A Decision-making approach to field service delivery ... - IEEE Xplore

A Decision-making approach to field service delivery under mixed maintenance policy Rui Zhou, Yaoguang Hu, Shasha Xiao, Jingqian Wen School of Mechanical Engineering

Beijing Institute of Technology Beijing, China [email protected],[email protected],[email protected], [email protected]

Abstract—Maintenance service delivery for large-scale equipment has been studied for decades and is becoming a crucial way for equipment manufacturers to obtain profit and competitiveness. To delivery maintenance service, equipment manufacturers need to optimal delivery plans with considering of various constraints (e.g. maintenance policy, service capacities, service costs) as well as maintenance type. This study proposed a mathematical model and algorithm-based approach to solving a maintenance service decision-making problem in the mixed maintenance policy context and to assess its performance. The mathematical model was constructed with the constraints of technician number, skill level, maintenance time windows, and maintenance policy. A hybrid algorithm of simulated annealing algorithm (SA) and genetic algorithm (GA) is proposed to solve this practical maintenance service delivery problem. A case study of agricultural equipment maintenance demonstrated that the proposed methods are effective and the algorithms can provide reasonable solutions within an acceptable computational time. Our approach is effective and can be utilized to improve the performance of maintenance service decision-making for the mixed policy. Keywords—Maintenance; Service delivery; Mixed maintenance policy; Decision-making

I.

INTRODUCTION

Maintenance aims to combat the inevitable degradation of equipment over the operational time and keep them in working order [1]. It plays an import role in equipment operational effectiveness and economy. Moreover, with the emergence of the service economy, more and more companies offer service to customers rather than only providing products for higher revenue and stronger competitiveness. As for equipment manufacturers, field maintenance service is the main way of providing service to customers. An effective maintenance service delivery is advantageous to realize the win-win state between manufacturers and customers. However, a variety of constraints need to be considered in the service delivery process, such as maintenance policy, service capacities, service costs, etc. Of all the above constraints, maintenance policy is the fundamental issue which plays a key role in decisionmaking of field service delivery [2]. In general, maintenance strategies can be categorized into Corrective Maintenance (CM), Preventive Maintenance (CM) and Condition Based Maintenance (CBM) according to [1]. CM usually occurs when the equipment breaks down with

c 978-1-4673-8644-9/16/$31.00 2016 IEEE

higher maintenance costs and downtime losses. PM is scheduled to keep the equipment operates in a specified state through systematic inspection, testing, or replacement. Nowadays, with the development of sensor technology and decreasing cost, CBM is drawing more attention from various industries for its avoiding rigidity of traditional PM and lag of CM. Lot of researches have been done for the decision-making of service delivery under different maintenance policies above mentioned. Besides, the majority of early studies on decisionmaking of maintenance service delivery focused on the fixed position equipment and under single maintenance policy [3, 4]. However, there will occur such phenomenon that CBM and CM or PM activities need to be tackled at the same time and the equipment may be with mobility like train, aircraft, etc. In this case, the existing methods seemed to be somewhat powerless. Consequently, in this paper, we deal with the occasion that mixed maintenance policies are employed and a decision-making method for service delivery in this case is presented. The remainder of this paper is structured as follows. Section 2 gives a review of related studies. In section 3, we describe the problem and specify the process of service delivery. Section 4 presents a method to address the problem. A case study is described and the experiment results and discussions are presented as well in section 5. The conclusions and discussions are stated in section 6. II.

