Flow-induced vibration and noise in automobile refrigeration system and flow pattern identification under different environmental thermal loads by multilayer perceptron

Refrigerant-induced vibration and noise near thermal expansive valve (TXV) in automobile air conditioning has gradually disturbed the driver and negatively affected the passenger experience. Nevertheless, the effect of environmental thermal load on vibration and noise is unclear and flow pattern identification of refrigeration needs to be accurate but low cost. In this paper, an experimental setup is conducted so as to investigate the effect of thermal load on flow patterns, vibration, and noise near TXV. Both vibration and noise are closely related to the flow pattern conversion during the starting process under different thermal loads. Besides, the vibration characteristics are analyzed on the frequency spectrum at different positions and axes. Increasing thermal load can raise the amplitude of vibration peaks but almost not affect the peak distribution. Finally, according to the connection of time-varying vibration signals and flow patterns, statistical features of triaxial vibration are used to train multilayer perceptron to identify flow patterns in TXV-inlet tube and TXV-outlet tube. The highest average accuracy for all flow patterns is found to be 99.8% (inlet) and 98.2% (outlet) for the test set. Besides, the specific accuracy for each flow pattern is found to be above 99% (inlet) and 96% (outlet), illustrating the model is capable of identifying flow patterns with few samples. The ensemble features of triaxial vibration can more accurately reflect the flow pattern transition compared with single-axis vibration. As seen in the above results, the specific flow patterns inducing larger vibrations can be avoided by adjusting thermodynamic parameters and redesigning refrigeration equipment in future designs.