CN-122016531-A - Cement-based material hardness prediction method, device, electronic equipment and storage medium

CN122016531ACN 122016531 ACN122016531 ACN 122016531ACN-122016531-A

Abstract

The invention provides a cement-based material hardness prediction method, a device, electronic equipment and a storage medium, belonging to the technical field of material detection, wherein the method comprises the steps of obtaining a back scattering electronic image of a cement-based material sample; and inputting the back scattering electronic image into a trained hardness prediction model, and outputting a pixel-level hardness distribution map by the trained hardness prediction model, wherein the trained hardness prediction model is a multi-task neural network comprising a mineral phase segmentation task and a pixel-level hardness regression task. The method avoids the damage and a large amount of time consumption of the traditional point-to-point indentation test on the sample, and improves the efficiency and the comprehensiveness of the hardness detection of the cement-based material.

Inventors

LIU ZHICHAO
XU ZHIMING
WANG FAZHOU
HU SHUGUANG

Assignees

武汉理工大学

Dates

Publication Date: 20260512
Application Date: 20260104

Claims (10)

1. A method for predicting hardness of a cement-based material, comprising: acquiring a back scattering electron image of a cement-based material sample; inputting the back scattering electronic image into a trained hardness prediction model, and outputting a pixel-level hardness distribution map by the trained hardness prediction model, wherein the trained hardness prediction model is a multi-task neural network comprising a mineral phase segmentation task and a pixel-level hardness regression task.
2. The method of claim 1, wherein obtaining a backscattered electron image of a cement-based material sample comprises: Acquiring microscopic images of the cement-based material sample through a back scattering electron mode of a field emission scanning electron microscope to obtain initial images; based on a preset gray reference target, performing gray normalization and correction operation of the initial image to obtain a calibration image; And sequentially executing noise suppression and local contrast enhancement processing on the calibration image to obtain the back scattering electronic image.
3. The method of claim 1, wherein the multi-tasking neural network employs an encoder-dual decoder architecture comprising an encoder, a first decoder and a second decoder, wherein, The input end of the encoder receives the back scattering electron image, and is used for carrying out feature extraction and downsampling on the back scattering electron image and outputting a multi-scale feature map; The input end of the first decoder is connected with the output end of the encoder and is used for carrying out up-sampling and feature fusion on the multi-scale feature map and outputting a mineral phase segmentation map; and the input end of the second decoder is connected with the output end of the encoder and is used for carrying out up-sampling and feature fusion on the multi-scale feature map and outputting an initial hardness predicted value.
4. A method of predicting hardness of a cementitious material as defined in claim 3, further comprising physically constraining the initial hardness prediction value to be modified, comprising: calculating the duty ratio of each mineral phase in the image according to the mineral phase segmentation map; based on physical rules, calculating a physical constraint hardness value according to the mineral phase proportion, the hardness priori value and the porosity of each mineral phase; and combining the physical constraint hardness value with the initial hardness predicted value to generate a hardness distribution diagram of the pixel level.
5. The method of claim 4, wherein calculating the physical constraint hardness value based on the mineral phase ratio, the hardness prior value of each mineral phase, and the porosity based on the physical rule comprises: the physical constraint hardness value H is calculated by adopting the following formula: Wherein, the Is the first The proportion of the individual mineral phases is calculated, Is the first A priori values of hardness of the individual mineral phases, In order for the porosity to be the same, As the weight coefficient of the light-emitting diode, To characterize the non-linear effect of porosity on hardness.
6. The method of claim 4, wherein the training process is optimized with a combined loss function, and wherein the step of calculating the combined loss function comprises: Calculating the segmentation loss between the predicted mineral phase and the real label; Calculating regression loss between the initial hardness predicted value and the real hardness; calculating a consistency loss between the initial hardness predicted value and the physical constraint hardness value; And carrying out weighted summation on the segmentation loss, the regression loss and the consistency loss to obtain the value of the combined loss function.
7. The method of claim 6, wherein the step of calculating a combined loss function further comprises: Calculating the partial derivative of the initial hardness predicted value to the mineral phase ratio as a first partial derivative, and calculating the partial derivative of the initial hardness predicted value to the porosity as a second partial derivative; constructing a monotonicity penalty based on the first partial derivative and the second partial derivative; calculating the variance of the initial hardness predicted value as uncertainty loss through forward propagation sampling for a plurality of times; introducing the monotonicity loss and uncertainty loss into the combined loss function, and weighted summing with the segmentation loss, regression loss, and consistency loss.
8. A cement-based material hardness prediction apparatus, comprising: The image acquisition module is used for acquiring a back scattering electronic image of the cement-based material sample; The hardness prediction module is used for inputting the back scattering electronic image into a trained hardness prediction model, and outputting a pixel-level hardness distribution map by the trained hardness prediction model, wherein the trained hardness prediction model is a multi-task neural network comprising a mineral phase segmentation task and a pixel-level hardness regression task.
9. An electronic device comprising a memory and a processor, wherein, The memory is used for storing programs; The processor, coupled to the memory, for executing the program stored in the memory to implement the steps in the cement-based material hardness prediction method according to any one of the preceding claims 1 to 7.
10. A computer readable storage medium storing a computer readable program or instructions which when executed by a processor is capable of carrying out the steps of the cement-based material hardness prediction method according to any one of claims 1 to 7.

