Prediction of the reaction forces of spiral-groove gas journal bearings by artificial neural network regression models

Elia Iseli,Jürg Schiffmann

doi:10.1016/j.jocs.2020.101256

Abstract

This paper presents neural network regression models for predicting the nonlinear static and linearized dynamic reaction forces of spiral grooved gas journal bearings. The partial differential equations (PDEs) are sampled, based on a full factorial and randomly spaced parameter set. Feed-forward neural network (FNN) architectures are developed for modeling the PDEs and therefore replacing the time-consuming discrete and iterative solution procedure used to this date. A significant speed-up factor of >103 in computation time is achieved, compared to solving the PDE numerically. Furthermore, the FNN allows for multi-dimensional interpolation, which makes global system optimization easily possible. This is demonstrated by a real-case rotordynamic system optimization. By using the neural network meta-models, a complete rotordynamic system optimization time reduction of factor 300 is achieved.

Highlights

High-speed small scale turbomachinery is used in many different energy conversion systems such as domestic or commercial heat pumps [20], Organic Rankine cycles [18] or fuel cells [24]
In order to assess the accuracy of the artificial neural network (ANN) compared to the classical finite difference method (FDM) approach in predicting the rotordynamic performance, the system natural frequencies and the corresponding logarithmic decrements are computed for all compressibility numbers and eccentricities with the ANN and the FDM approach for a solution selected on the computed Pareto curve
At computations the speed-up factor levels-off at > for the static force models and > 103 for the dynamic force models. This means a significant reduction in bearing computation time, allowing for large parameter variations in minutes

Summary

Introduction

High-speed small scale turbomachinery is used in many different energy conversion systems such as domestic or commercial heat pumps [20], Organic Rankine cycles [18] or fuel cells [24]. The requirements of high rotational speeds and a high life-time expectation make gas lubricated bearings ideal for these systems. In order to develop stable compressor systems, accurate bearing models need to be available and parameter variations with thousands of computations have to be performed during an automated optimization process. To this date, the performance of HGJB have been predicted by solving the thin-film flow equation, called Reynoldsequation, in different ways. Fleming and Hamrock [4] used the narrow groove theory (NGT) for optimizing HGJB for maximum stability, quantified by ORCID(s):

Methods

Results

Conclusion