Application of smart models for prediction of the frost layer thickness on vertical cryogenic surfaces under natural convection

摘要: Abstract Frost layer growth on cryogenic surfaces is an important topic, and accurate prediction of the frost can be helpful for research heat mass transfer. Implementation artificial intelligence techniques making estimates useful worthwhile, as they are simple-to-use, reliable, cheap. In this study, four models, such multiple linear regression (MLR), neural network (ANN), least squares support vector machine (LSSVM), adaptive neuro fuzzy inference system (ANFIS), were developed to estimate thickness, δ, vertical surfaces. The inputs models surface temperature, Tw; air Ta; relative humidity, φ; time, t. To develop a data set including 711 points was gathered from literature randomly split into two groups: 498 samples train 213 test robustness capability models. evaluate performance aforementioned comparison carried out between results obtained actual measured in laboratory with different graphical statistical error analyses. ANFIS model found outperform other For model, tests R2 MSE gave values 0.9966996032 0.02329202, respectively, testing set. Additionally, indicate apply suggested new condition studied.