Method for generating micromechanical property distribution diagram of solid waste composite cement-based cementing material

Abstract

The invention discloses a method for generating a micromechanical property distribution diagram of a solid waste composite cement-based cementing material, and belongs to the technical field of performance prediction of solid waste composite cement-based cementing materials, and the method comprises the following steps: acquiring BSE, EDS and nanoindentation data of a sample; identifying indentation areas in the nanoindentation mark graph, and generating masks corresponding to the areas; performing statistics on the regional masks to obtain element content data of each indentation region; establishing a prediction model by adopting machine learning based on the regional element content and the elasticity data; performing super-pixel segmentation on the BSE image to match the scale of the indentation area; calculating element features of the super-pixel regions, and predicting the elastic modulus of each super-pixel region; and mapping the predicted elastic modulus to a corresponding region to generate an elastic modulus distribution diagram. According to the method, prediction of the two-dimensional space distribution of the internal elastic modulus of the solid waste composite cement-based cementing material can be realized, and data and visual support are provided for material heterogeneity research and multi-scale modeling.

Description

Technical Field

[0001] This invention belongs to the technical field of performance prediction of solid waste composite cement-based cementitious materials, specifically involving a method for generating microscopic mechanical property distribution maps of solid waste composite cement-based cementitious materials. Background Technology

[0002] The macroscopic mechanical properties and long-term durability of cement-based cementitious materials fundamentally depend on the spatial distribution of their phases and pore structures at the microscale, and the resulting microscopic mechanical properties. Therefore, achieving high-resolution characterization of the spatial distribution of their microscopic mechanical properties (such as elastic modulus) is a crucial step in establishing the relationship between microstructure and macroscopic properties. Nanoindentation technology, as a technique capable of directly measuring the mechanical properties of micro-areas, is widely used in the field of micromechanical characterization. However, its essence is point-by-point measurement, which suffers from low data acquisition efficiency and insufficient spatial coverage. Furthermore, the testing process may adversely affect the structure surrounding the micro-curve, making it difficult to efficiently and quickly obtain large-area spatial distribution maps of mechanical properties.

[0003] The aforementioned problems are even more pronounced for solid waste composite cementitious materials. Due to the complex chemical properties and mineral composition of solid waste admixtures (fly ash, slag, metakaolin, etc.), their reaction products are more diverse, and the spatial heterogeneity of their microstructure is more significant. Therefore, accurately characterizing and predicting the spatial distribution of the micromechanical properties of solid waste composite cementitious materials is of significant theoretical and engineering value for regulating the reaction process and scientifically designing the mix proportions.

[0004] On the other hand, the analytical technique combining backscattered electron imaging (BSE) and energy-dispersive X-ray spectroscopy (EDS) surface scanning can efficiently obtain information on the atomic number and elemental distribution of materials, providing an effective analytical means for phase identification and quantification. Existing research has shown a clear correlation between the elastic modulus of phases in cement-based materials, such as CSH gel, CASH gel, unhydrated cement clinker, and fly ash, and their chemical composition (e.g., calcium-to-silicon ratio, aluminum-to-silicon ratio) and density. This indicates a mapping relationship between the phase type, elemental content, and other characteristics obtained from BSE and EDS analysis and their microscopic mechanical properties. However, existing characterization methods lack effective integration: BSE and EDS can quickly provide large-area composition and phase distribution information, but cannot directly obtain mechanical properties; nanoindentation can accurately measure the mechanical properties of local micro-regions, but it is difficult to achieve large-area coverage.

[0005] Chinese invention patent with publication number CN105241904A discloses a method for phase analysis of fly ash based on energy dispersive X-ray spectroscopy. This patent overlays multiple elemental energy spectrum distribution images of the same test area to determine the phase type of each pixel in the test area. However, the accuracy of this method depends on the effectiveness of the image processing algorithm and has limited quantitative analysis capabilities, especially for complex phases, which limits its practical application.

