Improvement of genetic programming symbolic regression and its application in heat exchangers

In the present study, symbolic regression is used to analyze the experimental data to find a more accurate correlation for heat transfer rate of heat exchangers. In order to improve the performance of genetic programming toolbox GPLAB of MATLAB environment, four new function modules are added into GPLAB, including: structure simplification module; constants optimization module; expansion rate reduction module with "self-swap" genetic operator; small term search intensity enhancement module with "intro-new" genetic operator. The modified symbolic regression is then used in the experimental data reduction process for shell-and-tube heat exchanger with continuous helical baffles. The correlations obtained have higher predictive accuracy and are less sensitive to the disturbance variation of the arguments.