Simulation of Misaligned Journal Bearings Using Neural Networks

Utilization of smart systems, i.e. software tools that incorporate artificial intelligence (AI), in engineering applications increases. This fact is due to their ability to study the performance of complicated systems, producing results quicker and easier than typical analytical models. This article is focused on the advantages of using Artificial Neural Networks (ANNs) to solve the problem of a misaligned hydrodynamic journal bearing. Firstly, the Reynolds equation is solved using finite difference method (FDM) for different operating and misalignment conditions. The results are used to train four (4) artificial neural networks, one for each design parameter. Afterwards, the networks are tested for several operational characteristics and compared with the results of the finite difference method. The outcome is that the force and the torque can be predicted with maximum error of approximately 5% with less computational cost than the finite difference method.