A process monitoring method and system

Abstract

This invention provides a process monitoring method and system. The invention relates to the field of process monitoring technology. The method includes: acquiring infrared images of a large concrete slab; using ground-penetrating radar (GPR) to detect multiple abnormal areas of the concrete slab to obtain GPR images of each abnormal area; determining GPR information of multiple abnormal points in each abnormal area based on the GPR images of each abnormal area; clustering the GPR information of the multiple abnormal points in each abnormal area to obtain K clusters; determining multiple supplementary detection points in each abnormal area based on the GPR images and the K clusters; acquiring ultrasonic radar data of the multiple supplementary detection points; and determining whether the strength of the large concrete slab is qualified based on the GPR information of the multiple abnormal points in each abnormal area and the ultrasonic radar data of the multiple supplementary detection points. This method can efficiently and accurately assess whether the strength of a large concrete slab is qualified.

Description

Technical Field

[0001] This invention relates to the field of process monitoring technology, specifically to a process monitoring method and system. Background Technology

[0002] During airport construction, large concrete slabs serve as the core load-bearing components of critical infrastructure such as runways and aprons. Their internal structural strength and overall quality directly impact the safety and durability of the entire project, necessitating quality monitoring during construction and maintenance. Large concrete slabs are prone to hidden defects such as voids, cracks, and segregation due to various complex factors, including uneven material proportions, flawed pouring processes, improper curing conditions, and external environmental erosion. These defects not only significantly reduce the mechanical properties of the components but also pose a significant challenge to traditional visual inspection methods due to their deep embedding within the structure. Currently, while conventional methods such as rebound hammer testing and ultrasonic testing are widely used for inspecting the internal quality of concrete structures, they often suffer from limited depth of inspection, insufficient accuracy, or only reflecting surface information. This makes it difficult to comprehensively and accurately assess the distribution of spatial defects within large volumes of concrete and their true impact on overall strength. Existing technologies typically rely on single inspection techniques or limited point sampling inspections. This approach is not only inefficient but also prone to missing critical defects, leading to inaccurate strength assessments. Meanwhile, the interpretation of test data relies heavily on the personal experience of technicians, is easily influenced by subjective factors, and lacks systematic data fusion and intelligent analysis, making it difficult to form a unified and objective quantitative evaluation of the overall health status of concrete slabs.

[0003] Therefore, how to efficiently and accurately assess whether the strength of large concrete slabs is up to standard is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem this invention addresses is how to efficiently and accurately assess whether the strength of a large concrete slab is up to standard.

[0005] According to a first aspect, the present invention provides a process monitoring method, comprising: acquiring an infrared image of a large concrete slab; determining multiple abnormal regions of the concrete slab using an abnormal region determination model based on the infrared image of the large concrete slab; obtaining a ground-penetrating radar image of each abnormal region using ground-penetrating radar based on the multiple abnormal regions of the concrete slab; determining ground-penetrating radar information of multiple abnormal points in each abnormal region based on the ground-penetrating radar image of each abnormal region; clustering K clusters based on the ground-penetrating radar information of the multiple abnormal points in each abnormal region; determining multiple supplementary detection points in each abnormal region based on the ground-penetrating radar image of each abnormal region and the K clusters; acquiring ultrasonic radar data of the multiple supplementary detection points; and determining whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of the multiple abnormal points in each abnormal region and the ultrasonic radar data of the multiple supplementary detection points.

[0006] In one possible implementation, determining whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of multiple anomaly points in each anomaly region and the ultrasonic radar data of multiple supplementary detection points includes: determining N core test points based on the ground-penetrating radar information of multiple anomaly points in each anomaly region and the ultrasonic radar data of multiple supplementary detection points; acquiring experimental data from the N core test points; constructing a verification graph, which includes N core test point nodes and N edges between the core test point nodes, wherein the node characteristics of each core test point node are the experimental data of the core test point, the ground-penetrating radar information of multiple anomaly points in each anomaly region, and the ultrasonic radar data of the multiple supplementary detection points, and the edges between the core test point nodes represent the distance between the core test points; processing the verification graph based on a graph neural network to determine multiple core verification points; and determining whether the strength of the large concrete slab is qualified based on the experimental data of the N core test points and the experimental data of the multiple core verification points.