LITERATURE REVIEW

Maintenance mode has experienced breakdown maintenance, preventive maintenance, into the conditionbased maintenance which through advanced technology of predicting fault, condition monitoring, fault diagnosis, life prediction and reliability assessment, repair each equipment before a failure occurs as far as possible [5]. At the same time, decision-making method of maintenance service delivery has been extensively studied in the past years. Actually, decision-making for effective maintenance service delivery is of great complex for the various constraints (e.g. service cost, skill matching, downtime cost, etc.) and source data (e.g. equipment health degradation profile, condition parameter) which need to be considered concurrently. Researchers in this area have presented many mathematical models and intelligent algorithms to solve the problem. Reference [6] established an optimization model combining

1068

preventive maintenance plan with production scheduling for a single equipment, in order to get the production and maintenance plan which meet the delivery time and minimize the total cost. In [7], a newly decision support system for effective maintenance operations be developed which is helpful for joint production and maintenance scheduling. In general, approaches to joint maintenance scheduling and production have been studied from various aspects in [8, 9, 10]. However, the above presented approaches focused on the fixed equipment and under the single maintenance policy. In a large moving equipment context, existing researches [11, 2] focus on the optimization and evaluation for maintenance policies. In addition, approaches to fault diagnosis and optimization of maintenance period account for a large part of the previous research. As for many large-scale equipment, they are distributed geographically and operating travelling with time. Moreover, corrective maintenance and conditionbased maintenance will occur at the same time aiming at lower traveling costs of technicians. Few literatures have studied on the above mentioned scenario. Consequently, we consider an occasion that mixed maintenance policies are employed and maintenance service delivery method for equipment with mobility is presented in this paper. III.

PROBLEM DEFINITION

In this paper, we consider a real-world situation in which service stations of a manufacturer provide field maintenance service to the customers distributed geographically around the service area. Agricultural machinery manufacturing enterprises provide free components and maintenance service to customers who have bought their equipment during the terms of service. Considering the extensive geographical distribution of the service demand points, a manufacturer will usually establish several intermediate service stations and assign them in charge of the specific service delivery process. A station is in charge of one specific area within the given service resources. Service resources may be technicians, vehicles, tools, spare parts and so on. The capacity planning of service stations is set by manufacturers according to the equipment population of the accountable area. However, impacted by the concentration of agricultural equipment operation time, equipment failures will often outbreak in a short time. This means that a service station needs to tackle a large number maintenance requirements reasonably according to the fault type and resource availability within a limited time. A scientific and practical approach is urgently needed in this decision-making process. The decision-making process involves three roles: the customers, the service station and external systems. In the initial decision-making process, customers experience equipment failure and call the corresponding service station to report the maintenance requirements and provide necessary failure information to the operator at the service station. Decision-makers of the service station aggregate demand information that the system received within a period of time to perfect necessary information about a service requirement from

external systems. Geographic information system (GIS) and fault monitoring system (FMS) are two systems that are primarily applied in an external system to separately obtain positional information and physical data of faulty equipment. After the end of the perfecting process for demand information, decisions should be made by considering a series of constraints, such as maintenance type (corrective maintenance or condition-based maintenance), failure time interval, tasks priority, technician number, skill level, and so on. It can be found that the mentioned problem is complex and what even worse is that decision makers are experiencing in lack of a comprehensive and practical methodology. As a consequence, in most cases decisions are based on experience. Then, the decision scheme, including the arrangements and route planning of technicians, is given and output for implementation. The decision-making of maintenance tasks is completed at this time. IV.

METHOD

According to the research gap and motivation, we consider a case study approach that is appropriate for our research endeavor. The proposed cross-regional collaborative decisionmaking approach consists of modeling and solving in two phases. In this section, we introduce the mathematical model that is employed for maintenance service decision-making and present the modified genetic simulated annealing algorithm. A. Model In the context of mix-mode maintenance service delivery, decisions are made by decision-makers with considering constraints that exist in the practical case. Due to the existing gap between the abstract mathematical model and real-life situation, we made some necessary assumptions listed in following:

• • • • •

Remaining life of the potential failure of equipment are known by failure prediction; The cost per mile ,the number of service stations technician ,and operational plan are known; Maintenance skills of technicians and hourly charge are known by skill matching; The costs of case-based maintenance is less than that of breakdown maintenance. Each requirement is serviced by only one technician. After repair, equipment health status returned to normal.

