Prediction model of joint roughness coefficient based on Gaussian process regression

The joint roughness coefficient (JRC) estimation may produce a sufficiently unreliable result, due to limitation of single statistical parameter method for characterizing morphology. A model based on Gaussian process regression (GPR) combined with principal component analysis (PCA) was proposed for the quantitative evaluation of JRC. Notably, eight parameters were selected as indicators for the comprehensive expression of the rock joint roughness. In order to analyze the model’s performance, 112 published rock joint profiles were used as the database, of which 95 were chosen as training database and 17 as validation database. The reliability of the model was verified by comparing the predicted results with the measured data. Results show that the derived GPR model demonstrates promising performance (R^2=0.972, MSE=0.517) for estimating JRC, reflecting the high applicability for multi-parameter JRC prediction even when the number of training dataset is small. In general, the GPR model may provide a new way of thinking about estimating JRC values with artificially intelligent.