A novel qualitative proof approach of the Dulong-Petit law using general regression neural networks

Dulong-Petit law is an ordinary description of specific heat capacity, which states that the heat capacity per weight (i.e., mass-specific heat capacity) for a number of substances becomes close to a constant value. In our study, we trained 30 groups´ data of metal elementary substances to establish a general regression neural network (GRNN) model within NeuralTools Software to predict the constant of the Dulong-Petit law. We used 31 samples to test the robustness of the computer model. In our results, 100% of the tested samples showed accurate results within the permissible error range (30% tolerance).Based on the characteristic of the artificial neural network (ANN) model established by NeuralTools, we applied our model to analyze the weight of different independent variables and test the accuracy of the Dulong-Petit law qualitatively. Finally, we put forward a novel proof method to support the theories and laws of natural science using the ANN model.