A neural network approach to establish the minimum film thickness in a multipad adjustable bearing

Real-time fluid film modification technique is used in the active bearing technology to control the journal's equilibrium position for various operating conditions. This is achieved by the controlled movement of the flexible bearing element. The information of the minimum film thickness (MFT) is essential during the bearing element displacement to avoid the metal to metal contact. The bearing element-specific MFT develops in the multipad externally adjustable bearing with asymmetric combinations of adjustments. A single analytical expression fails to predict the multiple minimum film thickness for various bearing elements adjustment configurations. In this work, a neural network model is developed to predict the MFT in a multipad adjustable bearing having four adjustable bearing elements. Design of experiment and transformation technique is used to collect the data set between the eccentricity ratio 0.1 to 0.8. A feed-forward multilayer perceptron is used to model the MFT. The results show that an accurate estimation of the MFT is possible using a single neural network model.