[0007] In one possible implementation, determining N core test points based on ground-penetrating radar information of multiple anomaly points in each anomaly region and ultrasonic radar data of multiple supplementary detection points includes: using a generative adversarial network to generate a risk distribution map of the anomaly region based on ground-penetrating radar information of multiple anomaly points in each anomaly region and ultrasonic radar data of multiple supplementary detection points; and determining N core test points based on the risk distribution map of the anomaly region.

[0008] In one possible implementation, the abnormal region determination model is a convolutional neural network model.

[0009] According to a second aspect, the present invention provides a process monitoring system, comprising: an acquisition module for acquiring infrared images of a large concrete slab; an abnormal region determination module for determining multiple abnormal regions of the concrete slab based on the infrared images of the large concrete slab using an abnormal region determination model; a detection module for obtaining ground-penetrating radar images of each abnormal region using ground-penetrating radar detection based on the multiple abnormal regions of the concrete slab; an abnormal point information determination module for determining ground-penetrating radar information of multiple abnormal points in each abnormal region based on the ground-penetrating radar images of each abnormal region; a clustering module for clustering K clusters based on the ground-penetrating radar information of multiple abnormal points in each abnormal region; a supplementary detection point determination module for determining multiple supplementary detection points in each abnormal region based on the ground-penetrating radar images of each abnormal region and the K clusters; an ultrasonic data acquisition module for acquiring ultrasonic radar data of multiple supplementary detection points; and a strength determination module for determining whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of multiple abnormal points in each abnormal region and the ultrasonic radar data of the multiple supplementary detection points.

[0010] In one possible implementation, the strength determination module is further configured to: determine N core test points based on ground-penetrating radar information of multiple anomaly points in each anomaly region and ultrasonic radar data of multiple supplementary detection points; acquire experimental data of the N core test points; construct a verification graph, which includes N core test point nodes and edges between the N core test point nodes, wherein the node features of each core test point node are the experimental data of the core test point, the ground-penetrating radar information of multiple anomaly points in each anomaly region, and the ultrasonic radar data of the multiple supplementary detection points, and the edges between the core test point nodes represent the distance between the core test points; process the verification graph based on a graph neural network to determine multiple core verification points; and determine whether the strength of the large concrete slab is qualified based on the experimental data of the N core test points and the experimental data of the multiple core verification points.

[0011] In one possible implementation, the intensity determination module is further configured to: generate a risk distribution map of the abnormal region using a generative adversarial network based on ground-penetrating radar information of multiple abnormal points in each abnormal region and ultrasonic radar data of multiple supplementary detection points; and determine N core test points based on the risk distribution map of the abnormal region.

[0012] In one possible implementation, the abnormal region determination model is a convolutional neural network model.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring an infrared image of a large concrete slab; determining multiple abnormal regions of the concrete slab using an abnormal region determination model based on the infrared image of the large concrete slab; obtaining a ground-penetrating radar image of each abnormal region based on the multiple abnormal regions of the concrete slab using ground-penetrating radar; determining multiple abnormal point ground-penetrating radar information of each abnormal region based on the ground-penetrating radar image of each abnormal region; clustering the multiple abnormal point ground-penetrating radar information of each abnormal region to obtain K clusters; determining multiple supplementary detection points of each abnormal region based on the ground-penetrating radar image of each abnormal region and the K clusters; acquiring ultrasonic radar data of the multiple supplementary detection points; and determining whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of the multiple abnormal points of each abnormal region and the ultrasonic radar data of the multiple supplementary detection points.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned process monitoring method. The method includes: acquiring an infrared image of a large concrete slab; determining multiple abnormal regions of the concrete slab using an abnormal region determination model based on the infrared image of the large concrete slab; obtaining a ground-penetrating radar image of each abnormal region using ground-penetrating radar detection based on the multiple abnormal regions of the concrete slab; determining multiple abnormal point ground-penetrating radar information of each abnormal region based on the ground-penetrating radar image of each abnormal region; clustering the multiple abnormal point ground-penetrating radar information of each abnormal region to obtain K clusters; determining multiple supplementary detection points of each abnormal region based on the ground-penetrating radar image of each abnormal region and the K clusters; acquiring ultrasonic radar data of the multiple supplementary detection points; and determining whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of the multiple abnormal points of each abnormal region and the ultrasonic radar data of the multiple supplementary detection points.

