An intelligent post-collision safety assessment model based on SHAP-attention mechanism

In view of the current situation that there are practical limitations in the prevention of passenger car and guardrail collision accidents, this study innovates to construct a post-accident intelligent safety assessment model based on SHAP-attention mechanism. By fusing the physical response data of collision, the weight matrix constructed by SHAP values is embedded into the attention mechanism to form an interpretable safety factor function composed of a quadratic decay term of velocity and an exponential decay term of energy change rate. When t=0.35s is used for the first time, the acceleration jump point accurately anchors the collision timing, when t =0.5s (i.e. 0.15s after collision), the safety factor decays exponentially from the initial value ≈1.0 to the critical value of human body injury 0.05, which constructs the physical evidence chain, accurately defines the accident responsibility identification and reduces the cost of judicial disputes; and establishes the attenuation slope of safety factor and the quantification of personnel injury grade, which promotes the innovation of insurance premium actuarial system. This study transforms unpreventable accidents into quantifiable safety benefits, providing core support for intelligent transportation "post-event evaluation drives pre-prevention and control."