AI techniques for evaluating misaligned journal bearing performance: An approach beyond the Sommerfeld number

Abstract

This paper presents two novel artificial intelligence-based approaches for evaluating the performance of heavily loaded marine journal bearings including shaft misalignment. Traditionally, the Sommerfeld number has been used as a key parameter to evaluate the performance similarity between different bearings. However, this method has limitations, particularly when dealing with complex elastic geometries, heavily loaded journal bearings and shaft misalignment. The first proposed approach leverages neural networks to analyze key bearing performance parameters and provide a more accurate and comprehensive assessment of bearing performance similarity, including additional parameters beyond the Sommerfeld number limitations. The second method utilizes artificial intelligence convolutional networks to assess the bearing similarity based on their simulated pressure profiles under isoviscous and isothermal hydrodynamic lubrication regime. The effectiveness of the proposed methods is demonstrated and compared to the traditional Sommerfeld number method, discussing various potential applications and extensions of this concept.