CN-121998822-A - Modal conversion method and device for cement hydration image data

CN121998822ACN 121998822 ACN121998822 ACN 121998822ACN-121998822-A

Abstract

The invention provides a method and a device for transforming the mode of cement hydration image data, which can be used in the technical field of artificial intelligence, and the method comprises the steps of scanning a hardened cement paste sample, and constructing a sample data set according to CT sample data and BSE sample data obtained by scanning; through training a data set, training a preset diffusion model, constructing a modal transformation model, carrying out modal transformation on CT data to be transformed to generate BSE data, realizing modal transformation from an X-CT image of a cement hydration microstructure to the BSE image, effectively combining the time-space imaging advantage of the X-CT with the high resolution and contrast advantage of the BSE, greatly improving the spatial resolution of a hydration image, ensuring the evolution imaging of the whole time scale, comprehensively reflecting the whole process of cement hydration, reducing the cost, effectively distinguishing hydration products with similar density, accurately capturing the details in the hydration process, and realizing the four-dimensional imaging of the cement hydration process.

Inventors

LIU ZHAOLIN
LI QINGBIN
HU YU
LI HAOXIN
MA RUI
ZHANG FENGQIANG

Assignees

清华大学

Dates

Publication Date: 20260508
Application Date: 20251202

Claims (11)

1. A method of modal transformation of cement-hydration image data, the method comprising: preparing a hardened cement paste sample; Scanning the hardened cement paste sample to obtain three-dimensional imaging sample data of cement hydration and scanning sample data of an electron microscope; Constructing a sample data set according to the three-dimensional imaging sample data and the scanning sample data of the electron microscope, wherein the sample data set comprises a training data set; Training a preset diffusion model through the training data set, and constructing a modal transformation model; And carrying out modal transformation on the three-dimensional imaging data to be transformed through the modal transformation model to generate scanning data of the electron microscope.
2. The method of claim 1, wherein preparing a sample of hardened cement paste comprises: Preparing hardened cement paste of different ages; And fixing and slicing the hardened cement paste to obtain a hardened cement paste sample.
3. The method of claim 1, wherein the obtaining of the three-dimensional imaging sample data and the electron microscope scan sample data of the cement hydration by scanning the hardened cement paste sample comprises: scanning and imaging the hardened cement paste sample through a computer tomography technology to generate the three-dimensional imaging sample data; And carrying out carbon spraying treatment on the hardened cement paste sample subjected to the computed tomography, and scanning the hardened cement paste sample subjected to carbon spraying through an electron microscope back scattering technology to generate the scanning sample data of the electron microscope.
4. The method of claim 1, wherein constructing a sample dataset from the three-dimensional imaging sample data and electron microscope scan sample data comprises: According to the three-dimensional imaging sample data and the electron microscope scanning sample data, a sample mapping relation is established, wherein the sample mapping relation comprises the three-dimensional imaging sample data and the corresponding electron microscope scanning sample data; Performing spatial registration and distortion correction on the three-dimensional imaging sample data and the corresponding scanning sample data of the electron microscope through an image registration algorithm to generate paired sample data; Constructing a sample data set according to the paired pairs of sample data; the sample dataset is divided into a training dataset and a validation dataset.
5. The method for modal transformation of cement-hydration image data according to claim 1, further comprising, prior to said training a preset diffusion model with said training dataset, constructing a modal transformation model: and performing data augmentation processing on the training data set to generate an augmented training data set, wherein the data augmentation processing comprises rotation, region cutting, random turnover, random gray scale adjustment and random brightness adjustment.
6. The method for modal transformation of cement-hydration image data according to claim 1, wherein training a preset diffusion model by the training data set to construct a modal transformation model includes: taking the scanning sample data of the electron microscope as a supervision signal, and performing forward diffusion on the scanning sample data of the electron microscope according to a preset noise signal according to a preset time step to generate noise image data; according to the preset time step, reversely generating according to the noise image data to obtain a predicted noise signal; taking three-dimensional imaging sample data corresponding to the scanning sample data of the electron microscope as constraint conditions, and generating conditions according to the predicted noise signals and the noise image data to obtain restored scanning data of the electron microscope; And carrying out back propagation according to a loss result between the restored scanning data of the electron microscope and the scanning sample data of the electron microscope through a loss function, iteratively updating model parameters, and constructing the modal transformation model.
7. The method of modal transformation of cement-hydration image data of claim 1, further comprising: Performing model test on the modal transformation model through the constructed test data set to obtain scanning test data of an electron microscope; And comparing the volume fractions according to the scanning test data and the actual measurement data of the electron microscope, and generating a model verification result.
8. A device for modal transformation of cement hydration image data, the device comprising: a cement paste sample preparation unit for preparing a hardened cement paste sample; the sample data acquisition unit is used for acquiring three-dimensional imaging sample data of cement hydration and scanning sample data of an electron microscope by scanning the hardened cement paste sample; A sample data set construction unit for constructing a sample data set according to the three-dimensional imaging sample data and the scanning sample data of the electron microscope, wherein the sample data set comprises a training data set; the model training unit is used for training a preset diffusion model through the training data set to construct a modal transformation model; And the modal transformation unit is used for carrying out modal transformation on the three-dimensional imaging data to be transformed through the modal transformation model to generate scanning data of the electron microscope.
9. A computer readable medium having stored thereon a computer program, characterized in that the program, when executed by a processor, implements a method of modal transformation of cement-hydration image data according to any one of claims 1 to 7.
10. A computer device comprising a memory for storing information including program instructions and a processor for controlling execution of the program instructions, wherein the program instructions when loaded and executed by the processor implement the method of modal transformation of cement-hydration image data of any one of claims 1 to 7.
11. A computer program product comprising computer programs/instructions which when executed by a processor implement the method of modal transformation of cement-hydration image data of any one of claims 1 to 7.

