Method and device for detecting performance of solid waste-based cementitious material stabilized aggregate base and medium

Abstract

The application discloses a kind of solid waste-based cementitious material stable aggregate base performance detection method, device and medium, comprising: collecting the multi-modal original detection data of solid waste-based cementitious material stable aggregate base, based on multi-modal original detection data, construct multi-modal detection data set;Based on the preset deep learning network, the multi-modal detection data set is carried out feature extraction, and the multi-dimensional performance characterization feature is obtained, the multi-dimensional performance characterization feature is input into the trained gate expert network, and the performance prediction mean and performance fluctuation mean corresponding to the multiple road performances of solid waste-based cementitious material stable aggregate base are obtained;Based on performance prediction mean and performance fluctuation mean, the performance reliability analysis of solid waste-based cementitious material stable aggregate base is carried out, and the standardized performance test result is generated and output.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method, apparatus, and medium for testing the performance of solid waste-based cementitious material stabilized aggregate base course. Background Technology

[0002] Solid waste-based cementitious aggregate stabilized base courses, as the core structural form for the resource utilization of solid waste in road engineering, have been widely used in highway construction at all levels in my country. However, due to the large dispersion and poor batch stability of industrial solid waste components such as steel slag, slag, and fly ash, the road performance of their base courses is significantly affected by the coupling of material composition, preparation process, and service environment. In the existing technology, the performance testing of solid waste-based cementitious aggregate stabilized base courses mostly relies on traditional indoor mechanical tests, which have the disadvantages of long testing cycles, high labor costs, and the ability to obtain discrete data of a single index. Moreover, the existing data processing methods have not achieved effective fusion of multi-source heterogeneous test data, and cannot quantify the uncertainty in the performance prediction process, resulting in insufficient accuracy and reliability of test results, making it difficult to meet the needs of rapid on-site testing and full life cycle performance evaluation in engineering projects.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, and storage medium for testing the performance of stabilized aggregate base courses based on solid waste cementitious materials, thereby at least solving the technical problems in the prior art of testing the performance of stabilized aggregate base courses based on solid waste cementitious materials, such as difficulties in fusing multi-source heterogeneous data, insufficient extraction of deep performance characteristics, inability to quantify performance prediction uncertainties, and low reliability of test results.

[0005] According to one aspect of the present invention, in order to achieve the above-mentioned objective, a method for testing the performance of a solid waste-based cementitious material stabilized aggregate base course is provided, comprising: Multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course were collected, and a multimodal test dataset was constructed based on the multimodal raw test data; Feature extraction is performed on the multimodal detection dataset based on a pre-set deep learning network to obtain multi-dimensional performance characterization features. The multi-dimensional performance characterization features are then input into a pre-trained gated expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances of the corresponding solid waste-based cementitious aggregate base course. Based on the performance prediction mean and performance fluctuation mean, a reliability analysis of the performance of stabilized aggregate base course of solid waste-based cementitious materials is conducted, and standardized performance test results are generated and output.

[0006] Furthermore, before constructing the multimodal detection dataset, automated preprocessing is performed on the original multimodal detection data. The automated preprocessing includes outlier truncation, missing value completion, unit normalization, and time series data denoising.

[0007] Furthermore, outlier truncation employs a robust Z-score and interquartile range method, traversing the multimodal raw detection data to identify outliers and truncates them to a preset reasonable boundary; missing value completion uses the median interpolation method of neighboring samples to complete the missing data caused by signal interruption.

[0008] Furthermore, feature extraction is performed on the multimodal detection dataset based on a pre-defined deep learning network to obtain multi-dimensional performance characterization features, specifically including: Graph structure modeling of aggregate gradation curve data is performed based on graph attention network to obtain the stability characteristics of aggregate skeleton; The mechanical response evolution characteristics are obtained by processing the time-series loading data of mechanical properties using a one-dimensional convolutional neural network. The base section image data is segmented at the pixel level based on the instance segmentation model to obtain the internal void distribution characteristics. By splicing and integrating the stability characteristics, mechanical response evolution characteristics, and internal void distribution characteristics of the aggregate skeleton, a multi-dimensional performance characterization feature is obtained.

[0009] Furthermore, the instance segmentation model is an improved YOLOv8seg model, which replaces the spatial pyramid pooling fast module in the YOLOv8seg model with a separable kernel attention pyramid module; the internal void distribution features extracted based on the improved YOLOv8seg model include void ratio, void area, void perimeter, maximum void width, and void roundness.

[0010] Furthermore, the multi-dimensional performance characterization features are input into a trained gated expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics of the solid waste-based cementitious aggregate base course, specifically including: Multi-dimensional performance characterization features are concatenated with scene embedding features to obtain comprehensive input features. Scene embedding features are obtained by vectorizing target layer, traffic level, and environmental zoning parameters. Calculate the fit weights of multiple gating experts based on comprehensive input features; By using the fit degree weighting to perform weighted fusion of the exclusive characterizations output by multiple gating experts, the average performance prediction and average performance fluctuation of multiple road performances of the corresponding solid waste-based cementitious aggregate base course are obtained.

