关键词: 超前钻探, 破岩能量, 逻辑回归概率, 地层硬度指标, 概率分类, 地层识别

Abstract: The traditional geological prediction method of advanced drilling usually takes the change rate of one specific drilling parameter as the main basis for stratum identification. The rock breaking of drill bit is a complicated mechanical process. Stratum identification with single drilling parameter results in great uncertainty. The cooperative effect of multiple parameters in drilling process should be considered. Firstly, the advanced drilling data was preprocessed, including standardization, frequency distribution analysis and sensitivity analysis, to select the key drilling parameters sensitive to stratum changes. Secondly, based on the principles of energy conservation, binary disordered logistic regression analysis and multi-parameter variability analysis, three comprehensive identification indexes of rock breaking energy, logistic regression probability and stratum hardness index were established respectively. Finally, a stratum identification model was established by probability classification method based on Bayesian principle, and the model parameters were obtained by ROC analysis method, and the stratum identification based on multiple drilling parameters and probability classification method was realized. Taking the tunnel project with complex geological conditions as an example, the application of the proposed stratum identification method is introduced. The results show that: Three comprehensive indexes have great performance in cross-hole stratum identification, and the identification accuracy exceeds 80%. The rock breaking energy and the logic regression probability are suitable for the cross-hole stratum identification with short distance, and the average identification accuracies are 86.3% and 84.1%, respectively. The logical regression probability index has strong identification capability for the weak interlayer, and the identification accuracy reaches 94.2%. The stratum hardness index is suitable for the cross-hole stratum identification with long distance, and the maximum identification accuracy of limestone is 93.2%.

Key words: advanced drilling, rock breaking energy, logistic regression probability, stratum hardness index, probability classification, stratum identification