Composite refractory brick heat insulation performance prediction method based on firework algorithm improved BP network

The invention provides a firework algorithm improved BP network-based composite refractory brick heat insulation performance prediction method. The method comprises the steps of obtaining original data, constructing a finite element simulation model, designing a simulation experiment scheme, performing simulation to obtain a machine learning data sample, training and constructing an FWA-BP network model, and verifying the FWA-BP network model. Aiming at the problem of predicting the influence of different design parameters on the heat insulation performance of the composite refractory brick, a finite element simulation model is constructed based on a very small number of experimental results, and a data sample for machine learning is obtained through simulation; and with the BP network model as the center, an initial weight and a threshold value are optimized in combination with an intelligent algorithm-fireworks algorithm FWA, and finally a novel FWA-BP network model which is high in prediction precision and