[0011] Furthermore, a reliability analysis of the performance of the stabilized aggregate base course based on solid waste-based cementitious materials was conducted based on the average performance prediction and the average performance fluctuation, specifically including: The reliability index of various road performance characteristics of solid waste-based cementitious material stabilized aggregate base course is calculated based on the performance prediction mean and the performance fluctuation mean. The reliability index is compared with the preset reliability threshold to obtain the reliability judgment result of the stable aggregate base course of solid waste-based cementitious material. Based on the reliability index and preset performance grading thresholds, a weighted comprehensive scoring algorithm is used to evaluate the overall road performance of solid waste-based cementitious aggregate base courses.

[0012] Furthermore, after generating and outputting standardized performance test results, the standardized performance test results are compared with the actual measured performance data on site. Based on the comparison results, the parameters of the preset deep learning network and the gated expert network are incrementally updated using the maximum likelihood loss function.

[0013] According to one embodiment of the present invention, a performance testing device for solid waste-based cementitious aggregate base course is also provided, comprising: a data acquisition module for acquiring multimodal raw test data of the solid waste-based cementitious aggregate base course and constructing a multimodal test dataset based on the multimodal raw test data; an extraction module for extracting features from the multimodal test dataset based on a preset deep learning network to obtain multidimensional performance characterization features, inputting the multidimensional performance characterization features into a trained gated expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics of the solid waste-based cementitious aggregate base course; and an analysis module for performing performance reliability analysis of the solid waste-based cementitious aggregate base course based on the performance prediction mean and performance fluctuation mean, and generating and outputting standardized performance test results.

[0014] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of the present invention.

[0015] In this embodiment of the invention, by collecting multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course and constructing a unified dataset, the standardized integration and unified management of multi-source heterogeneous test data are achieved. A pre-set deep learning network is used to extract multi-dimensional performance characterization features, fully exploring the intrinsic correlation between aggregate structural characteristics, mechanical response laws, and base course road performance. A gated expert network simultaneously outputs the predicted mean and fluctuation mean of multiple road performance parameters, quantifying the material dispersion and experimental uncertainty during the testing process. Performance reliability analysis is performed based on the predicted mean and fluctuation mean, significantly improving the accuracy and reliability of the test results, greatly shortening the testing cycle, and reducing testing costs. This solves the technical problems in the existing technology of solid waste-based cementitious material stabilized aggregate base course performance testing, such as difficulty in fusing multi-source heterogeneous data, insufficient extraction of deep performance features, inability to quantify performance prediction uncertainty, and low reliability of test results. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart of a method for testing the performance of a solid waste-based cementitious material stabilized aggregate base course according to one embodiment of the present invention; Figure 2 This is a structural block diagram of a solid waste-based cementitious material stabilized aggregate base performance testing device according to one embodiment of the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] According to an embodiment of the present invention, an embodiment of a method for testing the performance of a solid waste-based cementitious material stabilized aggregate base course is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0020] This method embodiment can be executed in an electronic device or similar computing device that includes a memory and a processor. Taking operation on a vehicle terminal as an example, the vehicle terminal may include one or more processors (processors may include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), digital signal processing (DSP) chips, microcontroller units (MCUs), field-programmable gate arrays (FPGAs), neural network processors (NPUs), tensor processors (TPUs), artificial intelligence (AI) type processors, etc.) and a memory for storing data. Optionally, the vehicle terminal may also include transmission devices, input / output devices, and display devices for communication functions. Those skilled in the art will understand that the above structural description is merely illustrative and does not limit the structure of the vehicle terminal. For example, the vehicle terminal may include more or fewer components than described above, or have a different configuration than described above.

[0021] The memory can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the solid waste-based cementitious material stabilized aggregate base performance testing method in this embodiment of the invention. The processor executes various functional applications and data processing by running the computer program stored in the memory, thereby realizing the aforementioned solid waste-based cementitious material stabilized aggregate base performance testing method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to a mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0022] The transmission device is used to receive or send data via a network. Specific examples of the network mentioned above may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0023] Display devices can be, for example, touchscreen liquid crystal displays (LCDs) and touch displays (also referred to as "touchscreens" or "touch displays"). The LCD allows users to interact with the user interface of the mobile terminal. In some embodiments, the mobile terminal has a graphical user interface (GUI), which allows users to interact with the GUI through finger contact and / or gestures on a touch-sensitive surface. Optional human-computer interaction functions include: creating web pages, drawing, word processing, creating electronic documents, playing games, video conferencing, instant messaging, sending and receiving emails, call interfaces, playing digital video, playing digital music, and / or web browsing, etc. Executable instructions for performing the above human-computer interaction functions are configured / stored in one or more processor-executable computer program products or readable storage media.

