Research On The Approach Of Multi-objective Intelligent Railway Location Design In Valley Area

Railway location design is the core mission of the railway construction engineering,which has a long-term and important influence on politics,economy,environment and other aspects along the line of the project.With the accumulation of railway construction experience and the improvement of the engineering technology in China,the modern concept of the railway location design has developed into a scientific problem that needs to comprehensively consider the coupling effect of economy,risk,environment,comfort and other factors.Especially in the western mountainous regions of China,complex terrains such as hills and canyons and various adverse geological conditions and hazards occur more and more frequently in the study region,which bring great difficulties to the design work and seriously reduce the efficiency and quality of the railway location planning and design.Therefore,in this paper,an approach of multi-objective intelligent railway location design for canyon regions based on the theory of deep reinforcement learning is proposed.which could consider the influence of economic and geological hazard risk on the railway.It could automatically generate and evaluate the railway location design results,provide more alternative schemes for designers and be of benefit to improve the design efficiency.In this paper,the principle and algorithm of deep reinforcement learning are analyzed and discussed firstly,and the DRL model oriented to railway location design is constructed.The elements of reinforcement learning such as the environment,states,actions,reward functions and constraint conditions are defined.An improved D3 QN algorithm based on priority experience storage mechanism is proposed.Taking the construction cost as the optimization target,the feasibility and effectiveness of the model are verified by experiments.Then,for the familiar geological hazards such as landslide and rockfall,the weight of evidence model is used to evaluate the hazard susceptibility of the target region and analyze the spatial distribution characteristics of hazards and the influence law of various evaluation indexes on the susceptibility.On this basis,combining the susceptibility of space hazard with the vulnerability of structures,the geological hazard risk value is obtained as the evaluation index to quantify the hazard risk of the railway alignment.Finally,combined with the theory of multi-objective optimization and deep reinforcement learning,taking the engineering construction cost and the hazard risk value as the objective function,the economic-risk double objective optimization model is established and solved by envelope MORL algorithm.The results show that the method could search the Pareto frontier of two objectives,take into account both the cost and risk of the railway,generate a better route scheme and provide new design ideas for designers.