A Case-based Reasoning Model For Construction Desigh Of Main Railway Infrastructure In Perilous And Mountainous Areas

The area where Sichuan-Tibet Railway passes through is a typical mountainous and perilous area with high altitude,large elevation difference and extremely complex geological conditions,and thus its route is difficult to spread.Many regions can be passed only through long deep-buried tunnels and long-span bridges.These tunnel and bridge projects usually cost large investment and last long time span while calling for large work volume,therefore they are defined as main infrastructure.However,it is difficult to understand the geological conditions of construction sites fully and completely by current survey technology.In this case,it is inevitable that the most common experience-oriented decision-making method for conceptual design of main projects leads to design errors due to the subjective factor of designers as well as the complexity and uncertainty of geological conditions,which may end up in casualties and economic losses.As a result,new decisionmaking techniques are required to guide the conceptual design of railway main infrastructures in mountainous and perilous areas.Case-based Reasoning(CBR)is a branch of artificial intelligence.It can be adopted to solve some existing problems quickly and effectively according to some historical similar issues.The process of solving problems is consistent with expert decision-making ideas,and the experience and knowledge contained in historical cases can be made use of effectively.For decades,plenty of successful cases have been accumulated in construction practice of railway networks in China,which provides the possibility for the application of CBR.Therefore,an aided decision-making model for conceptual design of main railway infrastructures(tunnels and bridges)in mountainous and perilous areas is proposed based on CBR in this paper.The main research content is as followed:(1)Basic theories of CBR was discussed,including advantages and characteristics of CBR and the connotation and function of related key technologies.The applicability of CBR in railway engineering is demonstrated by combining the characteristics of railway engineering in mountainous and perilous areas as well as some problems existing in engineering practice.(2)Through theoretical analysis and literature study,the category of feature attributes were summarized its quantification criteria.On this basis,tunnel and bridge cases are structurally and quantificationally represented by frame representation method.(3)A similar case retrieval method of CBR model was proposed.Firstly,a method to calculate the weight of railway main infrastructure feature attributes is proposed based on Cloud Theory.Afterwards,grey-KNN algorithm based on relative distance among case vectors is proposed to retrieve similar cases.(4)A case reuse method based on PSO-BP neural network was proposed.Firstly,analysed and summarized shortcomings of typical BP neural network,followed by demonstrating the necessity of adopting PSO algorithm to optimize BP neural network.On this basis,the weight and threshold value of BP algorithm are optimized by making use of the global optimization ability of PSO algorithm,which improves the convergence speed,learning ability and reasoning performance of the neural network.Finally,the flow of case reuse module was designed.(5)A case study based on a vital project in Sichuan-Tibet Railway,Erlangshan Tunnel was carried out.39 tunnel cases in Southwest mountainous areas of China were collected to establish a case base and further applied CBR model proposed in this study to reason the excavation scheme for conceptual design of Erlangshan tunnel,resulting in excellent effect.