[0024] Figure 1 This is a flowchart of a method for testing the performance of a solid waste-based cementitious material stabilized aggregate base course according to one embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps: Step S110: Collect multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course. Based on the multimodal raw test data, construct a multimodal test dataset. The specific content is as follows: In step S110, multimodal raw test data of solid waste-based cementitious material stabilized aggregate base layer is first collected. The multimodal raw test data covers two major categories: indoor standard specimen test data and field physical base layer test data. It comprehensively covers seven dimensions: material composition parameters, preparation process parameters, service condition parameters, mechanical property time-series sensing data, deformation performance sensing data, durability performance test data, and base layer cross-sectional image data. The material composition parameters include the proportion, activity index, and fineness modulus of solid waste-based cementitious materials such as steel slag, blast furnace slag, and fly ash, as well as the gradation curves, crushing values, and needle-like and flaky content of coarse and fine aggregates; the preparation process parameters include mixing time, mixing temperature, compaction work, compaction degree, and curing age; the service condition parameters include the number of freeze-thaw cycles, the number of wet-dry cycles, ambient temperature, ambient humidity, and base layer moisture content; the mechanical property time-series sensor data are continuously collected by pressure sensors and displacement sensors at a fixed sampling frequency of 100Hz, including the load-displacement time-series curve of the entire unconfined compressive strength loading process, the load-deformation time-series curve of the entire splitting tensile strength loading process, and the step-by-step loading and unloading time-series curve of the resilient modulus test; the deformation performance sensor data includes continuous displacement monitoring data of the drying shrinkage test and synchronous temperature deformation monitoring data of the thermal shrinkage test; the durability performance test data includes the mass loss rate of the erosion resistance test and the strength retention rate after freeze-thaw cycles.

[0025] For the acquisition of cross-sectional image data of the base course, an industrial CMOS camera with a resolution of 2448×2048 was used in conjunction with a ring-shaped shadowless light source. Before acquisition, the solid waste-based cementitious aggregate base course specimens were cut into standard cross-sections with a thickness of 50mm perpendicular to the compaction direction. The cross-sections were then sanded to remove surface laitance and loose particles, ensuring that the boundaries between the aggregate particles and the cementitious matrix were clearly distinguishable. During imaging, the camera lens was vertically aimed at the base course cross-section, with the distance between the lens and the cross-section fixed at 300mm. Three images with different lighting intensities were acquired for each cross-section. Local reflection interference was eliminated by image overlay, ultimately obtaining cross-sectional image data that accurately reflects the aggregate distribution, void morphology, and hydration state of the cementitious material within the base course.

[0026] After the acquisition of multimodal raw detection data is completed, automated preprocessing is performed on the multimodal raw detection data. The automated preprocessing includes outlier truncation, missing value completion, dimension normalization, and time series data denoising. The outlier truncation method combines robust Z-score and interquartile range. It iterates through all numerical fields of the original multimodal detection data, calculates the robust Z-score for each data point, and determines the data point as an outlier when the absolute value of the robust Z-score is greater than 3. The outlier is truncated to a preset reasonable boundary of the corresponding parameter, rather than directly deleting the entire sample, thus avoiding the loss of effective data due to the inherent discreteness of solid waste materials. Missing value completion uses the median interpolation method. For missing data with a continuous length of no more than 5 sampling points caused by signal interruption or equipment failure during the acquisition process, the median of the 3 nearest neighboring samples with the closest preparation parameters and operating parameters is selected for completion. Dimensional normalization uses an extreme value normalization algorithm to map all numerical parameters to the [0,1] interval, eliminating the impact of differences in the dimensions and orders of magnitude of different parameters on subsequent data processing. Temporal data denoising uses a moving average filtering algorithm with a window length of 7 to smooth the mechanical performance temporal sensing data, filter out high-frequency random noise from the sensor, and retain the true trend of changes in the mechanical properties of the base layer.

[0027] Based on the preprocessed multimodal raw detection data, a multimodal detection dataset is constructed. First, a unique sample number is assigned as the primary key to each detection sample. The material composition parameters, preparation process parameters, service condition parameters, mechanical property time-series data, deformation performance data, durability performance data, and base section image data corresponding to each sample are structurally encapsulated, establishing a one-to-one correspondence between sample data and sample number. Then, all samples are classified and stored according to three dimensions: sample number, curing age, and service condition type. A relational database is used to store structured numerical parameters, while a distributed file system is used to store unstructured time-series and image data. Storage path indexes for time-series and image data are established in the database to enable rapid retrieval and access to multimodal data. Finally, the multimodal detection dataset is divided into a training set, a validation set, and a test set in a 7:2:1 ratio. The training set is used for parameter training of the subsequent deep learning network, the validation set is used for hyperparameter tuning during network training, and the test set is used to evaluate the final performance of the network.

