Optimization of Variable-Depth Pocketed and Bionic Textured Pad Thrust Bearing Using Machine Learning Models and Genetic Algorithm

Abstract Surface texturing is employed to improve the performance behaviors of fluid-film bearings. In this article, an attempt has been made to investigate the influence of different types of textures on the pad surface. Three configurations have been considered, namely, a bionic texture inspired by a honeycomb structure, a variable-depth pocket, and a combination of bionic texture and variable-depth pocket. Performance parameters, namely, minimum film thickness and coefficient of friction, have been computed at different operating conditions. Due to the large number of possible cases, the prediction of performance behaviors has been made using different machine learning models. Lastly, a genetic algorithm is employed to find the optimized configuration at different operating conditions with fitness evaluation based on the prediction model. Optimized texture/pocket yields substantial improvement in minimum film thickness and considerable reduction in coefficient of friction. The optimized values are then compared with computed values, yielding less than 4% error, and verified with the help of a pin-on-disc apparatus.