In vehicle crash safety research, accurately capturing the dynamic response characteristics of a vehicle is a core requirement for Computer-Aided Engineering (CAE) analysis. When benchmarking occupant restraint systems using LS-DYNA software, it is essential to integrate data collected from at least three tri-axial accelerometers placed at different locations on the vehicle body to describe the vehicle's motion during a crash. However, deploying such sensors at multiple points in actual crash tests not only increases the project's economic costs but may also significantly reduce the accuracy of the CAE results due to potential sensor failures. To address this issue, this paper proposes a fusion algorithm based on the Kalman filter. This algorithm enables real-time correction and precise reconstruction of faulty data by collaboratively processing data from two normally functioning tri-axial accelerometers. The corrected data is then input into the CAE model, significantly enhancing the accuracy of simulation calculations. Experimental validation shows that this method effectively prevents data distortion caused by accelerometer malfunctions, ensuring the reliability and consistency of the simulation results. This study not only reduces the economic cost of crash testing but also provides strong technical support and solutions for improving vehicle safety CAE analysis.