ANFIS based evolutionary concept for estimating nucleate pool boiling heat transfer of refrigerant-ester oil containing nanoparticles. (December 2018)

• Record Type: Journal Article
• Authors: Saee, Ali Dehghan; Baghban, Alireza; Zarei, Fariba; Zhang, Zhien; Habibzadeh, Sajjad
• Is Part Of: International journal of refrigeration. Volume 96(2018)
• Page Start: 38
• Page End: 49
• Publication Date: 2018-12
• Subjects: Refrigeration -- Adaptive neuro-fuzzy inference system -- Pool boiling heat transfer -- Outlier detection -- Evolutionary algorithms; Froid -- Système d'inférence neuro-flou adaptatif -- Transfert de chaleur par ébullition libre -- Méthode de détection des valeurs aberrantes -- Algorithmes évolutionnaires; Refrigeration and refrigerating machinery -- Periodicals; 621.56
• DOI: 10.1016/j.ijrefrig.2018.08.002
• Languages: English
• ISSNs: 0140-7007

Abstract

Highlights: ANFIS is developed applying different evolutionary algorithms to predict pool boiling heat transfer. A number of 405 empirical data were obtained from literature to train and test the models. This model is an easy-to-use tool to estimate nucleate pool boiling heat transfer properties of refrigerant-oil mixtures with nanoparticles. Abstract: Performance and improvements of refrigeration systems are greatly related to the nucleate pool boiling heat transfer of refrigerant-oil mixtures containing nanoparticles (h). Empirical correlations have been used previously in order to estimate this parameter. Improved Adaptive Neuro-Fuzzy Inference System (ANFIS) is developed in this research applying different evolutionary algorithms to predict pool boiling heat transfer. To that end, 405 data samples were collected to construct the model and evaluate its performance based on each evolutionary algorithm. PSO-ANFIS model has the most accurate structure compared to other algorithms with R 2 = 0.998 and RMSE = 0.031 for all the 405 data samples. Moreover, the applicable domain of developed models was investigated and the dubious samples in the databank were indicated by using a technique based on Leverage algorithm.