[0028] Step S120: Based on a preset deep learning network, feature extraction is performed on the multimodal detection dataset to obtain multi-dimensional performance characterization features. These multi-dimensional performance characterization features are then input into a trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics of the corresponding solid waste-based cementitious aggregate base course. The specific details are as follows: In step S120, feature extraction is performed on the multimodal detection dataset based on a preset deep learning network to obtain multi-dimensional performance representation features. In this embodiment, the preset deep learning network adopts a multi-branch heterogeneous architecture, designing dedicated feature extraction branches for the inherent characteristics of different modal data, thereby achieving deep feature mining and unified representation of multi-source heterogeneous data. The preset deep learning network includes a graph attention network branch, a one-dimensional convolutional neural network branch, and an improved instance segmentation network branch. These three branches process the corresponding data types in the multimodal detection dataset in parallel, outputting performance-related features of different dimensions respectively.

[0029] Based on a graph attention network, graph structure modeling is performed on the aggregate gradation curve data in the multimodal detection dataset to obtain aggregate skeleton stability features. First, the aggregate gradation curve is transformed into an undirected graph structure. The mass proportion of aggregates in different particle size ranges is used as the graph node features, and the difference in mass proportion between adjacent particle size ranges is used as the edge weights to construct a particle size relationship graph that characterizes the overlap relationship between aggregate particles. This particle size relationship graph is input into a graph attention network containing three graph attention layers. Through the attention mechanism, the contribution weights of different particle size ranges to the stability of the aggregate skeleton are automatically learned, focusing on capturing the main skeleton structure formed by coarse aggregates and the filling effect of fine aggregates on the skeleton voids. The output of the graph attention network is a 128-dimensional feature vector, which is the aggregate skeleton stability feature. This feature can quantitatively reflect the skeleton bearing capacity and structural integrity of the stable aggregate base layer of solid waste-based cementitious materials.

[0030] The mechanical performance time-series loading data in the multimodal detection dataset is processed using a one-dimensional convolutional neural network (CNN) to obtain mechanical response evolution features. The CNN comprises four convolutional blocks, each consisting of a one-dimensional convolutional layer, a batch normalization layer, and a ReLU activation function layer. The kernel sizes are 16, 8, 4, and 2, respectively, with a stride of 1 for each. Load-displacement and load-deformation time-series data obtained from unconfined compressive strength, splitting tensile strength, and resilient modulus tests are input into the CNN. Multi-scale convolution operations are used to extract mechanical response features at different loading stages, capturing the complete evolution of the base specimen from elastic deformation to plastic failure during loading. The output of the CNN is a 256-dimensional feature vector, representing the mechanical response evolution features. These features reflect the strength development characteristics, deformation modulus, and failure toughness of the base material.

[0031] The improved YOLOv8seg instance segmentation model is used to perform pixel-level segmentation of the base section image data in the multimodal detection dataset to obtain the internal void distribution features. In this embodiment, a separable kernel attention pyramid module replaces the spatial pyramid pooling fast module in the original YOLOv8seg model, enhancing the model's ability to extract fine-grained void features. The base section image is input into the improved YOLOv8seg model, which automatically segments the aggregate particles, cementitious material matrix, and internal void regions in the image, outputting a pixel-level mask for each void. Based on the segmentation results, quantitative parameters such as porosity, average void area, void perimeter distribution, maximum void width, and void roundness are calculated. These parameters are then input into a fully connected layer for feature compression, resulting in a 128-dimensional internal void distribution feature. This feature comprehensively reflects the microstructural features and density of the base layer.

[0032] The aggregate skeleton stability characteristics, mechanical response evolution characteristics, and internal void distribution characteristics are spliced ​​and fused to obtain multi-dimensional performance characterization features. Specifically, the 128-dimensional aggregate skeleton stability characteristics, 256-dimensional mechanical response evolution characteristics, and 128-dimensional internal void distribution characteristics are spliced ​​in the channel dimension to form a 512-dimensional comprehensive feature vector, which is the multi-dimensional performance characterization feature. This feature simultaneously integrates the macroscopic structural characteristics, mechanical response laws, and microscopic void distribution information of the base layer, and can comprehensively characterize the comprehensive road performance of the solid waste-based cementitious material stabilized aggregate base layer.

[0033] The multi-dimensional performance characterization features are input into a trained gated expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics corresponding to the solid waste-based cementitious aggregate base course. In this embodiment, the gated expert network contains four different types of gated experts, corresponding to four domains: mechanical performance, deformation performance, durability performance, and comprehensive performance. Each gated expert consists of two fully connected layers with 256 hidden layer nodes and ReLU activation function. First, the multi-dimensional performance characterization features are concatenated with the scene embedding features to obtain the comprehensive input features. The scene embedding features are obtained by vectorizing three discrete parameters—target layer, traffic level, and environmental zoning—through the embedding layer, resulting in a 64-dimensional feature.

