Modeling study of uniaxial compressive strength prediction from similar analog drilling test signals

Uniaxial compressive strength (UCS) of rock is one of the important parameters to characterize the properties of rock mass in geotechnical and underground engineering. In addition, in order to perceive and predict the UCS of rock in underground coal mines accurately and quickly, a GA−BP (Genetic Algorithm-Backpropagation) neural network prediction model of UCS based on vibration signals from drilling is constructed based on drilling tests of nine specimens of similar materials with different ratios in the chamber. By varying the number of hidden layers, population and training function of GA−BP neural network, the factors affecting the prediction model and the results are discussed and analyzed to determine the optimal prediction model structure. The results indicate that there is a responding relationship between the vibration signal with drilling and the UCS of similar simulated materials, and the accuracy of the constructed prediction models are above 70%, and the research method of predicting the UCS with the perception of vibration signal with drilling has certain feasibility; The model results are optimal when trainlm is chosen for the training function, the hidden layer is 8, and the number of populations is 20,with coefficients of determination of 0.761 and 0.745 for the training set and the test set, respectively, the root-mean-square errors are 6.039 MPa and 4.254 MPa, and the mean absolute errors are 6.574 MPa and 4.716 MPa, respectively. The UCS prediction method proposed in this paper may provide a new idea for the intelligent identification of rock mechanical properties.