Optimization of high-level preventive maintenance scheduling for high-speed trains

• Published in: Reliability engineering & system safety Vol. 183; pp. 261 - 275
• Main Authors: Lin, Boliang, Wu, Jianping, Lin, Ruixi, Wang, Jiaxi, Wang, Hui, Zhang, Xuhui
• Format: Journal Article
• Language: English
• Published: Barking Elsevier Ltd 01.03.2019; Elsevier BV
• Subjects: 0-1 programming; Case studies; Computer simulation; EMU train; High speed rail; High-level maintenance planning; High-speed railway; Levels; Linearization; Locomotives; Maintenance management; Optimization; Preventive maintenance; Railroads; Reliability engineering; Scheduling; Sensitivity analysis; Simulated annealing; Trains; High-speed railway; High-level maintenance planning; EMU train; 0-1 programming; Simulated annealing
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2018.11.028

•A non-linear piecewise state function is designed to describe maintenance states.•A high-level maintenance planning model for high-speed trains is formulated.•The original model is linearized using a big M method.•Develop a simulated annealing algorithm to solve the relaxation model.•Conduct a real-world case study with 124 trains from the China Shanghai Railroad. For safety reasons, a high-speed train needs to carry out the preventive maintenance when its accumulated running mileage or time reaches a predefined threshold. This paper formulates the train high-level preventive maintenance planning problem as a 0-1 programming model. A novel state function is designed to describe whether a train is under maintenance or not. By using this function, the constraint for restricting the total number of trains under maintenance can be formulated reasonably well. A linearization technique is also employed to refine the original non-linear state function into a linear one. To handle large-scale instances, a simulated annealing algorithm is proposed for solving the problem and is applied to a real-world case study from Shanghai Railway Bureau (SRB), a regional railway operator under China Railway. The optimized results yield a total cost of 3,619,200 standard train-km in terms of remaining mileage. In addition, sensitivity analyses based on the real-world case study reveal some interesting insights. We have delivered the results to SRB as a useful and efficient decision support tool for their high-level maintenance planning work.