Creep Parameter Inversion for High CFRDs Based on Improved BP Neural Network Response Surface Method

The creep parameters of rockfill materials obtained from engineering analogy method or indoor tests often cannot accurately reflect the long-term deformation of high Concrete Faced Rockfill Dams (CFRDs). This paper introduces an optimized inversion method based on Multi-population Genetic Algorithm improved BP Neural Network and Response Surface Method (MPGA-BPNN RSM). The parameters used for inversion are determined by parameter sensitivity analysis based on the statistical orthogonal test method. MPGA-BPNN RSM, validated by Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and squared correlation coefficient (R2), etc., completely reflects the response between the creep parameters and the settlement calculation values obtained by Finite Element Method (FEM). MPGA optimized the objective function to obtain the optimal creep parameters. The results show that the settlement values of Xujixia CFRD calculated by FEM using the inversion parameters has great consistency with the monitored values both in size and in distribution, suggesting that the model parameters obtained by the introduced creep parameter inversion method are feasible and effective. The introduced method can improve the inversion efficiency and the prediction accuracy in FEM applications.

Biot's equations-based reservoir parameter inversion using deep neural networks

Reservoir parameter inversion from seismic data is an important issue in rock physics. The traditional optimisation-based inversion method requires high computational expense, and the process exhibits subjectivity due to the nonuniqueness of generated solutions. This study proposes a deep neural network (DNN)-based approach as a new means to analyse the sensitivity of seismic attributes to basic rock-physics parameters and then realise fast parameter inversion. First, synthetic data of inputs (reservoir properties) and outputs (seismic attributes) are generated using Biot's equations. Then, a forward DNN model is trained to carry out a sensitivity analysis. One can in turn investigate the influence of each rock-physics parameter on the seismic attributes calculated by Biot's equations, and the method can also be used to estimate and evaluate the accuracy of parameter inversion. Finally, DNNs are applied to parameter inversion. Different scenarios are designed to study the inversion accuracy of porosity, bulk and shear moduli of a rock matrix considering that the input quantities are different. It is found that the inversion of porosity is relatively easy and accurate, while more information is needed to make the inversion more accurate for bulk and shear moduli. From the presented results, the new approach makes it possible to realise accurate and pointwise inverse modelling with high efficiency for actual data interpretation and analysis.

A Regularization Homotopy Strategy for the Constrained Parameter Inversion of Partial Differential Equations

The main difficulty posed by the parameter inversion of partial differential equations lies in the presence of numerous local minima in the cost function. Inversion fails to converge to the global minimum point unless the initial estimate is close to the exact solution. Constraints can improve the convergence of the method, but ordinary iterative methods will still become trapped in local minima if the initial guess is far away from the exact solution. In order to overcome this drawback fully, this paper designs a homotopy strategy that makes natural use of constraints. Furthermore, due to the ill-posedness of inverse problem, the standard Tikhonov regularization is incorporated. The efficiency of the method is illustrated by solving the coefficient inversion of the saturation equation in the two-phase porous media.

Intelligent Parameter Inversion of Fractional-Order Model Based on BP Neural Network

To explore creep parameters and creep characteristics of salt rock, an Ansys numerical model of salt rock sample was established by using fractional creep constitutive model of salt rock, and an orthogonal test scheme was designed based on uniaxial compression test of salt rock samples. A large number of training data were obtained by combining the numerical model with the experimental scheme, and the model parameters were inverted by using the BP neural network. The model parameters are used for forwarding calculation, and the results are in good agreement with the measured strain data. This shows that the model parameter inversion method proposed in this paper can obtain reasonable parameter values and then accurately predict the creep behaviour of salt rock, which provides a good technical basis for related engineering practice and scientific research in the future.