Machine Learning-Based Optimizer for Tilting Pad Journal Bearing Inlet Flowrate

Abstract

Abstract In view of the green economy and energy transition, the reduction of the environmental impact of the power generation sector plays a key role. Fluid film bearings are the most common bearing for industrial turbomachinery and new design requirements have a direct impact on bearings operation. In fact, to achieve higher levels of efficiency, bearings must support higher specific loads and higher peripheral speeds. Furthermore, there is great interest in reducing the oil flowrate required for the bearing operation as much as possible. In this work, an optimization strategy for reducing the flowrate fed to tilting pad journal bearings (TPJBs) is proposed. An artificial neural network (ANN) is trained to estimate the static and dynamic performance of the bearings. The training dataset is built with a Reynolds-based thermo-hydrodynamic model. The trained ANN is then used in a constrained optimization that has the goal of minimizing the oil flowrate while ensuring safe bearing operation. Predictions are compared with experimental data from compressor mechanical running tests. The proposed model is an effective tool that can help industry achieve the goals required by the energy transition and can help in the development of optimized fluid film bearings.