Abstract: The geometry shape of fusion-solidification zone in electron beam welding of TC4 titanium alloy was predicted based on artificial neural network.Based on adaptive network model, network structure, and training algorithm and times, a BP neural network mapping model from focused current, beam current and welding speed to fusion penetration, fusion width, top weld width, depth-to-width ratio, weld reinforcement, nailhead half-angle was established.Training samples were obtained from abundant process experiments and normalized for reducing prediction error.The predicted results indicate that maximum relative error is less than 5%, and the prediction requirement can be satisfied.