Prediction of heat exchanger performance in cryogenic oscillating flow conditions by support vector machine

Heat transfer characteristics in cryogenic oscillating flow is essential to the development of high-efficiency cryocoolers. In this study, the heat exchanger performance in cryogenic oscillating-flow conditions was experimentally studied. Based on the collected experimental results and the material properties that can affect the targeted space-cycle averaged Nusselt number, machine learning models were established for data processing. It was found that the support vector machine models have distinguishable improvement in predicting accuracy compared with the commonly used non-dimensional correlations in the exponential form. For the two different data sets trained by standard support vector machine and support vector machine with leave-one-out method, the latter achieves better accuracy with the maximum error of 12.4% and the R-square value of 0.922. The results indicate that the support vector machine can be applied as a cost-effective and accurate data processing method for the heat transfer characteristics in cryogenic oscillating flow.

Automated summary

This study experimentally studied the heat exchanger performance in cryogenic oscillating-flow conditions and established machine learning models for data processing. Support vector machine models showed distinguishable improvement in predicting accuracy compared with non-dimensional correlations in the exponential form.