Share

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

Author(s) : ZENDEHBOUDI A., WANG B., LI X.

Type of article: Article

Summary

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

Details

  • Original title: Application of smart models for prediction of the frost layer thickness on vertical cryogenic surfaces under natural convection.
  • Record ID : 30021185
  • Languages: English
  • Source: Applied Thermal Engineering - vol. 115
  • Publication date: 2017/03/25
  • DOI: http://dx.doi.org/10.1016/j.applthermaleng.2017.01.049

Links


See other articles in this issue (31)
See the source

Indexing