A rockburst prediction model based on extreme learning machine with improved Harris Hawks optimization and its application

Abstract

As sudden, random, and uncertain rock dynamic disasters, rockbursts often threaten the lives of construction workers. Therefore, developing new rockburst intensity prediction methods is particularly important for the design and construction of hard rock geotechnical engineering projects. In this paper, a rockburst prediction method based on extreme learning machine (ELM) with improved Harris Hawks optimization (IHHO) was proposed for more accurate rockburst intensity predictions. First, 136 sets of typical rockburst case data were selected and subjected to normalization to get dimensionless data. Then, chaotic mapping and crossover and mutation operators were used to improve the Harris hawks optimization (HHO) and enhance its global search capability. Then 9 test functions were used to test, compare, and analyze the performance of genetic algorithm (GA), particle swarm optimization (PSO), HHO, and IHHO. Finally, a system was built based on the constructed rockburst intensity level prediction model and MATLAB programming. The comprehensive rockburst intensity level prediction system was applied to the headrace tunnels of Jinping-II Hydropower Station, contrasting the results of IHHO-ELM rockburst prediction model with those of FCM-MFIS model, six conventional machine learning models and the single-index rockburst criterion. The results show that its accuracy was as high as 94.12%, and has a higher convergence speed and higher prediction accuracy and may prove a new way of rockburst intensity level prediction.