This paper constructed the projectile-barrel coupling surrogate model of the artillery whole lifecycle based on artificial neural network to accelerate the computational efficiency of projectile-barrel dynamic responses, and based on the obtained surrogate model, analyzed the influence of inner diameter expansion of barrel along the axial direction on the coupling dynamic responses, including displacement, velocity, collision, etc. The overall steps are: selecting appropriate design inputs’ parameters and determining their probability distribution types and ranges; using Latin hypercube sampling to generate experimental design samples, and introducing them into the projectile-barrel coupling simulation model to obtain corresponding outputs at each sample point, thereby forming the whole sample datasets; building the projectile-barrel coupling surrogate model based on feedforward neural networks, in which the optimal network parameters are obtained by using particle swarm optimization algorithm; eventually applying the formed sample datasets to train, test, and evaluate the network.