چكيده لاتين
Railway tracks, as one of the critical and costly components of railway infrastructure, require precise and efficient management of maintenance and repair operations to ensure system reliability and network availability. This study develops a comprehensive mathematical model for planning of preventive and condition-based maintenance operations for railway tracks. The objective of the model is to minimize the total costs associated with maintenance and repair activities, track possession costs, and the correction of random failures over the planning horizon, while adhering to constraints such as budget, available resources, allowable limits of track quality indices, minimum system reliability and network availability.
To achieve this goal, the railway network is divided into segments, and quality indices, including the standard deviation of the longitudinal level, ballast fouling index, head check crack depth of rails, and lateral rail wear, are calculated for each segment. These indices are monitored throughout the planning period using prediction and recovery models, which depend on the type of maintenance performed in each period. Compliance with the allowable limits of these indices serves as the basis for planning of preventive maintenance. Preventive operations encompass routine maintenance and inspections, tamping, stone blowing, ballast screening, rail grinding, and rail replacement. Additionally, the expected number of random failures per period is estimated using a non-homogeneous Poisson distribution for rail weld failures, with preventive maintenance operations having a mitigating effect on their occurrence.
Given the complexity and nonlinearity of the mathematical model, a hybrid genetic algorithm was developed in MATLAB, utilizing a heuristic method to generate initial solutions and enhance the algorithm’s operators. To evaluate the efficiency of this algorithm, its results were compared with two alternative approaches: first, the direct solution of the nonlinear model in GAMS; and second, the solution of a simplified mixed-integer programming (MIP) model with a homogeneous Poisson approximation in GAMS, followed by the use of its allocation to compute the objective function of the original model. As a case study, data from one kilometer of the Arak railway region were collected and modeled. The results demonstrated that the proposed algorithm consistently yielded lower objective function values compared to the other methods. For small-scale problems, the objective function value of the proposed algorithm matched that of the direct solution in GAMS, but for larger problems, the direct solution’s objective function was up to 63.4% higher than that of the proposed algorithm. Furthermore, the allocation derived from the MIP model produced an objective function value up to 5% higher than the proposed algorithm in the worst case. In terms of computational time, for the largest problem dimension examined in the case study, the proposed algorithm converged in 103 seconds, while the direct solution of the nonlinear model in GAMS required 235 seconds, and the MIP model solution in GAMS took 7204 seconds.
The proposed model enables the planning of preventive maintenance operations by considering track quality indices, reliability and availability requirements, and the mitigating effect of these operations on random failures. The model’s adaptability to section-to-section variation, due to the variability in track structure, traffic, environment, maintenance history, etc. provides significant flexibility in implementing diverse maintenance policies and aligning with the maintenance and repair schedules of required machinery. The proposed algorithm ensures efficient resolution of the model across various problem dimensions with reasonable computational times, making it a robust tool for optimizing railway track maintenance management.
