Probabilistic slope stability analysis of Heavy-haul freight corridor using a hybrid machine learning paradigm

Abstract

With the rising freight demand, specialized heavy-haul railway corridors allow heavier trains to transport heavy freight, improving productivity and lowering unit costs. Generally, a heavy-haul corridor necessitates a significant investment, thus the risk assessment of a rail-track system must be extensively evaluated during the design phase. From the standpoint of serviceability, this study presents a probabilistic slope stability analysis of heavy-haul freight corridor using an efficient hybrid computational technique. The present approach, i.e., ANN-MPA, is an amalgamation of an artificial neural network (ANN) and marine predators algorithm (MPA). The newly constructed ANN-MPA was used to perform probabilistic analysis of a 12.293 m high embankment of heavy-haul freight corridor of Indian Railways with a design axle load of 32.5 MT. The concept of probability theory and statistics were used to map the soil uncertainties through the first-order second-moment method. The results of the proposed ANN-MPA model were evaluated and compared with other hybrid ANNs constructed with seven distinct swarm intelligence algorithms. In the validation phase, the proposed ANN-MPA outperformed (R2 = 0.9931 and RMSE = 0.0233) other hybrid ANNs and was used to perform probabilistic analysis of a 12.293 m high embankment. The reliability index and the probability of failure were computed under seismic and non-seismic conditions, taking into consideration the influence of uncertainties in soil parameters. Using the proposed approach, the failure probability of the 12.293 m high soil slope under different seismic conditions can be evaluated rationally and efficiently.