Comparison of Neural Networks and Fuzzy System for Estimation of Heat Transfer Between Contacting Surfaces

Abstract

Neural networks can be used in various subjects, such as the discovery of relationships, identification, system modelling, optimization and nonlinear pattern recognition. One of the interesting applications of this algorithm is heat transfer estimation between contacting surfaces. In the current investigation, a comparison study is done for temperature transfer function estimation between contacting surfaces using Group Method of Data Handling (GMDH) neural networks and ANFIS (Adaptive Neuro Fuzzy Inference System) algorithm. Different algorithms are trained and tested by means of input–output data set taken from the experimental study and the inverse solution using the Conjugate Gradient Method (CGM) with the adjoint problem. Eventually, the optimal model has been chosen based on the common error criteria of root mean square error. According to the obtained results among different models, ANFIS with gaussmf membership function has the best algorithm for identification of TCC between two contacting surfaces with 0.1283 error. Also, the inverse method has the lowest error for thermal contact conductance estimation between fixed contacting surfaces with root mean square error of 0.211.