[0006] Therefore, effectively integrating backscattered electron microscopy, energy dispersive spectroscopy (EDS), and nanoindentation techniques, using a limited but precise set of nanoindentation sites as connecting coordinates, and establishing a machine learning prediction model between micromechanics and its chemical composition and phase content, thereby extrapolating and predicting the distribution of micromechanical characteristics across the entire imaging region, is a feasible approach to solving current characterization challenges. For solid waste composite systems with more complex compositions and structures, developing this method that integrates multimodal data and machine learning prediction is particularly important for efficiently and accurately assessing the spatial distribution of their micromechanical properties. Summary of the Invention

[0007] To address the limitations of existing micromechanical research methods, such as nanoindentation testing primarily using discrete measurement points with limited spatial coverage, which hinders the efficient acquisition of continuous distribution information of the microscopic elastic modulus of materials, and the lack of effective integration between backscattered electron, energy dispersive spectroscopy (EDS), and micromechanical testing methods, resulting in complex overall analysis processes and high technical barriers, this invention aims to provide a method for generating micromechanical property distribution maps of solid waste composite cementitious materials. Based on nanoindentation testing, backscattered electron imaging, and EDS elemental surface scan data, this invention constructs a mapping relationship between regional-scale elemental characteristics and elastic modulus, and extends this mapping relationship to the microstructural scale, thereby achieving overall spatial prediction and visualization of the material's microscopic elastic modulus.

[0008] The objective of this invention is achieved by including the following steps: S1. Obtain backscattered electron images, energy dispersive spectroscopy (EDS) elemental surface spectrum images, and nanoindentation test data of the corresponding regions of solid waste composite cement-based cementitious material samples. S2. Based on the nanoindentation marking image, image processing software is used to identify each nanoindentation region and generate a region mask image corresponding to each nanoindentation point. This transforms discrete nanoindentation test points into standardized spatial analysis regions, achieving accurate conversion of discrete nanoindentation test points into standardized spatial analysis regions and solving the problem that traditional discrete indentation points cannot be used for spatial mechanical analysis. S3. Based on the region mask image from step S2, perform region-by-region statistical analysis on the surface spectrum images of each energy spectrum analysis element, calculate the average gray value of each element in each region, accurately obtain quantitative data on the element content in each region, and achieve precise matching between the element analysis and mechanical testing regions. S4. Using the element content of the region as the input feature and the experimentally measured elastic modulus of the corresponding region as the output target, a mapping model between the element content of the region and the elastic modulus is established using machine learning methods. A prediction model for solid waste composite cement-based cementitious materials is constructed, and the trained elastic modulus prediction model is saved. S5. Based on backscattered electron images, superpixel segmentation is performed on the microstructure of the material to divide the entire micro image into multiple spatially continuous small regions, so that the size of the small regions matches the scale of the nanoindentation region. S6. Calculate the corresponding element features for each superpixel region, and input the element features into the elastic modulus prediction model trained in step S4 to predict the predicted elastic modulus of each superpixel region. S7. Map the predicted elastic modulus of each superpixel region to the corresponding pixel region, so that pixels in the same superpixel region are given the same elastic modulus value. Do not perform interpolation or smoothing, and retain the region boundary information, thereby generating an elastic modulus distribution map at the microscale of the material.

[0009] Preferably, the backscattered electron image acquisition conditions in step S1 are: accelerating voltage 15kV, working distance 11mm; and the energy dispersive spectroscopy elemental surface spectrum image acquisition conditions are: processing time set to 4, dead time not exceeding 20%, scanning parameters per frame 256 µs / pixel, and 4 scanning frames.

[0010] Preferably, in step S2, the edge recognition tool of image processing software (Imagej) is used to adjust the tolerance parameter and accurately identify the reaction edge of the target cementitious particles.

[0011] Preferably, in step S3, the element content is obtained by calculating the average element intensity in each strip mask region for each element's surface spectrum, and finally obtaining a quantitative curve of the element content changing with the strip, and saving it as a CSV.

[0012] Preferably, in step S4, the machine learning method uses a regression model to establish a nonlinear mapping relationship between the regional element content and the elastic modulus, and the regression model is an ensemble learning model based on decision trees.

[0013] Preferably, in step S5, the superpixel segmentation is based on the principles of pixel grayscale similarity and spatial continuity. By adjusting the segmentation quantity parameter, the average number of pixels in a single superpixel region after segmentation is made similar to the average number of pixels in the nano-indentation region.

[0014] Preferably, in step S6, the elemental features of each superpixel region are calculated in a manner consistent with that of the nanoindentation region.

