Accurate prediction of frost growth is rather difficult because of its typically strong nonlinear and time-dependent process, and the measured experimental data usually contain many noisy signals. To solve this problem, a novel machine learning method-Support Vector Machine (SVM) based on Structure Risk Minimization principle is introduced to develop models for the prediction, during the rost growth, of frost thickness, total heat flux and frost mass concentration. The predicted results are found to be in good agreement with the measured experimental data, with mean relative error less than 0.62% for the total heat flux, 2.42% for the frost mass concentration, and 5.94% for the frost thickness. Compared with the multivariate nonlinear regression (NLR) model, the SVM models show better capability in solving nonlinear, time-dependent and noise-signal-interfered problem. This demonstrates that the SVM technique can be well used in predicting the frost growth characteristics, and accordingly, help optimize air-to-refrigerant system.
展开▼
