Enhancing bonding reliability of solid propellant grain based on FFTA and PSO-GRNN

Bonding quality at the interface of solid propellant grains is crucial for the reliability and safety of solid rocket motors. Although bonding reliability is influenced by numerous factors, the lack of quantitative characterization of interface debonding mechanisms and the challenge of identifying key factors have made precise control of process variables difficult, resulting in unpredictable failure risks. This paper presents an improved fuzzy failure probability evaluation method that combines fuzzy fault tree analysis with expert knowledge, transforming process data into fuzzy failure probability to accurately assess debonding probabilities. The predictive model is constructed through a general regression neural network and optimized using the particle swarm optimization algorithm. Sensitivity analysis is conducted to identify key decision variables, including normal force, grain rotation speed, and adhesive weight, which are verified experimentally. Compared with classical models, the maximum error margin of the constructed reliability prediction model is only 0.02%, and it has high stability. The experimental results indicate that the main factors affecting debonding are processing roughness and coating uniformity. Controlling the key decision variable as the median resulted in a maximum increase of 200.7% in bonding strength. The feasibility of the improved method has been verified, confirming that identifying key decision variables has the ability to improve bonding reliability. The proposed method simplifies the evaluation of propellant interface bonding reliability under complex conditions by quantifying the relationship between process parameters and failure risk, enabling targeted management of key decision variables.