Compressor model, Uncertainty, Semi-empirical, Interval score, Extrapolation
There are various empirical and semi-empirical models to estimate the performance of compressors. While some studies support the use of empirical methods for their high accuracy with the available experimental observations of compressor performance, others claim that semi-empirical methods can estimate the performance at extrapolation conditions more reliably. To understand if these claims are true, quantitative comparison of various types of compressor models was conducted based on uncertainty analysis. But these methods evaluate the accuracy and uncertainty of the models separately, and it is difficult for model users to comprehend their results. In this paper, the comparison of model accuracy and model uncertainty is combined together using a scoring method for probabilistic forecasting methods called interval score. The calculation of interval scores is widely used in the meteorological field to compare the accuracy of weather models with uncertainty outputs. This study uses interval scores of different compressor mass flow rate models to compare the performance five different empirical and semi-empirical compressor models with data from two compressors. The results of the comparison show that the most reliable model is a model that does not use redundant empirical coefficients nor physical principles, but even the most reliable model may fail to explain the compressor performance well under extrapolation.