[0015] This invention provides a process monitoring method and system. The method includes: acquiring infrared images of a large concrete slab; identifying multiple abnormal regions of the concrete slab using an abnormal region determination model based on the infrared images; obtaining ground-penetrating radar (GPR) images of each abnormal region using GPR detection based on the multiple abnormal regions; determining GPR information of multiple abnormal points in each abnormal region based on the GPR images of each abnormal region; clustering the GPR information of the multiple abnormal points in each abnormal region to obtain K clusters; determining multiple supplementary detection points in each abnormal region based on the GPR images of each abnormal region and the K clusters; acquiring ultrasonic radar data of the multiple supplementary detection points; and determining whether the strength of the large concrete slab is qualified based on the GPR information of the multiple abnormal points in each abnormal region and the ultrasonic radar data of the multiple supplementary detection points. This method can efficiently and accurately assess whether the strength of a large concrete slab is qualified. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a process monitoring method provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a large concrete slab provided in an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of an infrared thermal imager provided in an embodiment of the present invention;

[0019] Figure 4 A schematic diagram of a process for determining whether the strength of a large concrete slab is qualified, provided for an embodiment of the present invention;

[0020] Figure 5 This is a schematic diagram of a process for determining N core test points according to an embodiment of the present invention;

[0021] Figure 6 This is a schematic diagram of a process monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0023] In this embodiment of the invention, the following are provided: Figure 1 The process monitoring method shown includes steps S1 to S8:

[0024] Step S1: Obtain an infrared image of a large concrete slab.

[0025] Large concrete slabs refer to large slab-type concrete components used in airport runways or aprons, made of high-strength concrete, and capable of bearing the loads of aircraft take-off and landing and resisting environmental erosion. Figure 2 This is a schematic diagram of a large concrete slab provided in an embodiment of the present invention.

[0026] Infrared images of large concrete slabs are obtained by taking pictures of large concrete slabs with an infrared thermal imager. Figure 3 This is a schematic diagram of an infrared thermal imager provided in an embodiment of the present invention.

[0027] Infrared images of large concrete slabs can clearly show the temperature field distribution of different areas on the surface and inside the concrete slab. Different colors represent different temperature areas. Infrared images of large concrete slabs can identify potential defects inside the concrete, such as voids, cracks, or uneven material distribution.

[0028] The integrity of the internal structure of a concrete slab directly affects its thermal conductivity. When there are voids inside the concrete slab, the thermal conductivity of the void area is significantly lower than that of the surrounding solid concrete area due to the strong insulating properties of air, and this will appear as a distinct low-temperature anomaly in infrared images. When there are cracks in the concrete slab, the cracks may introduce air or moisture and alter the local heat conduction path, leading to a temperature gradient between the cracked area and the normal area.

[0029] Step S2: Based on the infrared image of the large concrete slab, use the anomaly region determination model to determine multiple anomaly regions of the concrete slab.

[0030] The abnormal region determination model is a convolutional neural network model. The input of the abnormal region determination model is the infrared image of the large concrete slab, and the output of the abnormal region determination model is multiple abnormal regions of the concrete slab.

[0031] Convolutional neural network models include Convolutional Neural Networks (CNNs). A CNN is a deep learning model inspired by the biological visual system. Through a combination of convolutional layers, pooling layers, and fully connected layers, CNNs can automatically learn the spatial hierarchical features of images.

[0032] Multiple abnormal areas in the concrete slab were identified by the model in infrared images of large concrete slabs, where the temperature characteristics differed significantly from those of normal areas.

[0033] Anomalies can occur in large concrete slabs due to factors such as internal voids, moisture accumulation, or uneven material density, which can alter their thermal conductivity. These anomalies can appear as spots or streaks of localized overheating or undercooling on infrared images.

[0034] Infrared images of large concrete slabs can comprehensively record the overall temperature distribution of the slab. Concrete areas with different structural states can exhibit varying temperature characteristics due to differences in thermal conductivity. For example, when there are voids within the concrete, heat transfer is hindered in that area, resulting in a temperature difference compared to the surrounding solid concrete. Similarly, when cracks exist, the temperature at the cracks will also be abnormal due to factors such as air circulation or moisture penetration. This allows the model to filter out areas with structural anomalies from the infrared images of large concrete slabs based on these temperature differences.