[0034] Based on the comprehensive input features, the fitness weights of multiple gating experts are calculated. A fitness score for each gating expert relative to the current detection sample is calculated using a gating network. This fitness score is obtained by performing a matrix operation on the concatenated vector of multi-dimensional performance representation features and scene embedding features, along with the corresponding gating expert's weight vector, and then adding a bias term. The fitness scores of all gating experts are normalized using the Softmax function to obtain the fitness weight for each gating expert. A temperature coefficient of 1.0 is set to control the sharpness of the weight distribution, ensuring that experts with higher fitness scores receive a higher weight proportion.

[0035] The specific representations output by multiple gated experts are weighted and fused using the fitness weights to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics corresponding to the solid waste-based cementitious aggregate base course. Each gated expert outputs a 256-dimensional specific representation. The specific representations of all gated experts are weighted and summed according to their corresponding fitness weights to obtain the final fused feature. The fused feature is input into multiple independent task output heads. Each task output head corresponds to a road performance index, including unconfined compressive strength, splitting tensile strength, resilient modulus, drying shrinkage coefficient, thermal shrinkage coefficient, and erosion loss rate. Each task output head outputs two values: the predicted mean and logarithmic variance of the corresponding performance index. The logarithmic variance is converted into the performance fluctuation mean through exponential operation. Finally, the performance prediction mean and performance fluctuation mean of all road performance indices are output, realizing the simultaneous prediction and uncertainty quantification of multiple road performance characteristics of the solid waste-based cementitious aggregate base course.

[0036] Step S140: Based on the average performance prediction and the average performance fluctuation, perform a reliability analysis of the stabilized aggregate base course of solid waste-based cementitious materials, generate and output standardized performance test results, the specific content of which is as follows: In step S140, a reliability analysis of the solid waste-based cementitious material stabilized aggregate base course is performed based on the predicted average performance value and the average performance fluctuation value, generating and outputting standardized performance test results. In this embodiment, the performance reliability analysis focuses on quantifying the performance fluctuation range and, in conjunction with the current technical specifications for highway base course construction, assesses the safety margin of various road performance characteristics of the solid waste-based cementitious material stabilized aggregate base course. This avoids misjudgments of test results due to material dispersion and provides a scientific basis for engineering quality control.

[0037] First, the reliability index of various road performance characteristics of the solid waste-based cementitious aggregate base course is calculated based on the predicted average performance and the average performance fluctuation. The reliability index characterizes the probability that the corresponding performance indicator meets the specification requirements. A higher value indicates that the performance indicator not only meets the average requirements but is also less affected by factors such as material batch fluctuations and test errors, demonstrating higher performance stability. During the calculation, the minimum specification requirement value for each road performance characteristic is used as a benchmark threshold. Combined with the predicted average performance and the average performance fluctuation, the risk of actual performance falling below the specification threshold is quantitatively assessed. Differentiated benchmark reliability requirements are set for performance indicators of different importance. The benchmark reliability requirements for core mechanical indicators such as unconfined compressive strength and splitting tensile strength are higher than those for deformation indicators such as drying shrinkage coefficient and thermal shrinkage coefficient.

[0038] The reliability index is compared with a preset reliability threshold to obtain the reliability judgment result of the stabilized aggregate base course performance of the solid waste-based cementitious material. The preset reliability threshold is set in stages according to the engineering grade, base course layer, and service environment. Specifically, the reliability threshold for the base course of expressways and Class I highways is higher than that for Class II and lower highways, the reliability threshold for the upper base course is higher than that for the lower base course, and the reliability threshold for cold and rainy regions is higher than that for temperate regions. When the reliability index of a certain road performance is greater than or equal to the corresponding preset reliability threshold, the performance is judged to be qualified; when the reliability index is less than the preset reliability threshold, the performance is judged to be unqualified, and the performance index is marked as a risk item, indicating that further verification or adjustment of the construction process is required.

[0039] After obtaining the reliability assessment results of the solid waste-based cementitious aggregate base course, the comprehensive road performance of the solid waste-based cementitious aggregate base course is evaluated using a weighted comprehensive scoring algorithm based on the reliability index and preset performance grading thresholds. The weighted comprehensive scoring algorithm assigns weights according to the contribution of each road performance characteristic to the overall service performance of the base course, with mechanical properties having the highest weight, followed by durability, and then deformation performance. The reliability index of each performance characteristic is multiplied by its corresponding weight and then summed to obtain the comprehensive performance score. Based on the comprehensive score, the base course performance is divided into four levels: excellent, good, qualified, and unqualified. Different levels correspond to different engineering application recommendations: excellent levels can be directly used for high-grade highway base courses, qualified levels can be used for low-grade highways or temporary projects, and unqualified levels are prohibited from use in engineering construction.

