Abstract: Aiming at the crack defects caused by the complex nonlinear mapping relationship between the key process parameters and the prepared coating in the process of laser cladding, the network model between the process parameters and the coating crack density was constructed by combining genetic algorithm and BP neural network, and the network model was trained, fitted and predicted by MATLAB software.The results show that the minimum prediction error of the BP neural network optimized model by genetic algorithm is 3.06%, the average error is controlled within 11.57%, and the mean square error is 0.0008.The prediction accuracy of the model is high and the performance is stable.It verifies the feasibility of combining the theoretical model with the actual test, and reduces a large number of repetitive tests required for the study of coating crack process.It is of great significance to prepare crack free nickel base cladding coating.