Quality guided image recognition towards industrial materials diffusion

Abstract

Ordered porous materials, especially mesoporous materials, molecular sieves and organometallic framework materials, are widely used in the fields of adsorption, membrane separation and catalytic reaction processes due to their unique properties. Molecular sieve porous material is a new type of engineering material with excellent performance, and its microstructure is one of the key factors affecting macroscopic physical properties and use effect. At present, in order to further broaden its application in the fields of separation, catalysis, sensors and micro-devices, at present, microscopic molecular level control of porous material composition and structure, macroscopically controlling its appearance and appearance has become a research of porous materials. An important development direction. However, the quantitative characterization of the molecular structure of molecular sieve porous materials and its influence on physical properties has always been the focus and difficulty in the field of materials science and engineering. At the same time, the microstructure of porous materials of molecular sieves is the key factor affecting its macroscopic physical properties, and the analysis of spatial structure characteristics. Has important research significance. Based on this paper, a method for predicting effective diffusion coefficient in porous materials by convolutional neural network is proposed. The training samples of porous material microstructure are generated by computer stochastic simulation, and the corresponding effective diffusion coefficients are calculated by finite element method. The image quality subjective evaluation model is used to control the microscopic picture precision of the molecular sieve porous material. The combination of the two methods can quickly and accurately calculate the effective diffusion coefficient.