Investigation on the influence of the Bohai Sea Geological Environment on the submarine stratum stability based on data mining: an intelligent prediction model
Xin Li,Chao Jia,Hualin Zhao,Yongbo Teng,Yi Zhang,Pengpeng Zhang
DOI: https://doi.org/10.1007/s11356-022-22793-4
Abstract:The Bohai Sea has unique characteristics of the marine geological environment, with strong water closure and weak self-purification ability. Assessment of the submarine stratum stability plays an essential part in ocean engineering construction, preventing marine geological disasters and controlling environmental pollution for the Bohai strait. Nevertheless, limited studies have investigated quantitative classification of the submarine stratum stability or the influence of Bohai Sea's engineering geological environment on the submarine stratum stability using laboratory experiments and the data mining method. Thus, this study carried out physical and mechanical experiments based on the submarine borehole soil samples from the Bohai Sea. And a database containing 60 groups of submarine soil samples' mechanical parameters was established. The physical and mechanical parameters of the Bohai sea's submarine soil, i.e., depth, water content, void ratio, density, internal friction angle, and cohesion, are the prediction indicators. And the submarine soil samples' unconfined compressive strength and elastic modulus are used for quantifying the stability grade of Bohai Sea's submarine stratum. Furthermore, the submarine stratum stability prediction model was established based on the multiple linear regression analysis method and the nonlinear regression analysis method (i.e., BPNN, normal cloud, and TOPSIS method). The testing results indicate that the submarine stratum stability prediction model obtained by the linear regression analysis does not have obvious linear characteristics and gives a poor prediction performance (MAE=0.9333, MAPE=0.3, RMSE=1.2111). In contrast, the nonlinear analysis methods presented a higher fitting version. Compared to the cloud model and the TOPSIS method, the BPNN model with the lowest testing error (MAE=0.2, MAPE=0.1, RMSE=0.4472) presented the best prediction accuracy. The proposed submarine stratum prediction model has been validated as an effective submarine stratum stability prediction method and has good prospects for further ocean engineering applications.