Research On Shunting Scheduling Of EMU First-Level Maintenance In A Stub-end Depot

After running for a period of time or mileage,an electric multiple units(EMU)needs to enter a depot for maintenance.The efficient EMU maintenance is the key for the smooth implementation of high-speed railway network.In recent years,shunting scheduling of EMU first-level maintenance(SSEFM)has been widely concerned,and it is becoming one of the important topics in the field of railway system scheduling and optimization.Considering that there are many forms of actual maintenance operations,three kinds of problems are gradually studied in this paper,including the multiple constraints-based SSEFM problem(MC-SSEFM),the MC-SSEFM with variable trains,and the MC-SSEFM with variable trains under stochastic uncertainties.Finally the effectiveness of the proposed models and algorithms are verified with case studies.As there are multiple constraints such as time,resource,layout,route,and conflict constraints in the SSEFM,it is necessary to study the MC-SSEFM to establish a mixed integer linear programming(MILP)model with the goal of maximizing the reservation time in the storage area and minimizing the overstay time in the cleaning and inspection areas.Among them,the maximization goal may promote an EMU to complete all maintenance tasks as soon as possible and stay in the storage area in advance to prepare for leaving the depot.The minimization goal may improve the effective utilization rate of operation tracks to ensure the depot more efficient.Secondly,a three-level constructive heuristic(TCH),including operation track allocation,bottleneck track allocation and track occupancy repair,is proposed on the basis of a given train and route sequence.To test the performance of the proposed TCH,experimental results show that the proposed TCH performs the best compared with the CLPEX solver.Since there are variable marshalling in maintained EMUs,it is necessary to consider the EMU variable marshalling constraint on the basis of the above problem to study the MC-SSEFM with variable trains,so as to ensure the appropriate maintenance routes and tracks are selected according to the marshalling situation.Secondly,a two-segment encoding method based on the train and route sequence,and a decoding framework based on the above TCH are designed.The crossover and mutation operators are integrated to improve the ability of exploration and exploitation,and an enhanced particle swarm optimization(EPSO)algorithm is proposed.To test the performance of the proposed EPSO,experimental results show that the proposed EPSO performs the best compared with the CLPEX solver and other six state-of-the-art algorithms.Due to the uncertainties in processing/arrival/departure times,it is necessary to study the MC-SSEFM with variable trains under stochastic uncertainties on the basis of the above problem to obtain a robust schedule.Firstly,two uncertain parameters are introduced to describe the disturbance degree of random variables and the allowable violation degree of constraints respectively,a general robust optimization method for general linear programming problems when random variables obey a certain probability distribution is studied.Then,this method is used to transform the stochastic MILP model to be a deterministic robust counterpart model that can be solved.Finally a three-segment encoding method based on the train,route and track sequence and a decoding strategy based on the above TCH are designed.By embedding heuristic rules for initialization,an adaptive iterative local search(AILS)algorithm is proposed.The proposed AILS includes problem-specific neighborhood structures,a variable neighborhood descent method,and an adaptive perturbation mechanism.To test the performance of the proposed AILS,experimental results show that the proposed AILS performs the best compared with the CLPEX solver and other six state-of-the-art algorithms.Taking an actual EMU operating depot as an industrial case,three kinds of SSEFM problems are studied by applying the actual data of first-level maintenance from this depot to the above MILP models and proposed algorithms.Experimental results based on case study show that the proposed models and algorithms effectively avoid shunting conflicts to generate a feasible schedule that is applicable to three scenarios with reducing maintenance time and improving maintenance efficiency.This demonstrates that the proposed models and algorithms have an engineering value and application prospects in the field of stabilizing and optimizing the SSEFM.