with the collaboration of Iranian Society of Mechanical Engineers (ISME)
Keywords = CatBoost
Agricultural and off-Road Machinery

Modelling Plate Penetration in Bekker Soil Model Using Machine Learning for Off-Road Vehicles

Articles in Press, Accepted Manuscript, Available Online from 26 October 2025

https://doi.org/10.22067/jam.2025.93736.1386

A. Nazari Chamki, A. Mardani, A. Hosainpour

Abstract The study of soil behaviour in wheel interaction is complex due to the wheel's geometry and the varying soil conditions. Traditional measurements of soil parameters, such as the Bevameter and the cone penetrometer, are time-consuming and labour-intensive. This research presents a machine learning-based approach to predict soil sinkage in plate penetration tests, providing a suitable alternative to conventional methods. A soil bin with controlled experimental conditions was used to collect data, which was measured by a load cell and a magnetic encoder at a constant penetration rate of 4 mm s-1. Two main machine learning models were selected; XGBoost and CatBoost. Hybrid versions of these models were developed using the Shrike Bird Optimisation Algorithm (SBOA). The results showed that the hybrid models outperformed the base models. The SBOA-CatBoost hybrid model achieved the highest accuracy on the training data with a coefficient of determination of 0.99, a mean square error of 2.81, and a mean absolute error of 0.79. The findings of this study highlight the potential of machine learning as a cost-effective and efficient alternative to traditional methods for measuring soil parameters. Further research is recommended to validate these models in different soil types and conditions.