Pipe bending forming, as a critical technology in precision manufacturing, is widely used in industrial fields, but defects such as fracture and back caused by multi-parameter coupling seriously restrict the quality and efficiency of forming. This study takes the hydraulic die-less bending machine as the research object, and systematically investigates influence law of process parameters on the quality of the forming by combining theoretical analysis and experiments, and establishes a database of process parameters. Based on the plane strain hypothesis, the exponential hard model and the Mises yielding criterion, the theoretical formulas of key parameters such as neutral layer displacement, springback angle, and wall thickness change rate are derived, and their engineering application is revealed. Through 300 groups of multi-variable experiments, the influence of bending speed, mold gap, and lubrication conditions on the quality of the forming is clarified and optimization suggestions are proposed. The nonlinear mapping model of process parameters and the quality of the forming is constructed by combining the BP neural network, and the parameter combination is optimized by the particle swarm algorithm to realize intelligent control. Experimental verification shows that the intelligent system reduces the Springback angle to 0.7° in the bending of the battery cooling pipe of energy vehicles, shortens the production cycle by 37.8%, and increases the qualified rate to 98%; in the small radius bending of aerospace fuel pipes the springback error is only 0.47°, which is significantly better than the traditional method. In addition, the economic benefit analysis shows that the material loss rate is by 58.3%, and the comprehensive cost is reduced by 26.7%. The research results provide a systematic parameter optimization method and intelligent control strategy for bending forming of pipes, which has significant engineering application value.