averages]

2020 t'J-5: дао gagg

■ Idle Driving Working

Cost anil revenue [annual, ¥K[

1Б. 000 JO. 000

2020 Й0 — Cost — E

Fig. 3 value of KPIs in the base experiment

In order to observe the impact of different maintenance policies and technician number on one chart, we set the simulation time to 60 years though the system we simulated will never remain unchanged for 60 years. We adopted three technicians and policy of replacing only when can't be repaired as the base case. Then two contrast simulation are designed and implemented. We first change the service technician number from the initial value 3 to others to observe the changes of above KPIs and the results are shown in Table 1. After that we changed the maintenance policy from initial to "replacement after several maintenance periods" and the results are show Fig.4

Technician Number Equipment availability technician utilization Profit

3(Base Case) 95% 100% 10500

1 20% 100% 1500

2 60% 100% 7500

4 98% 95% 9900

5 99% 87.50% 9000

The technician's number is changed from one to five and according performance values are recorded in Table 1. The value of KPIs in the base experiment case is shown in the first line of Table 1. It can be found from Table 1 that when technician number=3 the whole efficiency of the service station reaches its maximum. Besides, fewer service technicians will lead to lower equipment utilization but reducing the profit of service station; whereas, more technicians will increase the service cost though equipment availability will also be improved. Obviously IPS2 providers can do capacity planning reasonably with our model.

Equipment availability [annual averages]


2020 2025 2030 2035 ■ Working B On Maintenance

Service crew utilization [annual averages]

■-■A: On Г.;

■ Failed

2J20 £025 2035

■ Idle ■ Driving Wording Cost and revenue [annual, If K]

ooo pi 000 5. 000

— Cost —Rever

2030 лГ-л--■ —f г 0 £ i t

Fig. 4 value of KPIs with different maintenance policies

After that we designed another set of experiments and observed the performance value changes along with the service policy. The service technician number is three and remains unchanged in this group of experiment. At the beginning of the experiment, we take "replacement after two maintenance period" as the maintenance policy and the corresponding statistical data are displayed in Fig. 4 from the simulation time 2015 to 2035. Then we change the replacement after 5 maintenance periods at 2035 and the performance value has a significant change. Especially service costs dropped significantly and the profit nearly increased nearly by one third. Afterwards at the simulation time of 2055 we change the replacement period to 8 maintenance periods whereas the value of performance indicators didn't have a marked change. There's no doubt that this set of experiments is of great significance for field service providers to develop a suitable maintenance policy.

6. Conclusions and future research

In this paper, a multi-agent based decision-making approach for field service delivery has been presented which can provide multiple suggestions to the decision-makers of service providers for capacity planning and field service delivery policy development. The approach has been implemented as a decision-making tool for service providers and the performance diagrams in this paper are clear and detailed enough which can be a scientific and simple guidance for decision-makers. Furthermore, different service resources and maintenance policies can be input in our approach and which will support for different scenarios and provide multiple suggestions to the decision-makers of service providers. Therefore our approach enables IPS2 providers to test and evaluate the effect of different capacity planning schemes as well as maintenance policies in advance which will reduce the decision risks and costs. Moreover, compared to static approaches, the presented approach in this paper offers significant advantages by taking dynamics, stochastics and complexity of actual field service delivery system into consideration. All those make our approach more comprehensive and feasible to the decision-making of service delivery in the IPS2 context.

However, it is still necessary to search for a more appropriate and flexible approach for decision-making of service delivery. This article is not perfect for the service network construction. Therefore we should introduce more elements of actual service network to our model, for example, top managers, horizontal service providers, spare parts managers and so on. In addition, state of technicians in the service work should be more detailed including training and laid off. And more attributes of service technicians should be added such as skill level, working time, and some others. Generally speaking, more efforts should be made toward to facilitate field service delivery of IPS2 and this will be the focus of our future studies.


This research is supported by the National High-Tech. R&D Program, China (Project No. 2013AA040402). We express our sincere thanks to the Lovol International Heavy Industry Co., Ltd.


[2] Lagemann H, Meier H. Robust capacity planning for the delivery of Industrial Product-Service Systems[J]. Procedia CIRP, 2014, 19: 99-104.

[3] Hertz P, Cavalieri S, Finke G R, et al. A simulation-based decision support system for industrial field service network planning[J]. Simulation, 2013: 0037549713512685.

[4] Tukker A. Eight types of product-service system: eight ways to sustainability? Experiences from SusProNet[J]. Business policy and the environment, 2004, 13(4): 246-260.

[5] Meier H, Völker O, Funke B. Industrial product-service systems (IPS2)[J]. The International Journal of Advanced Manufacturing Technology, 2011, 52(9-12): 1175-1191.

[6] Meier H, Steven M, Funke B, et al. Complexity and Flexibility of IPS2 across various Planning Levels[M]//Functional Thinking for Value Creation. Springer Berlin Heidelberg, 2011: 315-319.

[7] Hertz P N A E. Industrial field service network planning[D]. Diss., Eidgenössische Technische Hochschule ETH Zürich, Nr. 21462, 2013, 2013.

[8] Rondini A, Tornese F, Gnoni M G, et al. Business Process Simulation for the Design of Sustainable Product Service Systems (PSS)[M]//Advances in Production Management Systems: Innovative Production Management Towards Sustainable Growth. Springer International Publishing, 2015: 646-653.

[9] Alrabghi A, Tiwari A. State of the art in simulation-based optimisation for maintenance systems[J]. Computers & Industrial Engineering, 2015, 82: 167-182.

[10] Finke G R, Hertz P, Verhasselt S. Categorization and parameterization of network elements in after-sales field services[C]//Proceedings of the APMS 2011 international conference-advances in production management systems, Stavanger, Norway. 2011.

[11] Hertz P, Finke G R, Verhasselt S. A Process Reference Model for Service Delivery in After-Sales Field Service Networks[C]//APMS 2011 International Conference-Advances in Production Management Systems. 2011.

[12] Lagemann H, Boßlau M, Meier H. The Influence of Dynamic Business Models on IPS 2 Network Planning-An Agent-Based Simulation Approach[J]. Procedia CIRP, 2015, 30: 102-107.

[13] Papakostas N, Papachatzakis P, Xanthakis V, et al. An approach to operational aircraft maintenance planning[J]. Decision Support Systems, 2010, 48(4): 604-612.

[14] Chan F T S, Zhang T. The impact of Collaborative Transportation Management on supply chain performance: A simulation approach . Expert Systems with Applications. 2011, 38(3): 2319-2329.


[16] Dekker R. Applications of maintenance optimization models: a review and analysis[J]. Reliability Engineering & System Safety, 1996, 51(3): 229-240.

[17] Meier H, Boßlau M. Design and engineering of dynamic business models for industrial product-service systems[M]//The Philosopher's Stone for Sustainability. Springer Berlin Heidelberg, 2013: 179-184.

[18] Baykasoglu, A. and V. Kaplanoglu, An application oriented multi-agent based approach to dynamic load/truck planning[J]. Expert Systems with Applications, 2015. 42(15-16): p. 6008-6025.

[1] Goossens A J M, Basten R J I. Exploring maintenance policy selection

using the Analytic Hierarchy Process; an application for naval ships[J]. Reliability Engineering & System Safety, 2015, 142: 31-41.