A multi-fidelity prediction model for vertical bending moment and total longitudinal stress of a ship based on composite neural network

In ship engineering,the prediction of vertical bending moment(VBM)and total longitudinal stress(TLS)during ship navigation is of utmost importance.In this work,we propose a new prediction paradigm,the multi-fidelity regression model based on multi-fidelity data and artificial neural network(MF-ANN).Specifically,an ANN is used to learn the fundamental physical laws from low-fidelity data and construct an initial input-output model.The predicted values of this initial model are of low accuracy,and then the high-fidelity data are utilized to establish a correction model that can correct the low-fidelity prediction values.Hence,the overall accuracy of prediction can be improved significantly.The feasibility of the multi-fidelity regression model is demonstrated by predicting the VBM,and the robustness of the model is evaluated at the same time.The prediction of TLS on the deck indicates that just a small amount of high-fidelity data can make the prediction accuracy reach a high level,which further illustrates the validity of the proposed MF-ANN.

Numerical simulation of hydraulic fracture propagation in laminated shale reservoirs

The main area of the Jiaoshiba anticline of the Fuling shale gas field was taken as the research object,laboratory rock mechanical experiments and direct shear experiments were conducted to clarify the mechanical anisotropy characteristics and parameters of rock samples with rich beddings.Based on the experimental results,a 3D fracture propagation model of the target reservoir taking mechanical anisotropy,weak bedding plane and vertical stress difference into account was established by the discrete element method to analyze distribution patterns of hydraulic fractures under different bedding densities,mechanical properties,and fracturing engineering parameters(including perforation clusters,injection rates and fracturing fluid viscosity).The research results show that considering the influence of the weak bedding plane and longitudinal stress difference,the interlayer stress difference 3–4 MPa in the study area can control the fracture height within the zone of stress barrier,and the fracture height is less than 40 m.If the influence of the weak bedding plane is not considered,the simulation result of fracture height is obviously higher.Although the opening of high-density bedding fractures increases the complexity of hydraulic fractures,it significantly limited the propagation of fracture height.By reducing the number of clusters,increasing the injection rate,and increasing the volume and proportion of high-viscosity fracturing fluid in the pad stage,the restriction on fracture height due to the bedding plane and vertical stress difference can be reduced,and the longitudinal propagation of fractures can be promoted.The fracture propagation model was used to simulate one stage of Well A in Fuling shale gas field,and the simulation results were consistent with the micro-seismic monitoring results.