[0015] Preferably, in step S7, the predicted elastic modulus mapping is performed by directly assigning the predicted elastic modulus value of each superpixel region to all pixels in that region, without interpolating or smoothing the elastic modulus between adjacent regions, thereby preserving the region boundary information.

[0016] The solid waste composite cement-based cementitious material is a pure silicate cement, a fly ash composite silicate cement system, a slag composite silicate cement system, or a metakaolin composite silicate cement system. It can also be used in other cementitious material systems whose phase composition can be distinguished by backscattered electron and energy dispersive spectroscopy analysis of elemental surface spectrum images.

[0017] The beneficial effects of this invention are: 1. This invention, by integrating backscattered electron and energy spectrum analysis of elemental surface image, nanoindentation technology and machine learning model, bypasses the limitations of low efficiency and small characterization range of nanoindentation technology, and realizes the rapid generation of large-area micromechanical distribution maps based on limited calibration data, significantly improving the efficiency of obtaining micromechanical features; 2. A quantitative machine learning model was established between chemical composition / phase content and microscopic elastic modulus, clarifying the microscopic intrinsic relationship between "composition-structure-performance", making it theoretically feasible to predict microscopic mechanical distribution characteristics based on backscattered electron and energy spectrum analysis of elemental surface image. 3. By using superpixels that match the size of the nanoindentation points as the basic analysis unit, the accuracy of characterization and analysis are aligned. This ensures the reasonableness of the prediction results while reducing data noise interference and improving analysis efficiency. 4. The method of this invention has strong versatility and is especially suitable for solid waste composite cement-based material systems with complex phases and high heterogeneity, providing a powerful microscopic analysis tool for studying the microscopic composition-structure-performance relationship of materials; 5. This invention, based on nanoindentation marking images, utilizes image processing software to identify each nanoindentation region, generating a region mask image corresponding to each nanoindentation point, thus transforming discrete nanoindentation test points into standardized spatial analysis regions. Based on the aforementioned region mask images, the surface spectrum images of each energy dispersive spectroscopy element are statistically analyzed region by region, calculating the average gray value of each element within each region, accurately obtaining quantitative data on the element content of each region, and achieving precise matching between elemental analysis and mechanical testing regions. Using the region element content as input features and the experimentally measured elastic modulus of the corresponding region as output targets, a dedicated mapping model between the region element content and elastic modulus is established using machine learning methods, and the trained elastic modulus prediction model is saved. Based on the aforementioned mapping model, combined with multi-source image data, a microscopic mechanical property distribution map of solid waste composite cementitious materials is generated. Attached Figure Description

[0018] Figure 1This is a schematic flowchart of the method of the present invention; Figure 2 This is a nanoindentation marking diagram from Example 1 of the present invention; Figure 3 This is a backscattered electron image from Embodiment 1 of the present invention; Figure 4 This is a characteristic X-ray energy spectrum elemental surface image of Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the superpixel grid division of the elemental composite image for energy spectrum analysis in Embodiment 1 of the present invention; Figure 6 This is a micromechanical thermogram of Embodiment 1 of the present invention; Figure 7 This is a backscattered electron image from Embodiment 2 of the present invention; Figure 8 This is a characteristic X-ray energy spectrum elemental surface image of Embodiment 2 of the present invention; Figure 9 This is a micromechanical thermogram of Embodiment 2 of the present invention; Figure 10 This is a backscattered electron image from Embodiment 3 of the present invention; Figure 11 This is a characteristic X-ray energy spectrum elemental surface image of Embodiment 3 of the present invention; Figure 12 This is a micromechanical thermogram of Embodiment 3 of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the embodiments and accompanying drawings, but this does not limit the present invention in any way. Any changes or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0020] Example 1 To verify the feasibility of the method of this invention, this embodiment uses a composite cementitious material sample as the solid waste composite cementitious material. In this sample, metakaolin accounts for 30% of the total mass of the cementitious material, and the water-cement ratio is 0.4. Specifically, the composite components are mixed evenly with cement in a certain proportion, and then water is added and stirred in a paste mixer at 120 r / min for 5 min to prepare metakaolin-cement composite paste. After stirring, the paste is poured into a mold and manually shaken slightly for 30 s to remove air bubbles. After curing at room temperature for 24 h, the sample is demolded and then placed in a standard curing room (temperature 20 ± 2 ℃, relative humidity ≥95%) until the sample reaches the specified curing age.

