| As one of the important basic building materials,cement is widely used in the process of modernization in our country.The construction of bridges,roads and buildings cannot be separated from cement.The progress of the cement industry is not only related to the development of the country,but also affects the quality of life of the people across the country.Basic theoretical research and applied research are of great significance to promote the progress of the cement industry.The application of deep learning technology to explore the kinetics of cement hydration,the evolution of cement microstructure and the changes in its physical and chemical properties has greatly accelerated the progress of cement research.As a result,many excellent cement simulation models have been born,which provide basic simulation services for a certain characteristic of cement.However,these models often exist as independent individuals,and it is difficult to interact to promote the development of the entire cement hydration simulation research.Therefore,the goal of this article is to build a platform that can provide core simulation services for cement hydration and can quickly integrate many current excellent simulation models.The platform can provide a closed-loop simulation experiment to provide cement enterprise users to solve the problem of resource waste in the production process of cement foundation;secondly,the platform is provided to cement-related scientific research personnel to use,so that scientific researchers can quickly and easily integrate their own new The researched simulation model is integrated into the system,which promotes the complementary advantages of cement scientific research institutions and cement enterprises,and the situation of collaborative development of industry,university and research institute.The research content of this article mainly includes the following three aspects:(1)Design and implement the core cement hydration closed-loop experimental simulation engineThis dissertation uses deep learning methods to explore the key implementation technologies and simulation construction strategies of cement inversion models,realizes the generation and acquisition of cement data,the simulation of cement hydration evolution,the closed-loop experimental simulation engine for cement performance estimation,and the integration of time series image acquisition methods Build the core simulation channel of the system.In this dissertation,the STS-GAN texture synthesis method is used to build a 3D cement texture generation model,which provides effective image data support for cement hydration modeling and simulation,and cement performance estimation simulation.A deep belief network is used to build a high-precision cement hydration evolution simulation model.In simulating the cement hydration process in the real environment,the simulation evolution verification is performed on the generated cement microstructure texture image,and the simulation results are true and credible;the convolutional neural network is used to build a cement performance estimation simulation model to realize the cement resistance of the system cement image data.Non-destructive estimation of compressive strength;finally a closed-loop simulation experiment for the whole age from cement generation to evolution to performance evaluation is constructed.(2)A time series cement image layer interpolation method is proposed to obtain time series3 D images of cement hydrationFirst,the advantages of the moving least squares curve and surface fitting interpolation method for acquiring time series cement microstructure images are explored,and the strategy of applying the moving least squares method for acquiring cement time series images is proposed.At the same time,the cement image layer interpolation method was improved for the business scenarios of the two time series image acquisition of the virtual cement hydration simulation platform.In the business scenario for the acquisition of complete sequential cement image slices,it is proposed to use a parallel computing strategy to greatly improve the acquisition efficiency,and experiments are carried out to verify its effect.In the real-time interactive business scene of platform-oriented front-end time series cement images,a cement voxel mapping strategy is proposed to achieve efficient time-series image acquisition and efficient front-end 3D image interactive applications.Through experimental analysis and verification,it is proved that the time series cement hydration simulation image obtained by this method has excellent characteristics such as smooth evolution surface,high authenticity,good continuity,and smoother interaction.(3)Design and complete the development and testing of cement hydration simulation platformThis dissertation builds a set of models and service access solutions that are easy to integrate and flexible application management through the fusion application of Docker virtualization technology,microservice technology,Web GL 3D image rendering technology,and code generation technology.The basic platform,simulation center,and integration center of the virtual cement hydration simulation platform were designed and completed,and the development and testing of the virtual cement hydration simulation platform was finally realized.First of all,in view of the diversity characteristics of the inversion model,the combination of Docker virtualization technology and microservice technology is studied to complete the efficient deployment design of model services,and realize the automatic feature of the new inversion model uploading code to complete the service deployment.At the same time,code generation technology and table construction technology are applied to accelerate the integration of new services on the basic platform,making the integration of new services more efficient and faster.Secondly,the whole set of technologies of Spring MVC was used to realize the design and construction of the front and back ends,complete the design of the database and cement simulation architecture,and finally completed the development and testing. |
PDF Full Text Request