In order to realize the accurate prediction on characteristics of molt en pool in wide-band laser cladding, and achieve the real-time detection,evaluation and closed-loop control of las er cladding,a full factorial design method is used to conduct the experiments,and the experimental results are chos en randomly as sample data for neural network.Genetic algorithm is utilized to optimize the initial weights an d thresholds of back propagation (BP) neural network.The BP neural network prediction model is develo ped to express the relationship between the process parameters (laser power,powder thickness,scanning speed) a nd the characteristics of molten pool.The training set of data obtained in experiments is used to train the neu ral network to establish the perfect mapping relation between input and output of network.The testing set of data is used to verify the performance of the trained network.Simulation results indicate that the predict ion model of characteristic parameters of molten pool in wide-band laser cladding has sufficient accuracy. The neural network prediction model is of great significance for the real-time monitoring of laser cladding p rocess and the quality control of cladding layers.