Abstract: A machinerock relationship model is developed to investigate the correlation between tunneling parameters and stratum conditions during shield tunneling.  Statistical analysis of tunneling parameters in Nanjing metro line 6, operating under composite strata, is conducted. Initially, a back propagation(BP) neural network model is established and optimized using the artificial bee colony(ABC) algorithm, thereby creating an ABCBP neural network model capable of identifying the tunnel face strata and describing the composite ratio of these strata using tunneling parameters. Subsequently, strata identification and composite ratios prediction for two composite strata within the tunneling section are conducted. The following findings emerge: (1) The shield tunneling parameters′ fluctuation range and average values exhibit discernible patterns that vary with the strata at the tunnel face. (2) The model achieves high identification accuracies of upper soft and lower hard strata, moderatelyweathered argillaceous sandstone, and silty clay, with recall rate reaching 94.1%, 96.6%, and 96%, respectively, resulting in an overall accuracy of 95%. (3) Compared with similar machine learning models, the ABCBP model demonstrates superior performance in average absolute error, root mean square error, and sample regression values, indicating high prediction accuracy and stability.

Key words: metro shield tunnel, stratum identification, composite strata, tunneling parameters, neural network, composite ratio, machine learning, artificial bee colony algorithm