[0021] Samples were taken from the center of the cement paste, with a sample size of approximately 10 mm × 10 mm × 10 mm. These samples were immersed in isopropanol for 7 days, with the solution changed at regular intervals to terminate the cement hydration reaction and maintain microstructural stability. Subsequently, the samples were thoroughly dried in a 40 °C drying oven until weight stability was achieved. The dried samples were then placed in a vacuum storage chamber to prevent carbonization and dust contamination. To enhance the porosity contrast in backscattered electron imaging and avoid damaging the microstructure through polishing, the samples underwent vacuum impregnation and embedding treatment with low-viscosity epoxy resin.

[0022] The samples were coarsely polished sequentially using 320, 500, 1200, and 2400 grit sandpaper, followed by progressive fine polishing using diamond suspensions of 9.00, 3.00, 1.00, and 0.25 μm. The polishing grit size, time, and pressure were as follows: 9 μm—5 min / 12 N, 3 μm—15 min / 15 N, 1 μm—30 min / 20 N, and 0.25 μm—3 h / 25 N. The samples were then sputter-coated with gold.

[0023] As attached Figure 1 As shown in this embodiment, the method for generating the microscopic mechanical property distribution map of solid waste composite cementitious material includes the following steps: S1. Acquire backscattered electron images, energy dispersive spectroscopy (EDS) images of the solid waste composite cementitious material sample, as well as nanoindentation test data of the corresponding regions. Field emission scanning electron microscopy (SEM) was used to acquire images in backscattered mode under high vacuum conditions. Acquisition parameters were set as follows: accelerating voltage 15 kV, working distance 11 mm, and resolution of both backscattered electron and EDS images at 1024 pixels × 704 pixels. An Oxford Instruments Xplore30 EDS spectrometer was used to acquire elemental spectra, with a scan time of 256 μs / pixel, 4 collection frames, 1024 channels, a processing time of 4 seconds, a dead time of less than 20%, and a characteristic X-ray output count rate stable at 80,000–100,000 cps per second. Collected elements included silicon (Si), aluminum (Al), calcium (Ca), magnesium (Mg), iron (Fe), sulfur (S), and oxygen (O). The collected elemental surface spectra are first saved as qualitative images, and then quantitatively processed through the built-in Quantmap function to output quantitative elemental surface spectra data in terms of atomic percentages for subsequent heat map analysis. By pressing a nano-indentation indenter into the polished surface of a cement slurry sample and simultaneously recording the relationship between load and indentation depth during the loading process, a load-displacement response curve is obtained, thereby characterizing the mechanical response features of different micro-regions of the hydrated cement slurry. Based on this load-displacement curve, the micromechanical property parameters of the material's micro-regions can be obtained. A regularly distributed array of indentations was selected on the surface of each sample for testing. 225 indentation points (15×15 array) were measured at 15µm intervals on each sample surface. The maximum indentation displacement h_max = 250 nm, the loading rate was 0.5 mN / s, the holding force was 60 seconds, and the unloading rate was 0.5 mN / s. Based on the indentation loading-unloading curves, the test data were analyzed using classical nanoindentation theory to calculate micromechanical parameters such as indentation modulus and hardness. S2. Based on the nanoindentation marker images, image processing software is used to identify each nanoindentation region and generate a region mask image corresponding to each nanoindentation point, thus transforming the discrete nanoindentation test points into standardized spatial analysis regions. Since nanoindentation test points are discretely distributed at the microscale, directly performing subsequent elemental and mechanical feature analysis based on a single pixel or point is easily affected by factors such as inconsistent regional scales and unclear boundaries, making it difficult to achieve comparability between different regions. To solve the above problems, this embodiment identifies the nanoindentation marker regions and generates corresponding region mask images, transforming the discrete nanoindentation test points into standardized analysis regions with clear spatial ranges, thereby providing a unified spatial benchmark for subsequent quantitative statistics and modeling analysis. This embodiment specifically uses ImageJ software to process the nanoindentation marking images. First, the nanoindentation marking images are imported into ImageJ software. The Magic Wand tool in ImageJ is used to identify the nanoindentation marking regions. By adjusting the tolerance parameter, the nanoindentation marking regions are clearly distinguished from the surrounding background regions. Then, each nanoindentation marking region is selected sequentially and converted into a binary mask image. The generated nanoindentation region mask images are saved in a predetermined order, ensuring that each mask image corresponds one-to-one with the corresponding nanoindentation test point. This completes the spatial standardization processing