Optimization of thermal performance in closed-loop wickless heat pipes with helical coil evaporators by neuro-genetic approach

• The performance of heat pipes is influenced by complex interactions of parameters. • Traditional methods often overlook the nonlinearity among the parameters. • A neuro-genetic optimization approach is introduced. • The method combines Artificial Neural Networks with a Genetic Algorithm. • The thermal resistance is reduced by 51% compared to the lower bound. The closed-loop wickless heat pipe (CLWHP) is an advanced passive thermal management device that effectively transfers heat through phase change of the working fluid. Its performance is affected by a complicated interaction of design and operational parameters, specifically supplied power, inclination angle, filling ratio, and curvature ratio of the helically coiled evaporator. Traditional empirical approaches often fail to capture the nonlinear, multidimensional relationships among these parameters, thereby limiting design optimization. This study introduces a neuro-genetic approach that combines Artificial Neural Networks (ANNs) with Genetic Algorithms (GAs) to predict and minimize the thermal resistance of CLWHP. Experimental data from 192 systematic tests across four curvature ratios (0.0451-0.0924), three inclination angles (30–70°), four filling ratios (50–90%), and four power levels (50–125 W) were collected. A feed-forward ANN with two hidden layers was trained using Levenberg-Marquardt backpropagation, achieving exceptional predictive accuracy (R = 0.983, MAPE =1.63%). The trained ANN was integrated with a GA optimization framework to identify optimal parameter combinations minimizing thermal resistance, avoiding undesirable designs and operational settings. The performance optimization reveals that curvature ratio is the dominant factor (contributing 39% to performance improvement) through enhanced Dean Vortices and secondary flow mixing, while power, filling ratio, and inclination must be carefully balanced to avoid diminishing returns. Optimized Thermal resistance reduced by 51.7 % and 16.8% compared to the lower and upper bounds of the key parameters, respectively. This validated framework provides a reliable tool for optimizing passive two-phase thermal systems, significantly reducing experimental requirements costs while attaining superior thermal performance.