Abstract
Abstract Shock response spectrum (SRS) is the primary method of simulating explosive separation impact environment. SRS is obtained from shock experiments in traditional research, which consumes lots of materials. In this study, a prediction model based on machine learning (ML) is established to rapidly obtain SRS. 1053 sets of data obtained by shock experiments were used in the model. A total of 15 features are selected, in which 7 accelerations of different frequencies on the SRS are taken as target features. 7 types of ML algorithm are used for the prediction and the best one in each target is integrated in the completed model. The results indicate that the optimal average percentage errors at frequencies of 100Hz, 500Hz, 1000Hz, 3000Hz, 5000Hz, 7000Hz, and 10000Hz are 5.9%, 6.8%, 8.1%, 19.5%, 20.0%, 22.0%, and 19.7%, respectively. The error range of SRS fit by the 7 prediction targets is within ±20%. This study is contributed to improving the debugging efficiency of SRS.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call