Before presenting the mathematical notations are given as follows.

model, related

Indices

n∈ N

m∈M yi

xijk

The number of the equipment required for maintenance The number of technicians The import degree coefficient of equipment i Whether arrange technician k to perform maintenance task from i to j, if do the arrangement, it will equal 1, or else will equal 0

2016 IEEE 11th Conference on Industrial Electronics and Applications (ICIEA)

1069

C1 C2 C3 C4 bi d kij hi jki pk g ki

[qi1, qi 2 ] zi Fi

The cost per mile The unit cost for beyond time window Unit cost of technician for waiting Unit cost of downtime for equipment hourly fee of task i Distance of i to j by technician k Standard hour for service of equipment i

Consequently, the problem in this paper can be mathematically stated as follows:

If technician k has the ability to serve equipment i it will be 1,or else it will be ∞ Skill level of technician k and

Min Z = ¦¦¦C d x +¦¦¦ x b +¦W + ¦ D 1 kij ijk ijk i i i

pk ∈ {1, 2,3, 4,5} Work hours of technician k complete the service of equipment i and it equals to hi (1 + ∂pk ) jki ;here ∂ is the capacity coefficient of technician k; and higher skill corresponding to lower ∂ ; Time window of equipment i; as for condition-based maintenance, it is the predictive failure interval; Location set of equipment i during the time window; Idle time of equipment i during the time window, represented as { [ f i11, f i12 ],[ f i 21, f i 22 ], ……}; as for corrective maintenance Fi = Φ

tkia tkib v

time in mixed maintenance policy context as far as possible. What also should be noted is the method for determining the level of importance of the equipment i. It is related to the customer level a, maintenance type b and downtime loss c and can be represented as yi = a * b * c . In addition, we use two matrixes which represent service technician and mission costs, service technician and mission capability respectively as proposed in our previous research [12].

Time of technician k arrive at equipment i Time of technician k leave equipment i

Speed of service vehicle In addition to the above notations, there are two more equations worth to be mentioned.

N

N

M

N

i =1 j =1 k =1

N

M

i =1 j =1 k =1

N

N

i =1

i =1

(1)

s.t M

N

k =1

j =1

¦ ¦

xijk ≤ m, i = 0

¦ ¦ x = 1, j = {1, 2, ⋅⋅⋅, N} ¦ x =¦ x ≤ 1, k = {1, 2, ⋅⋅⋅, M} M

N

k =1 N

i =1 ijk

j =1

N

0 jk

i =1

i0k

xijk (tkib+ dkij / v) ≥ q j1 , j = {1, 2, ⋅⋅⋅, N} xijk (tkja + gkj ) ≤ q j 2 , j = {1, 2, ⋅⋅⋅, N}

(2)

(3) (4) (5) (6)

In this mathematical model, Eq. 1 is the objective function that minimizes the total cost of maintenance service delivery. Eq. 2 to Eq. 6 represent the constraint conditions. Eq. (2) represents the number of dispatched service technicians those who are performing a task outside should be smaller than the service station hired; Eq. (3) represents that each demand equipment can only be reached once by one service technician; Eq. (4) indicates that the service personnel should go back to the service station after completing all tasks he is responsible; Eq. (5) means that technician’s arriving time should be later than the early of failure time window for equipment j; Eq. (6) represents that technician’s leaving time should be earlier than the normal working time of equipment j.

