A Comparative Analysis of Hybrid Computational Models Constructed with Swarm Intelligence Algorithms for Estimating Soil Compression Index

Abstract

The determination of the compression index (Cc) of clay through oedometer tests is time-consuming and expensive. To replace the practice of conducting laboratory oedometer tests, this study presents a comparative analysis of hybrid machine learning models for estimating the soil Cc based on actual laboratory test data. Ten swarm intelligence algorithms, namely particle swarm optimization, ant colony optimization, artificial bee colony, grey wolf optimizer, moth flame optimizer, whale optimization algorithm, salp swarm algorithm, Harris hawks optimization, slime mould algorithm, and marine predator algorithm, were used to optimize the learning parameters of an artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS). Subsequently, 20 hybrid ANN and ANFIS models were constructed to obtain the best prediction model. A sum of 700 oedometer test data was acquired from an Indian Railways project to construct and validate the hybrid models. Besides, 30 new oedometer experiments were performed for external validation of the developed hybrid models. Experimental outcomes show that the proposed ANFIS and PSO hybrid model (ANFIS-PSO) attained the most accurate prediction of soil Cc, which is much superior to the developed hybrid ANN and ANFIS models. Based on the experimental results, the proposed ANFIS-PSO model demonstrates high potential as a robust alternative to the actual oedometer test to assist geotechnical engineers in the introductory stage of civil engineering projects.