Development of Reduced Order Model for Performance Prediction of Reciprocating Compressor

Abstract

The reduced order model (ROM) is one of the methods for quickly monitoring the response results when operating conditions change in complex systems. In the conventional method, human and numerical costs are incurred in pre-treatment, interpretation, post-processing to monitor the results of a response under a specific operating condition using numerical analysis. A simplified model is developed that faithfully reproduces higher-order models while lowering the degree of freedom (DOF) of complex systems. Training data should be collected for ROM configuration to predict the performance of the compressor. After setting the operating conditions inside and outside the actual operating area of the compressor as the analysis conditions of the 3D Fluid–Structure Interaction simulation results are collected as training data. The ROM for predicting the performance of the compressor is generated based on the training data. Although a large amount of data is required for high accuracy ROM generation, simulation can only obtain a small amount of training data with a long analysis time. Artificial Neural Network Method complements small amounts of training data by extending to large amounts of data. This paper is a study to establish a methodology for making ROM for predicting the performance of reciprocating compressor. The accuracy of the ROM is verified by comparing it with the experimental data.