Prediction model for blast-induced structural damage based on an optimized Bayes discriminant analysis model

Abstract

The damage caused to nearby structures by blast-induced ground vibrations (BIGV) in open-pit mines is essentially dynamic. Predicting this damage can be used to set the allowable range for BIGV. To improve the accuracy of the predictions of structural damage due to BIGV in open-pit mines, an optimized Bayes discriminant analysis (OBDA) model is proposed. In this paper, a stepwise discriminant analysis (SDA) was used to screen the variables, and the Bayes discriminant analysis was optimized with the jackknife method. The influence of different sample covariance matrices and prior probabilities on the OBDA model was considered based on good engineering practice. Moreover, nine comparative models were established for a comparative analysis. The OBDA model is discussed in depth. The rationality of the degree of damage of a structure was demonstrated statistically. The coefficient of variation method and improved CRITIC method were used to calculate the weights of the primary variables. The rationality of SDA was verified. The hypothesis that the covariance matrices were not equal was verified statistically, and the influence of the significance level on the SDA was also discussed. The OBDA model was applied to evaluate the damage to structures caused by BIGV in Tonglushan open-pit mine. The results show that: (1) The model has an excellent discriminatory effect, and its correct-judgment rate was 89.167%. (2) The OBDA model has better prediction performance compared with nine comparative models and with the previously used random forest and gradient-boosted machine models. The OBDA model may be a new option for predicting the damage caused to structures by BIGV in open-pit mining.