of the nanoindentation regions, providing accurate spatial constraints for subsequent regional elemental content statistics, phase content analysis, and elastic modulus modeling. S3. Based on the region mask image from step S2, perform region-by-region statistical analysis on the surface spectrum images of each energy dispersive spectroscopy element, calculate the average gray value of each element in each region, and obtain quantitative data on the element content of each region. Since there are many nanoindentation regions, and the distribution characteristics of multiple elements need to be statistically analyzed simultaneously in each region, manual or region-by-region measurement methods are inefficient and prone to introducing human error, making it difficult to meet the requirements of batch and consistent analysis. To solve the above problems, this embodiment adopts an automated region element statistical method based on the region mask to achieve efficient and uniform extraction of element content in each region. In this embodiment, the nanoindentation region mask images generated in step S2, along with the corresponding elemental surface spectrum images for energy dispersive spectroscopy (EDS), are processed as input data. Each region mask image defines the analysis area in binary form, and each elemental surface spectrum image represents the distribution intensity of the corresponding element at the microscale in grayscale form. While ensuring consistency in size and spatial location between the region mask images and the elemental surface spectrum images, each region mask is read sequentially and applied to the corresponding elemental surface spectrum image. By automatically calculating the pixel grayscale values ​​within the foreground range of each mask region, the average grayscale value of each element within the corresponding region is obtained, serving as a quantitative characterization parameter for the element content of that region. Following the region numbering order, the statistical analysis of the element content of each element is performed sequentially for all nanoindentation regions, and the statistical results are summarized into a regional element content data table. This achieves a quantitative expression of element content variations across spatial regions, providing standardized input data for subsequent regional elastic modulus modeling. S4. Using the elemental content of the region as input features and the experimentally measured elastic modulus of the corresponding region as the output target, a regression model (an ensemble learning model based on decision trees) machine learning method is used to establish a mapping model between the elemental content and elastic modulus of the region, and the trained elastic modulus prediction model is saved. Since the mechanical properties of cementitious materials at the microscale are significantly affected by the chemical composition characteristics within the region, and there is a complex nonlinear relationship between the content of different elements and the regional elastic modulus, traditional empirical formulas or single-factor analysis methods are difficult to accurately reflect the variation law of the regional elastic modulus. Therefore, this embodiment constructs a regional elastic modulus prediction model based on the aforementioned quantitative statistics of regional elemental content, realizing a quantitative mapping between microscopic chemical composition information and mechanical properties. This embodiment uses the elemental content data of each region obtained in step S3 as the model input features, where the elemental content parameters are used to characterize the chemical composition characteristics within the region; the elastic modulus obtained from the nanoindentation test of the corresponding region is used as the model output target value. By establishing a one-to-one correspondence between the elemental content data and the elastic modulus data of the same region, a regional sample dataset is constructed. Specifically, the multi-element content parameters of each region are used as input feature vectors, and the elastic modulus obtained from nanoindentation tests in the corresponding regions is used as the output target value. Specifically, this embodiment employs an ensemble learning regression model, XGBoost, based on gradient boosting decision trees, to establish a nonlinear mapping relationship between regional element content and elastic modulus. This model continuously optimizes the prediction error through iterative training with multiple decision trees, thereby improving the model's ability to fit complex nonlinear relationships. This establishes a quantitative mapping model between regional chemical composition characteristics and mechanical properties, enabling it to output corresponding elastic modulus prediction values ​​based on regional element content data, thus achieving the predictive characterization of the elastic modulus in regions where nanoindentation testing has not been performed. S5. Based on backscattered electron images, superpixel segmentation is performed on the microstructure of the material, dividing the entire microscopic image into multiple spatially continuous small regions, so that the size of the small regions matches the scale of the nanoindentation