B. Genetic simulated annealing algorithm In this study, we design a new hybrid method that was integrated by Genetic algorithm (GA) and Simulated annealing (SA) and referred to as the GSA to solve the problem. The GSA is an optimization method with a strong performance for optimization and faster convergence rate N M =Φ∪ ≠ Φ∩ > yC x g F (F f t ) based on GA in combination with simulated annealing i 4 ijk ki, i i ia1 kib ° algorithm. Considering our problem, we combine SA and GA ° j=1 k=1 using the idea of a SA’s update method and a GA’s crossover °° N M and mutation mechanism. The algorithm improves local yC x t Di = ® i 4 ijk kib, search capability of genetic algorithm and expands searches ° j=1 k=1 field by accepting some characteristics of deteriorative ° N M solutions under a certain probability, so as to effectively solve − − yC x max{min{( t f ),( t f )},0}, others ° i 4 ijk kib i12 kib ir 2 the premature convergence of genetic algorithm, and °¯ j=1 k=1 overcome the shortcomings which simulation consumes much Wi is the method of calculating waiting cost of technician k time by using SA to achieve equilibrium. The algorithm steps were developed in detail in our previous research. for equipment i ; Di is the method of calculating downtime To distinguish different service routes, each maintenance maintenance cost for equipment i. With these two functions, demand point is represented by a natural number, and zero is we can make arrangements for repair work in equipment idle

Fi = Φ ∪ tkia ∈ Fi ∪ Fi = Working time 0, ° N M Wi = ® ° ¦¦ C3 xijk min{(fi11 − tkia ),… (fir1 − tkia )}, others ¯ j =1 k =1

¦¦ ¦¦

¦¦

1070


taken to separate the routes, which can also be considered to be a service station point. Then we designed a fitness function to evaluate the individual and it can be represented as Fit ( z ) = Cmax − Z . Here Cmax is the biggest estimated

generated randomly. The results of the maintenance routes, the corresponding technician number and the total service delivery cost are listed in Table 2. The operation effect of the algorithm is seen in Fig. 1.

value as the objective function z. V. CASE STUDY AND RESULT ANALYSIS In this section, we employ the approach that was mentioned in the previous section to solve a practical problem. We describe the experimental set and algorithm preferences in agricultural equipment maintenance environments. The results of service delivery cost and routes are presented and analyzed. According to our problem and solution method, a simulation study is performed by Matrix Laboratory. There is one service station which is responsible for equipment maintenance service within the service area. The remaining life of potential equipment, operation plan and location of equipment are known and shown in Table 1. There are five technicians numbered as T1~T5 and the skill matrix is {3, 1, 2, 2, 2}. In addition, we set the downtime loss of equipment is 20/hour, hourly fee of CBM is 10 and the CM hourly fee is 25r. The unit cost of technician for waiting is 5.

It can be found from Fig. 1 that the proposed GSA gets convergence in 140 generations. The first column in Table 2 is the serial number of technician, and the other column is assignment of maintenance tasks and route planning for technicians which from calculated by GSA algorithm. The last line represents the optimal total cost by our method.

TABLE 1. EXPERIMENTAL DATA

TABLE 2. SERVICE DELIVERY RESULT

Customer 1

Mainte nance type CBM

Task prior -ity 1

Standard working hour 1

2

CM

4

2

3

CM

3.2

2.5

4

CM

4

2.5

5

CBM

0.6

1.2

6

CM

4.8

2

7

CM

4

3

8

CBM

2

0.5

9

CM

4

2

10

CM

4

3

11

CM

4

2

12

CBM

3.2

1

13

CM

3.2

2

14

CM

6

2

15

CBM

3.2

1

Position 1

114.7614,35.23 752 114.7612,35.23 651 114.7613,35.23 8 114.7621,35.23 715 114.7617,35.23 707 114.7619,35.23 709 114.7615,35.23 71 114.7618,35.23 72 114.7616,35.23 72 114.7617,35.23 714 114.7620,35.23 719 114.7616,35.23 707 114.7615,35.23 706 114.761,35.237 53 114.762,35.237 34

·Position n

114.7619,35.2 3723 114.7612,35.2 3615 114.7613,35.2 38 114.7621,35.2 3715 114.7617,35.2 3707 114.7619;35.2 3709 114.7615,35.2 371 114.76185,35. 23724 114.7616,35.2 372 114.7617,35.2 3714 114.7620,35.2 3719 114.7616,35.2 3707 114.76152,35. 23706 114.761,35.23 753 114.762,35.23 734