Description

Modal conversion method and device for cement hydration image data Technical Field The invention relates to the technical field of computers, in particular to the technical field of artificial intelligence, and particularly relates to a method and a device for modal transformation of cement hydration image data. Background The key to the performance study of cement concrete is its microstructure. The microstructure is derived from the development of cement hydration, and the direct characterization, particularly the nondestructive dynamic detection, of cement hydration has long been the focus of attention in the industry and is also the bottleneck of technology. In the related art, a microscopic image in the cement hydration process can be obtained with high resolution through a focused ion beam electron microscope (FIB-SEM) back scattering technology, and the microscopic structure of cement is accurately represented, but the FIB-SEM belongs to a lossy detection technology and has high cost, a sample can be polished and cut under the action of an ion beam, so that time information in the hydration process can not be reserved, and meanwhile, the cutting range of the sample is smaller, the obtained data result lacks extensive space representativeness, and the whole cement hydration process is difficult to comprehensively reflect. Another related art uses a super-resolution method to interpolate a low-resolution X-ray computed tomography (X-CT) image by an interpolation technique to generate new pixels, so as to improve the resolution of the X-CT image, but the X-CT image itself has a certain limitation, especially between substances with close densities (for example, between high-density csch and calcium hydroxide, ettringite), which are often difficult to be effectively distinguished due to physical limitations of X-ray attenuation. Even if the super-resolution algorithm is adopted to improve the resolution of the image, the difficulty of distinguishing substances is still difficult to overcome, and therefore, details in the hydration process cannot be accurately captured. The traditional interpolation and reconstruction algorithm can only improve the pixel resolution of the image to a certain extent, so that the spatial resolution of the image is difficult to truly improve, and the image cannot be accurately displayed on the representation of hydration details. Disclosure of Invention The invention aims to provide a mode conversion method of cement hydration image data, which realizes the mode conversion from an X-CT image of a cement hydration microstructure to a BSE image, and can effectively combine the time-space imaging advantage of the X-CT with the high resolution and contrast advantage of the BSE, thereby greatly improving the spatial resolution of the hydration image, ensuring the evolution imaging of the whole time scale, comprehensively reflecting the whole process of cement hydration, reducing the cost, effectively distinguishing hydration products with similar density, accurately capturing details in the hydration process and realizing the four-dimensional imaging of the cement hydration process. Another object of the present invention is to provide a device for modal transformation of cement hydration image data. It is yet another object of the present invention to provide a computer readable medium. It is a further object of the invention to provide a computer device. In order to achieve the above object, the present invention discloses a method for transforming cement hydration image data in a mode, comprising: preparing a hardened cement paste sample; scanning the hardened cement paste sample to obtain three-dimensional imaging sample data of cement hydration and scanning sample data of an electron microscope; according to the three-dimensional imaging sample data and the scanning sample data of the electron microscope, a sample data set is constructed, and the sample data set comprises a training data set; training a preset diffusion model through a training data set to construct a modal transformation model; and carrying out modal transformation on the three-dimensional imaging data to be transformed through a modal transformation model to generate scanning data of the electron microscope. Preferably, preparing a hardened cement slurry sample comprises: Preparing hardened cement paste of different ages; and fixing and slicing the hardened cement paste to obtain a hardened cement paste sample. Preferably, the method for obtaining three-dimensional imaging sample data and electron microscope scanning sample data of cement hydration by scanning a hardened cement paste sample comprises the following steps: scanning and imaging the hardened cement paste sample by using a computer tomography technology to generate three-dimensional imaging sample data; And (3) carrying out carbon spraying treatment on the hardened cement paste sample subjected to the computed tomography scanning, and scanning t