Geometry and cooling effects on an induction self-heating insert for rapid mold temperature control: Experimental, Multiphysics, and artificial neural network-based analysis

Control of mold-cavity surface temperature strongly affects filling, premature frozen-layer formation, and cycle time in thin-wall injection molding. This study evaluates an induction self-heating insert with a thin heating zone directly coupled to a cooling layer using experiments, a sequentially coupled electromagnetic-thermal model, and an Artificial Neural Network (ANN). The design space contains 25 Taguchi L25 geometries defined by T, W, and L; X and Y coordinates are added so that the ANN predicts the surface temperature field. Three cooling states are considered: No Water (NW), Stagnant Water (SW), and Flowing Water (FW). Experimental ANN targets are the mean of five repeated tests. The simulation reproduced the location of the hot zone; among four cases with reported quantitative differences, three differed by no more than about 10 °C and one showed a peak difference of about 20 °C. To prevent data leakage, all 846 points from each geometry were kept together during geometry-wise validation. The resulting R² values were 0.746, 0.653, and 0.493 for NW, SW, and FW, with Root Mean Square Error (RMSE) values of 17.03, 12.97, and 12.62 °C, respectively. The effects of T, W, and L depended on the cooling state and were not consistently monotonic. A proof-of-concept molding test showed complete filling of a Chip 2 part using Low-Density Polyethylene (LDPE); the mean spacing was 28.6 µm for a nominal 30 µm, and the mean height was 106.6 µm for a nominal 100 µm. These results show that combining insert geometry, cooling-water state, and a surrogate model can support local thermal design, delay frozen-layer formation, and improve filling in injection molding.