[0040] The standardized performance test results are generated based on the performance reliability assessment results and the comprehensive road performance level evaluation results. These standardized performance test results are compiled using a unified format and include basic test information, the predicted average performance value for each road performance characteristic, the performance fluctuation range, the reliability index, the individual performance assessment results, the comprehensive performance level, and engineering application recommendations. The basic test information includes the sample number, material composition, manufacturing process, maintenance age, service conditions, and test time. The performance fluctuation range is determined by a fluctuation of twice the predicted performance value, covering more than 95% of the performance distribution range.

[0041] The standardized performance test results are output. In this embodiment, the standardized performance test results support multiple output formats, including electronic test reports, paper test reports, and database storage records. Electronic test reports can be exported as PDFs, including comparison charts and reliability distribution curves for various performance aspects, facilitating intuitive viewing of the test results. Paper test reports are printed according to the format required by the engineering quality acceptance specifications and, after being stamped with the official seal of the testing institution, can serve as the basis for engineering acceptance. All test results are synchronously stored in a cloud database, establishing an association index between the test results and the original sample data, supporting retrospective querying and statistical analysis of historical test data. Simultaneously, the standardized performance test results can be interfaced with the highway engineering quality supervision and management system through a data interface to achieve real-time uploading and sharing of test data, improving the informatization level of engineering quality control.

[0042] Based on steps S110 to S140 above, in this embodiment of the invention, by collecting multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course and constructing a unified dataset, the standardized integration and unified management of multi-source heterogeneous test data are realized; a preset deep learning network is used to extract multi-dimensional performance characterization features, fully exploring the intrinsic correlation between aggregate structural characteristics, mechanical response laws and base course road performance; a gated expert network is used to simultaneously output the predicted mean and fluctuation mean of multiple road performances, quantifying the material discreteness and experimental uncertainty in the testing process; performance reliability analysis is performed based on the performance predicted mean and fluctuation mean, significantly improving the accuracy and reliability of the test results, greatly shortening the testing cycle, reducing testing costs, and thus solving the technical problems of difficulty in multi-source heterogeneous data fusion, insufficient extraction of deep performance features, inability to quantify performance prediction uncertainty, and low reliability of test results in the performance testing of solid waste-based cementitious material stabilized aggregate base course in the prior art.

[0043] The solid waste-based cementitious material stabilized aggregate base performance testing method of the present invention performs automated preprocessing on the original multimodal test data before constructing the multimodal test dataset. The automated preprocessing includes outlier truncation, missing value completion, dimension normalization, and time series data denoising, which can effectively improve data quality, eliminate dimension differences and noise interference, and provide an accurate and reliable data foundation for subsequent feature extraction and performance prediction.

[0044] Furthermore, outlier truncation employs a robust Z-score and interquartile range method to identify outliers by traversing the multimodal raw detection data and truncates them to a preset reasonable boundary. Missing value completion uses the median interpolation method of neighboring samples to complete the missing data caused by signal interruption. This method can accurately remove abnormal interference while preserving the integrity of valid data, significantly improving the accuracy and continuity of multimodal raw detection data.

[0045] Specifically, feature extraction is performed on the multimodal detection dataset based on a pre-defined deep learning network to obtain multi-dimensional performance characterization features. These features include: graph structure modeling of aggregate gradation curve data based on a graph attention network to obtain aggregate skeleton stability features; processing of mechanical performance time-series loading data based on a one-dimensional convolutional neural network to obtain mechanical response evolution features; pixel-level segmentation of base section image data based on an instance segmentation model to obtain internal void distribution features; and concatenation and fusion of aggregate skeleton stability features, mechanical response evolution features, and internal void distribution features to obtain multi-dimensional performance characterization features. These features can comprehensively and accurately characterize the base skeleton structure, mechanical response, and internal void distribution, providing highly recognizable multi-dimensional performance characterization features for performance prediction.

[0046] Furthermore, the instance segmentation model is an improved YOLOv8seg model, which replaces the spatial pyramid pooling fast module in the YOLOv8seg model with a separable kernel attention pyramid module. The internal void distribution features extracted based on the improved YOLOv8seg model include void ratio, void area, void perimeter, maximum void width, and void roundness, which can significantly improve the accuracy of fine-grained void feature extraction and simultaneously obtain comprehensive quantitative internal void distribution features such as void ratio, void area, void perimeter, maximum void width, and void roundness.

