The Speed Prediction Research of Peak Particle Vibration Velocity in Underwater Blasting Based on GWO-SVR

The blasting vibration produced in the blasting process of underwater engineering brings serious damage to the surrounding environment. Predicting blasting peak particle velocity (PPV) is one of the effective ways to alleviate the problem. To further improve the prediction accuracy of blast PPV, the grey wolf optimization (GWO) algorithm is used in this paper to optimize the penalty factor and radial basis kernel function parameters of support vector regression (SVR) model iteratively, and a blasting PPV prediction model was established. Taking the Dajin Island water intake open channel of the Phase 1 project of Guangdong Taishan Nuclear Power Station as the engineering background, according to the relevant parameters of blasting vibration recorded in 30 blasting tests, taking into account blasting design parameters and geological conditions, a database consisting of 12 inputs (hole length ( HL ), spacing ( S ), row spacing ( R s ), burden ( B ), stemming length ( l s ), the distance of blasting center ( B d ), height differential elevation ( H de ), seawater pressure ( P s ), powder factor ( PF ), maximum charge of single hole ( Q smax ), maximum charge per delay ( Q max ) and total charge ( Q tot ) and 1 output (PPV)) was established. Then, the grey wolf optimization-support vector regression (GWO-SVR) model, double-layer neural network, medium decision tree, and empirical SVR model are used to establish a prediction model for the PPV in underwater blasting respectively, and the prediction results are compared and analyzed. The results are as follows: the actual value - predicted value diagram and residual comparison show that the prediction effect of PPV based on the double layer neural network model is the worst in underwater blasting. The comparison of regression evaluation indexes shows that GWO-SVR is the best method for predicting PPV in underwater blasting; its R 2 is 0.9285, RMSE is 0.21424, MSE is 0.0459 and MAE is 0.1625. The research results can provide a theoretical reference for the construction of similar underwater blasting projects, and provide a scientific basis for delineating a reasonable safety warning range for similar underwater blasting projects.