A novel ML-assisted design and optimization of water-lubricated herringbone grooved journal bearings

Purpose The purpose of this study is to present an integrated framework for the design, analysis and machine learning (ML)–based optimization of water-lubricated herringbone grooved journal bearings (HGJBs) for underwater vehicle rotors operating at speeds up to 3,000 rpm under radial loading. Design/methodology/approach A numerical model is developed by solving the nonlinear incompressible Reynolds equation using the central finite difference method to evaluate key performance metrics such as load-carrying capacity, leakage rate and frictional torque. Artificial Neural Networks (ANNs) and Adaptive Neuro-Fuzzy Inference Systems (ANFISs) are employed to predict and optimize groove parameters – groove angle (Ga), groove depth (Gd) and groove number (Ng). ANN facilitates precise performance prediction, while ANFIS identifies optimal groove configurations and operating conditions. Findings ML-based optimization significantly enhances the static performance of HGJBs compared to conventional plain journal bearings. The study demonstrates that integrating numerical modeling with ML improves prediction accuracy and streamlines multiobjective optimization. Originality/value This work systematically integrates ML with numerical modeling to enhance hydrodynamic performance prediction and optimization of water-lubricated HGJBs. The proposed methodology provides a scalable approach for improving bearing performance, contributing to advancements in marine and underwater engineering.