[0047] Specifically, multi-dimensional performance characterization features are input into a trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics of the solid waste-based cementitious aggregate base course. This includes: concatenating the multi-dimensional performance characterization features with scene embedding features to obtain comprehensive input features, where the scene embedding features are vectorized from target layer, traffic level, and environmental zoning parameters; calculating the fit weights of multiple gating experts based on the comprehensive input features; and using the fit weights to weightedly fuse the exclusive characterizations output by multiple gating experts to obtain the performance prediction mean and performance fluctuation mean of multiple road performance characteristics of the solid waste-based cementitious aggregate base course. This adaptively matches different engineering scenarios, accurately outputting the predicted mean and fluctuation mean of multiple road performance characteristics of the solid waste-based cementitious aggregate base course, thus improving the scenario adaptability and reliability of performance prediction.

[0048] In this embodiment, the reliability analysis of the solid waste-based cementitious aggregate base course is conducted based on the performance prediction mean and the performance fluctuation mean. Specifically, this includes: calculating the reliability index of various road performance characteristics of the solid waste-based cementitious aggregate base course based on the performance prediction mean and the performance fluctuation mean; comparing the reliability index with a preset reliability threshold to obtain the reliability judgment result of the solid waste-based cementitious aggregate base course performance; and evaluating the comprehensive road performance of the solid waste-based cementitious aggregate base course based on the reliability index and the preset performance grading threshold using a weighted comprehensive scoring algorithm. This enables quantitative evaluation and scientific grading of base course performance, significantly improving the objectivity and engineering guidance of the test results.

[0049] Furthermore, after generating and outputting standardized performance test results, the standardized performance test results are compared with the field measured performance data. Based on the comparison results, the parameters of the preset deep learning network and gated expert network are incrementally updated through the maximum likelihood loss function, which can continuously optimize the model prediction accuracy and improve the generalization ability and long-term reliability of the method under different materials and working conditions.

[0050] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0051] This invention also provides a device for testing the performance of solid waste-based cementitious aggregate base course. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0052] Figure 2 According to one embodiment of the present invention, a device for testing the performance of solid waste-based cementitious aggregate base course includes: The acquisition module 201 is used to acquire multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course, and to construct a multimodal test dataset based on the multimodal raw test data; Extraction module 202 is used to extract features from the multimodal detection dataset based on a preset deep learning network to obtain multidimensional performance characterization features. The multidimensional performance characterization features are then input into a trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances of the corresponding solid waste-based cementitious material stabilized aggregate base course. Analysis module 203 is used to perform reliability analysis of the performance of solid waste-based cementitious material stabilized aggregate base course based on the performance prediction mean and the performance fluctuation mean, and to generate and output standardized performance test results.

[0053] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0054] According to one embodiment of the present invention, an electronic device is also provided, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the above-described method for testing the performance of solid waste-based cementitious aggregate base course during operation.

[0055] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program: Step S1: Collect multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course, and construct a multimodal test dataset based on the multimodal raw test data; Step S2: Based on the preset deep learning network, feature extraction is performed on the multimodal detection dataset to obtain multidimensional performance characterization features. The multidimensional performance characterization features are then input into the trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances of the corresponding solid waste-based cementitious aggregate base course. Step S3: Based on the performance prediction mean and performance fluctuation mean, perform a reliability analysis of the performance of the solid waste-based cementitious material stabilized aggregate base course, and generate and output standardized performance test results.

[0056] According to one embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the storage medium is located to execute the above-described method for testing the performance of solid waste-based cementitious aggregate base course.

[0057] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps: Step S1: Collect multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course, and construct a multimodal test dataset based on the multimodal raw test data; Step S2: Based on the preset deep learning network, feature extraction is performed on the multimodal detection dataset to obtain multidimensional performance characterization features. The multidimensional performance characterization features are then input into the trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances of the corresponding solid waste-based cementitious aggregate base course. Step S3: Based on the performance prediction mean and performance fluctuation mean, perform a reliability analysis of the performance of the solid waste-based cementitious material stabilized aggregate base course, and generate and output standardized performance test results.

[0058] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0059] According to one embodiment of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the above-described method for testing the performance of solid waste-based cementitious aggregate base course.

[0060] Optionally, in this embodiment, the above-mentioned computer program product can be configured as a computer program that performs the following steps: Step S1: Collect multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course, and construct a multimodal test dataset based on the multimodal raw test data; Step S2: Based on the preset deep learning network, feature extraction is performed on the multimodal detection dataset to obtain multidimensional performance characterization features. The multidimensional performance characterization features are then input into the trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances of the corresponding solid waste-based cementitious aggregate base course. Step S3: Based on the performance prediction mean and performance fluctuation mean, perform a reliability analysis of the performance of the solid waste-based cementitious material stabilized aggregate base course, and generate and output standardized performance test results.