region. Since nanoindentation tests can only obtain mechanical property information at the microscale of the material at finite discrete locations, it is difficult to guarantee spatial continuity and scale consistency if the discrete test results are directly used for the mechanical property characterization of the entire microscopic image. Therefore, this embodiment introduces a superpixel segmentation method based on backscattered electron images to spatially reconstruct the microstructure of the material, providing a regional basis for the subsequent establishment of the microscopic mechanical property field. The invention uses backscattered electron images as input objects and employs a superpixel segmentation method to process the entire microscopic image, dividing the image into multiple spatially continuous small regions with clear boundaries. The superpixel segmentation process ensures that the gray-level characteristics within the regions are relatively consistent, while making each small region spatially closely fit the boundary of the material's true microstructure. By appropriately setting the superpixel segmentation parameters, the spatial size of the small region is matched with the scale of the nanoindentation region, thereby ensuring that the subsequent regional elastic modulus prediction model based on nanoindentation data can be effectively applied at the superpixel scale. In this way, the entire backscattered electron image is transformed into a set of regions composed of multiple spatial analysis units consistent with the nanoindentation scale. To facilitate a clear and intuitive explanation of the spatial division of the superpixel region, this invention provides a schematic representation of the superpixel region on an elemental composite image obtained through energy dispersive spectroscopy analysis, as shown below. Figure 5 As shown, image segmentation algorithms can divide a microscopic image into multiple spatially continuous small regions with similar internal features. Each region serves as the basic analysis unit for subsequent elastic modulus prediction. It should be noted that... Figure 5 The number of superpixels shown is for illustrative purposes only. In actual analysis, a larger number of superpixel regions can be set according to the image resolution and the scale of the nanoindentation region. S6. Calculate the elemental characteristics of each superpixel region in the same manner as the nanoindentation region. Then, input the calculated elemental characteristics into the elastic modulus prediction model trained in step S4 to predict the elastic modulus of each superpixel region. After completing the construction of the regional elastic modulus prediction model and the superpixel segmentation of the entire backscattered electron image, in order to achieve continuous characterization of the elastic modulus at the microscale of the material, this embodiment applies the prediction model to the superpixel scale region to perform elastic modulus mapping on the entire microscopic image. Using each superpixel region obtained in step S5 as the basic spatial unit, extract the elemental content feature parameters of the corresponding superpixel region based on its spatial position in the backscattered electron image. Substitute the elemental content features as input into the regional elastic modulus prediction model constructed in step S4 to obtain the predicted elastic modulus value corresponding to each superpixel region. The predicted elastic modulus values ​​of each superpixel region are backfilled into the corresponding superpixel region according to their spatial location, thereby achieving a continuous distribution representation of the elastic modulus across the entire backscattered electron image. In this way, the discrete mechanical property information originally obtained only at the nanoindentation scale is extended to the overall elastic modulus distribution at the microstructural scale, providing a foundation for the spatial analysis of the micromechanical properties of materials. S7. The predicted elastic modulus value of each superpixel region is directly assigned to all pixels in that region without interpolation or smoothing of the elastic modulus between adjacent regions, thereby preserving the region boundary information and assigning the same elastic modulus value to pixels in the same superpixel region, thereby generating an elastic modulus distribution map at the microscale of the material. After obtaining the predicted elastic modulus values ​​for each superpixel region in the entire microscopic image, this embodiment further maps the predicted results back to the pixel scale to construct a micromechanical thermogram of the material, in order to achieve an intuitive characterization of the material's micromechanical properties. Specifically, using the superpixel segmentation results obtained in step S6 as the basis for spatial mapping, the predicted elastic modulus values ​​for each superpixel region obtained in step S6 are assigned to all pixels within the corresponding superpixel region, ensuring that pixels within the same superpixel region have consistent elastic modulus values. In this way, the true microstructural boundary features reflected by the superpixel segmentation are preserved without introducing spatial interpolation or smoothing.