Description

Cement-based material hardness prediction method, device, electronic equipment and storage medium Technical Field The invention relates to the technical field of material detection, in particular to a cement-based material hardness prediction method and device, electronic equipment and a storage medium. Background Cement-based materials are widely used as an important engineering material in construction, infrastructure and industrial products, especially in the production of concrete and cement composite materials, where the microstructure and micro-domain physical properties of the cement-based materials have a critical impact on their macroscopic properties. In the research and development and quality detection of cement-based materials, micro-zone hardness is taken as an important parameter for evaluating the performance of the materials, and can directly influence the characteristics of durability, compression resistance, corrosion resistance and the like. Conventional hardness testing methods, such as microhardness testing and nanoindentation testing, typically require contact testing on the surface of a sample of material, which is not only destructive to the material, but also complex and time consuming. Especially in the field of cement-based materials, the traditional method often has difficulty in accurately capturing the change of the hardness of the micro-areas, and the hardness difference of different mineral phases in the cement-based materials cannot be accurately reflected. In addition, technologies such as nano indentation have certain limitations in micro-region hardness test, and difficulties exist in selection and accuracy control of indentation positions, so that continuous and fine test processes cannot be realized. How to realize efficient and reliable micro-zone hardness prediction with definite physical significance on the premise of not depending on large-scale destructive sampling test is a technical problem to be solved in the field. Disclosure of Invention In view of the foregoing, it is desirable to provide a method, apparatus, electronic device, and storage medium for predicting hardness of a cementitious material that is effective and reliable without relying on extensive destructive sampling tests. In order to solve the above technical problems, in a first aspect, the present invention provides a method for predicting hardness of a cement-based material, including: acquiring a back scattering electron image of a cement-based material sample; inputting the back scattering electronic image into a trained hardness prediction model, and outputting a pixel-level hardness distribution map by the trained hardness prediction model, wherein the trained hardness prediction model is a multi-task neural network comprising a mineral phase segmentation task and a pixel-level hardness regression task. In one possible implementation, the acquiring the backscattered electron image of the cement-based material sample includes: Acquiring microscopic images of the cement-based material sample through a back scattering electron mode of a field emission scanning electron microscope to obtain initial images; based on a preset gray reference target, performing gray normalization and correction operation of the initial image to obtain a calibration image; And sequentially executing noise suppression and local contrast enhancement processing on the calibration image to obtain the back scattering electronic image. In one possible implementation, the multi-tasking neural network employs an encoder-dual decoder architecture, including an encoder, a first decoder, and a second decoder, wherein, The input end of the encoder receives the back scattering electron image, and is used for carrying out feature extraction and downsampling on the back scattering electron image and outputting a multi-scale feature map; The input end of the first decoder is connected with the output end of the encoder and is used for carrying out up-sampling and feature fusion on the multi-scale feature map and outputting a mineral phase segmentation map; and the input end of the second decoder is connected with the output end of the encoder and is used for carrying out up-sampling and feature fusion on the multi-scale feature map and outputting an initial hardness predicted value. In one possible implementation, the method further includes performing a physical constraint modification on the initial hardness prediction value, including: calculating the duty ratio of each mineral phase in the image according to the mineral phase segmentation map; Based on a physical rule, calculating a physical constraint hardness value according to the mineral phase proportion, the hardness priori value and the porosity of each mineral phase; and combining the physical constraint hardness value with the initial hardness predicted value to generate a hardness distribution diagram of the pixel level. In one possible implementation manner, the calcu