Design and multi-objective optimization of heavy-load lifting buffer airbag

To address the challenges of rapid deployment for heavy-duty unmanned platforms and impact protection during drop-landing, a buffer airbag system was designed to ensure the safe landing of the unmanned platform. First, theoretical modeling was used to preliminarily determine the geometric parameters of the buffer airbag and the size of the vent holes. A finite element simulation model that included the buffer airbag, the unmanned platform and the ground was established using the control volume method (CVM) to simulate the cushioning process of the airbag. The impact dynamics response during the landing attenuation process was analyzed, and the specific energy absorption (SEA) of the buffer airbag system was introduced to evaluate its cushioning performance. The analysis results indicate that during the landing buffer process, the unmanned platform reached a maximum overload value of 16.8g at 0.06 s, and the platform velocity reached a post-rebound maximum of 0.79 m/s at 0.1 s. At 0.18 s, the platform shifted laterally due to the center of gravity, and one side of the platform first contacted the ground. This led to fluctuations in overload value and velocity, but the overall landing process remained stable. The calculated SEA of the buffer airbag system was 1527 J/kg. Using the control variable method, the influence of key parameters on the buffer performance of the airbag was analyzed, revealing the effects of vent hole size, critical exhaust pressure, airbag base area and airbag height on the performance of the buffer. The study found that the influence of airbag height on the two key indicators (maximum overload and SEA) was consistent, presenting no optimization conflict. However, after the height exceeded 0.37 m, a diminishing returns effect appeared. When the other three variables were changed individually, they caused a conflict between the goals of minimizing maximum overload and maximizing SEA, making it impossible to improve both simultaneously through independent optimization. To resolve this issue, an optimal latin hypercube design was used for sampling, and a three-layer neural network was employed to construct proxy models for the objective functions. The root mean square error (RMSE) and coefficient of determination (R²) of these two types of proxy models were calculated to evaluate their regression performance. The NSGA-Ⅱ genetic algorithm was integrated for multi-objective optimization. The results showed that for the proxy models predicting maximum overload and SEA, the RMSEs were 0.4895 and 0.7262, and the R² values were 0.9833 and 0.9364, respectively. All R² values were greater than the industry benchmark of 0.9, indicating high reliability of the proxy models. When the vent hole area was 3952 mm², the critical exhaust pressure, the airbag base area were 158 kPa and 1.08 m², respectively. Meanwhile, the maximum overload was reduced from 16.8g to 14.5g, and the SEA remained stable at 1529 J/kg. Based on the characteristics of the airbag's buffer process, a gantry crane-based simulation test method for unmanned platform airdrop and lifting was designed. A piezoelectric IEPE acceleration sensor was used to measure the overload value of the unmanned platform. Experimental verification showed that the measured maximum overload value was 15.2g, with a mere 4.8% error compared to the simulation result, confirming the correctness of the optimization scheme. This research provides efficient and low-cost design technical support for the soft landing systems of unmanned platforms, enhancing the safety and efficiency of operational deployment.