Integration of discrete step size ANN with continuous differential evolution for predicting complex threshold scenarios of the steady-state thermal explosions

In the present work, an advance computational intelligence paradigm based on the artificial neural network is presented for analyzing the complex threshold scenarios of a steady-state thermal explosion, where temperature oscillations arise just below the system’s thermal threshold, leading to either a blow-up or settling down to a steady state. The solution of the governing differential equation is performed by using the artificial neural networks optimized with global search differential evolution enhanced by local refinements of principal axis maximization. The proposed scheme is applied for analyzing two scenarios: steady state with and without the blowing effect. In order to validate the accuracy of the designed methodology, a comparison of the proposed and exact solutions has been made. Moreover, the statistical interpretations are used to prove the worth, convergence, accuracy, stability, and robustness of the proposed algorithm.