Abstract:

In order to recognize the boulder geology accurately during shield tunneling and ensure construction safety, the recognition methods are studied by theoretical analysis and engineering data validation. And then the modulated specific energy(SM) is put forward and the SM recognition model and recognition matrix of boulder geology are constructed based on the shield tunneling specific energy and main tunneling parameters. By using the BP neural network technology, the monitored tunneling parameters are identified as training samples to establish neural network recognition model of the boulder geology, which has very high recognition accuracy. Finally, the recognition methods are verified by using shield tunneling data. The results show that: (1) The SM method has high fault tolerance, stability and specificity, whose prejudgment effect is better than that of the tunneling specific energy method. (2) The coincidence rates of neural network prejudgment to recognition matrix under 2 sets of monitored data reach 98.3% and 98.8%, respectively. (3) The dual recognition method for boulder geology by taking SM method as basis and neural network method as an auxiliary and reference has a good practical significance.

Key words: shield, tunneling parameters, boulder geology recognition, modulated specific energy, neural network