Optimization of vascular structure of self-healing concrete using deep neural network (DNN)

•An input representation method is proposed to effectively represent the concrete beams with 6 round pores in the middle span as well as benefit the optimization process.•With the concrete structures encoded as inputs, the DNN model is capable to predict peak load and toughness with high accuracy.•It is feasible to optimize concrete structures towards higher peak load and toughness by fixing the weights of DNN and training the inputs.