After analysis of the problem, this paper performed many times of simulation experiments and we set the parameters of the algorithm as follows: the number of particles=200 and the number of iterations=200. We set Cmax=10000 to the fitness function. The positions of the crossover and mutation are

Fig. 1.The optimal value iteration of GSA

Technician Number T1 T2 T3 T4

Routes 15Æ11 7-2 5 13Æ12Æ8Æ6Æ9Æ3Æ14

T5 Total cost

10Æ4Æ1 40106

VI. CONCLUSIONS AND FUTURE RESEARCH The purpose of this research is looking for a decisionmaking method for service delivery in mixed maintenance policy to minimize the service cost. The proposed intelligent method resolves the maintenance field service delivery problem under corrective maintenance and condition-based maintenance context. The related data of condition-based maintenance is got from external system employed in the equipment. A hybrid algorithm of genetic algorithm and simulated annealing algorithm is presented and solved our problem efficiently. It can be found from the decision-making results that our presented method is effective for this complex problem and can provide reasonable support for modern maintenance field service delivery. However, additional research is still needed to provide a more profitable and stable approach to maintenance service delivery in the nixed service policy. There are many deficiencies in this article, such as not considering technicians workload which leads to uneven technician assignment and collaborative scheduling service. Therefore, our study will consider more constraints, in order to be more in line with the actual operation of the equipment and improve service level of the whole service system to achieve service-oriented manufacturing. ACKNOWLEDGMENT


1071

This research is funded by the Republic of China’s National High-Technology Project (863) under contact number: 2013AA040402. Thank you for the cooperation of Foton Lovol International Heavy Industry Co., Ltd.

REFERENCES [1]

Alrabghi A, Tiwari A. State of the art in simulation-based optimisation for maintenance systems[J]. Computers & Industrial Engineering, 2015, 82: 167-182. [2] Zio E, Compare M. Evaluating maintenance policies by quantitative modeling and analysis[J]. Reliability Engineering & System Safety, 2013, 109: 53-65. [3] Yao X, Fernández-Gaucherand E, Fu M C, et al. Optimal preventive maintenance scheduling in semiconductor manufacturing[J]. Semiconductor Manufacturing, IEEE Transactions on, 2004, 17(3): 345-356. [4] Huanghui Zong, Hu Xiong, “Periodic preventive maintenance scheduling model for vehicle,” Computer Simulation, 2013,05: 167170,212. [5] Ya-mei, Chen Xuhua, “Model study of maintenance decisions based on the condition,” Science, Technology and Engineering, 2007,7 (4): 586588,593. [6] Fitouhi MC, Nourelfath M, “Integrating noncyclical preventive maintenance scheduling and production planning for a single machine ,”International Journal of Production Economics 2012; 136 (2):.. 344-51 [7] Ni J, Jin X. Decision support systems for effective maintenance operations[J]. CIRP Annals-Manufacturing Technology, 2012, 61(1): 411-414. [8] Luo M, Yan H C, Hu B, et al. A data-driven two-stage maintenance framework for degradation prediction in semiconductor manufacturing industries[J]. Computers & Industrial Engineering, 2015, 85: 414-422. [9] Xiao L, Song S, Chen X, et al. Joint optimization of production scheduling and machine group preventive maintenance[J]. Reliability Engineering & System Safety, 2016, 146: 68-78. [10] Moghaddam K S. Multi-objective preventive maintenance and replacement scheduling in a manufacturing system using goal programming[J]. International Journal of Production Economics, 2013, 146(2): 704-716. [11] Xiang Y, Cassady C R, Pohl E A. Optimal maintenance policies for systems subject to a Markovian operating environment[J]. Computers & Industrial Engineering, 2012, 62(1): 190-197. [12] Li X, Wen J, Zhou R, et al. Study on resource scheduling method of predictive maintenance for equipment based on knowledge [C]. ISKE, 2015

1072