[0024] Furthermore, the obtained pixel-level elastic modulus distribution results are visualized. A unified color mapping scale is used to represent different elastic modulus intervals with different colors, generating a continuous micromechanical thermogram to intuitively reflect the spatial distribution characteristics of the elastic modulus inside the material. The results are as follows Figure 3 The image shown is a backscattered electron image obtained in this embodiment. Figure 4 The image shown is the corresponding characteristic X-ray energy dispersive spectroscopy elemental surface spectrum image; as shown... Figure 6 The corresponding micromechanical thermogram is shown below.

[0025] Example 2 The method for generating the microscopic mechanical property distribution diagram of the solid waste composite cementitious material in this embodiment is based on Example 1, but differs from Example 1 in that the analytical object is different. This embodiment analyzes a cement-fly ash composite cement paste sample, with fly ash accounting for 30% of the total mass of the cementitious material; the results are as follows. Figure 7 The image shown is a backscattered electron image obtained in this embodiment. Figure 8 The image shown is the corresponding characteristic X-ray energy dispersive spectroscopy elemental surface spectrum image; Figure 9 This is the corresponding micromechanical thermogram.

[0026] Example 3 The method for generating the microscopic mechanical property distribution diagram of the solid waste composite cement-based cementitious material in this embodiment is based on Example 1, but differs from Example 1 in that the analytical object is different. This embodiment analyzes a cement-mineral powder composite cement paste sample, with mineral powder accounting for 30% of the total mass of the cementitious material; the results are as follows. Figure 10 The image shown is a backscattered electron image obtained in this embodiment. Figure 11 The image shown is the corresponding characteristic X-ray energy dispersive spectroscopy elemental surface spectrum image; as shown... Figure 12 The corresponding micromechanical thermogram is shown below.

Claims

1. A method for generating a microscopic mechanical property distribution map of solid waste composite cementitious materials, characterized in that, Includes the following steps: S1. Obtain backscattered electron images, energy dispersive spectroscopy (EDS) elemental surface spectrum images, and nanoindentation test data of the corresponding regions of solid waste composite cement-based cementitious material samples. S2. Based on the nanoindentation marking image, image processing software is used to identify each nanoindentation region and generate a region mask image corresponding to each nanoindentation point, so that the discrete nanoindentation test points are transformed into standardized spatial analysis regions. S3. Based on the region mask image from step S2, perform region-by-region statistical analysis on the surface spectrum images of each energy spectrum analysis element, calculate the average gray value of each element in each region, and obtain quantitative data on the element content in each region. S4. Using the element content of the region as the input feature and the experimentally measured elastic modulus of the corresponding region as the output target, a mapping model between the element content of the region and the elastic modulus is established using machine learning methods, and the trained elastic modulus prediction model is saved. S5. Based on backscattered electron images, superpixel segmentation is performed on the microstructure of the material to divide the entire micro image into multiple spatially continuous small regions, so that the size of the small regions matches the scale of the nanoindentation region. S6. Calculate the corresponding element features for each superpixel region, and input the element features into the elastic modulus prediction model trained in step S4 to predict the predicted elastic modulus of each superpixel region. S7. Map the predicted elastic modulus of each superpixel region to the corresponding pixel region, so that pixels in the same superpixel region are given the same elastic modulus value, thereby generating an elastic modulus distribution map at the microscale of the material.

2. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, The backscattered electron image acquisition conditions for step S1 are: accelerating voltage 15kV, working distance 11mm; the energy dispersive spectroscopy (EDS) elemental surface spectrum image acquisition conditions are: processing time set to 4, dead time not exceeding 20%, scanning parameters per frame 256 µs / pixel, and 4 scanning frames.

3. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, In step S2, the edge recognition tool of the image processing software is used to adjust the tolerance parameters and accurately identify the reaction edge of the target cementitious particles.

4. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, In step S3, the element content is obtained by calculating the average element intensity in each strip mask region for each element's surface spectrum, and finally obtaining a quantitative curve of the element content changing with the strip, which is then saved as a CSV.

5. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, In step S4, the machine learning method uses a regression model to establish a nonlinear mapping relationship between the regional element content and the elastic modulus. The regression model is an ensemble learning model based on decision trees.

6. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, In step S5, the superpixel segmentation is based on the principles of pixel grayscale similarity and spatial continuity. By adjusting the segmentation quantity parameter, the average number of pixels in a single superpixel region after segmentation is made similar to the average number of pixels in the nano-indentation region.

7. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, In step S6, elemental features are calculated for each superpixel region in the same manner as for the nanoindentation region.

8. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, In step S7, the predicted elastic modulus mapping directly assigns the predicted elastic modulus value of each superpixel region to all pixels within that region, without interpolating or smoothing the elastic modulus between adjacent regions, thus preserving the region boundary information.

9. The method for generating the microscopic mechanical property distribution map of the solid waste composite cementitious material according to claim 1, characterized in that, The solid waste composite cement-based cementitious material is a pure silicate cement, a fly ash composite silicate cement system, a slag composite silicate cement system, or a metakaolin composite silicate cement system.

Images