Abstract: In order to solve the problem of Oxy-hydrogen micro-flame technological parameters'effect on soldering quality, this paper uses orthogonal test to narrow the technological parameter range and 100 sets of real data are obtained through experiments. 92 groups of experimental data as input samples were imported into the MATLAB neural network module to establish BP neural network prediction model between soldering quality index and soldering technological parameters. After a number of optimization training, the remaining 8 sets of process parameters are combined into the optimized BP neural network model for model accuracy verification. The Oxy-hydrogen micro-flame soldering experiment was carried out on the PCB, and the error range between the predicted value and the true value was found to be no more than 4%. A set of optimal process parameters was obtained by using the trained BP neural network prediction model. The experimental results show that more than 98 percentage of average soldering coverage can be obtained by using this process parameter set, which can be used in the practical Oxy-hydrogen micro-flame automatic soldering.