Research On Intelligent Prediction Method Of TBM Rock Breaking Efficiency Based On Morphology Of Rock Chips

Tunnel boring machine(TBM)has widely been used in tunnel engineering due to the high advance rate,good tunnel shaping,simple operation and high safety.During the TBM tunneling,the rock cutting efficiency by the cutter is affected by the complicated geological conditions and operational parameters.Therefore,it is necessary to study the influences of the joint conditions and cutter spacing on rock breaking efficiency of TBM.In the present study,a series of two-dimensional indentation tests were conducted on jointed rock specimens.The cracking behaviors and the rock breaking modes by disc cutters under different joint rock mass characteristics and cutter spacing were analyzed.Subsequently,morphology of rock chips were measured.According to the experimental results,machine learning methods were introduced to establish a predict model for rock cutting efficiency of TBM.The main research contents are as follows:(1)In order to study the cracking behaviors and rock chips formation law of the rock mass during the tunneling,a series of two-dimensional rock indentation tests were carried out.After the preparation of rock samples,loading and data collection,etc.,a single rock breaking process of the TBM cutters under the confining pressure in the jointed rock mass was simulated.(2)Based on the two-dimensional rock breaking tests,the influence of joint inclination,joint spacing and cutter spacing on the crack propagation characteristics in rock masses was studied.Then the different rock fragmentation modes and morphology of rock chips were summarized.(3)The particle size distributions and shape sizes the rock chips formed in the tunneling process were analyzed.The roughness index and flatness were employed to estimate the rock cutting efficiency of TBM.After that,the effects of joint inclinations and joint spacing on the rock fragmentation mechanism and the rock cutting efficiency of TBM were studied.According to the experimental results,regression analyses on the roughness index,flatness and specific energy were carried out.The results show that it is a feasible method to estimate the rock cutting efficiency by the particle size distributions and morphology of rock chips.(4)BP neural network(GA-BPNN)optimized based on genetic algorithm,gradient boosting regression tree based on particle swarms optimized(PSO-GBRT)and support vector regression algorithm(PSO-SVR)were established to estimate the rock cutting efficiency.In these prediction models,joint rock mass parameters,TBM operational parameters and rock chips geometric parameters were set as input parameters.While specific energy was set as output parameter.At the same time,the prediction results obtained from these three models were compared and analyzed.It was found that the PSO-SVR model shows better prediction accuracy than the other two models.The analysis provides important references for the prediction of TBM rock cutting efficiency.