[0061] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0062] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0063] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0064] The units described as separate components may or may not be physically separate. Similarly, the components shown as units may or may not be physical units; they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0065] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0067] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for testing the performance of solid waste-based cementitious aggregate base course, characterized in that, include: Multimodal raw detection data of solid waste-based cementitious material stabilized aggregate base course were collected, and a multimodal detection dataset was constructed based on the multimodal raw detection data; Feature extraction is performed on the multimodal detection dataset based on a preset deep learning network to obtain multidimensional performance characterization features. The multidimensional performance characterization features are then input into a trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances corresponding to the solid waste-based cementitious material stabilized aggregate base course. Based on the predicted average performance and the average performance fluctuation, a reliability analysis of the stabilized aggregate base course of the solid waste-based cementitious material is performed, and standardized performance test results are generated and output.

2. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 1, characterized in that, Before constructing the multimodal detection dataset, automated preprocessing is performed on the original multimodal detection data. The automated preprocessing includes outlier truncation, missing value completion, unit normalization, and time series data denoising.

3. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 2, characterized in that, The outlier truncation method uses a combination of robust Z-score and interquartile range to identify outliers by traversing the original multimodal detection data and truncates the outliers to a preset reasonable boundary. The missing value completion method uses the median interpolation of neighboring samples to complete the missing data caused by signal interruption.

4. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 1, characterized in that, Based on the preset deep learning network, feature extraction is performed on the multimodal detection dataset to obtain the multidimensional performance characterization features, specifically including: Graph structure modeling of aggregate gradation curve data is performed based on graph attention network to obtain the stability characteristics of aggregate skeleton; The mechanical response evolution characteristics are obtained by processing the time-series loading data of mechanical properties using a one-dimensional convolutional neural network. The base section image data is segmented at the pixel level based on the instance segmentation model to obtain the internal void distribution characteristics. The stability characteristics of the aggregate skeleton, the evolution characteristics of the mechanical response, and the distribution characteristics of the internal voids are spliced ​​and fused to obtain the multi-dimensional performance characterization characteristics.

5. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 4, characterized in that, The instance segmentation model is an improved YOLOv8seg model, which replaces the spatial pyramid pooling fast module in the YOLOv8seg model with a separable kernel attention pyramid module; the internal void distribution features extracted based on the improved YOLOv8seg model include void ratio, void area, void perimeter, maximum void width, and void roundness.

6. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 1, characterized in that, The multi-dimensional performance characterization features are input into the trained gated expert network to obtain the predicted mean value of performance and the mean value of performance fluctuation for multiple road performance characteristics of the solid waste-based cementitious aggregate base course. Specifically, this includes: The multi-dimensional performance characterization features are concatenated with the scene embedding features to obtain comprehensive input features. The scene embedding features are obtained by vectorizing the target layer, traffic level, and environmental zoning parameters. Based on the comprehensive input features, calculate the fit weights corresponding to multiple gating experts; By using the fitness weights to perform weighted fusion on the exclusive representations output by the multiple gating experts, the predicted mean of performance and the mean of performance fluctuation of the corresponding road performance of the solid waste-based cementitious aggregate base course are obtained.

7. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 1, characterized in that, Based on the predicted average performance and the predicted average performance fluctuation, a reliability analysis of the stabilized aggregate base course of the solid waste-based cementitious material is conducted, specifically including: The reliability index of various road performance characteristics of the solid waste-based cementitious material stabilized aggregate base course is calculated based on the predicted average performance value and the performance fluctuation average performance value. The reliability index is compared with a preset reliability threshold to obtain the reliability judgment result of the solid waste-based cementitious material stabilized aggregate base course. Based on the reliability index and the preset performance grading threshold, the comprehensive road performance of the solid waste-based cementitious material stabilized aggregate base course is evaluated using a weighted comprehensive scoring algorithm.

8. The method for testing the performance of solid waste-based cementitious material stabilized aggregate base course according to claim 1, characterized in that, After generating and outputting the standardized performance test results, the standardized performance test results are compared with the actual measured performance data on site. Based on the comparison results, the parameters of the preset deep learning network and the gated expert network are incrementally updated using the maximum likelihood loss function.

9. A device for testing the performance of solid waste-based cementitious aggregate base course, characterized in that, include: The acquisition module is used to acquire multimodal raw test data of solid waste-based cementitious material stabilized aggregate base course, and to construct a multimodal test dataset based on the multimodal raw test data; The extraction module is used to extract features from the multimodal detection dataset based on a preset deep learning network to obtain multidimensional performance characterization features. The multidimensional performance characterization features are then input into a trained gating expert network to obtain the performance prediction mean and performance fluctuation mean of multiple road performances corresponding to the solid waste-based cementitious material stabilized aggregate base course. The analysis module is used to perform a reliability analysis of the performance of the solid waste-based cementitious material stabilized aggregate base course based on the predicted average performance value and the average performance fluctuation value, and to generate and output standardized performance test results.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device where the storage medium is located to perform the solid waste-based cementitious material stabilized aggregate base performance testing method according to any one of claims 1 to 8.

Images