[0035] Convolutional neural networks (CNNs) can perform intensive sliding scans and feature extraction on infrared images of large concrete slabs using multi-level convolutional kernels. The model's shallow convolutional layers effectively capture local temperature gradient changes and edge contour information in the infrared image. As the network depth increases, deeper convolutional layers can non-linearly combine and abstract these low-level features to form high-level semantic feature maps that represent complex temperature distribution patterns. For example, the model can learn to identify regions with a specific circular shape and a significantly higher central temperature than the surrounding area, associating them with internal voids in the large concrete slab. It can also identify irregular strip-shaped texture features with lower temperatures and associate them with cracks and water seepage. Finally, by parsing and spatially mapping these high-level feature maps, the CNN can accurately separate areas that do not conform to normal concrete thermal characteristics from the background, thereby identifying multiple abnormal areas in the concrete slab.

[0036] Step S3: Based on the multiple abnormal areas of the concrete slab, ground-penetrating radar is used to detect and obtain ground-penetrating radar images of each abnormal area.

[0037] The ground-penetrating radar image of each anomalous area is generated by scanning multiple anomalous areas of the concrete slab using ground-penetrating radar equipment. It is a three-dimensional image that can reflect the internal structural condition of the corresponding anomalous area.

[0038] In ground-penetrating radar (GPR) images, stripes and patches of different gray levels or colors represent different structural features inside the concrete. For example, solid concrete areas, void areas, and crack areas will present different image features, and each GPR image corresponds one-to-one with the location of the corresponding abnormal area on the concrete slab.

[0039] Step S4: Based on the ground-penetrating radar image of each abnormal region, determine the ground-penetrating radar information of multiple abnormal points in each abnormal region.

[0040] In some embodiments, an anomaly information determination model can be used to determine ground-penetrating radar (GPR) information of multiple anomalies in each anomaly region. The anomaly information determination model is a convolutional neural network model. The input to the anomaly information determination model is the GPR image of each anomaly region, and the output of the anomaly information determination model is the GPR information of multiple anomalies in each anomaly region.

[0041] Ground-penetrating radar (GPR) information for multiple anomaly points within each anomaly region is detailed radar detection information extracted after analyzing the GPR image of each anomaly region, focusing on multiple points with anomalous structural features within that region. This information includes the specific coordinates of each anomaly point within its corresponding anomaly region, the radar wave reflection intensity at that point, and the propagation time of the reflected wave.

[0042] Convolutional neural networks (CNNs) possess the ability to accurately capture local features in images. Anomalies in ground-penetrating radar (GPR) images can exhibit specific abrupt changes in grayscale or morphological characteristics in reflective stripes, features that clearly differ from normal areas. CNNs, through the kernels of convolutional layers, can extract local edge textures and morphological features from the image. Small kernels can identify the edge contours of reflected waves, while large kernels can capture the overall shape of anomaly regions. Pooling layers can retain key features and reduce data volume through downsampling. Fully connected layers in CNNs can integrate global features. By combining typical feature patterns of anomalies in the training data, fully connected layers can accurately locate pixels in the image that match the anomaly characteristics and output GPR information such as coordinates, reflection intensity, and depth of these anomalies.

[0043] Step S5: Based on the ground-penetrating radar information of multiple abnormal points in each abnormal region, K clusters are obtained.

[0044] K-means clustering is an iterative clustering analysis algorithm. It divides sample data into K distinct clusters based on a pre-defined number of clusters, K, through an iterative process, ensuring that each sample belongs to the cluster whose mean (cluster center) is closest to it. K-means clustering achieves its clustering goal by minimizing the sum of squared distances from samples within each cluster to the cluster center, effectively grouping similar data objects together. In some embodiments, the value of K can be pre-set manually.

[0045] The K clusters are sets of K anomalies with similar characteristics obtained by clustering ground-penetrating radar (GPR) information from multiple anomalies in each anomaly region using the K-means clustering algorithm. The GPR information from anomalies within each cluster exhibits high similarity in key features, such as similar depth ranges, radar wave reflection intensity ranges, and anomaly structure types.

[0046] The ground-penetrating radar information characteristics of anomalies differ significantly between different clusters. For example, the depth of anomalies in one cluster is concentrated between 10-20 cm, while the depth of anomalies in another cluster is concentrated between 30-40 cm. Each cluster has a corresponding cluster center, and the characteristic value of the cluster center is the average of the characteristic values ​​of all anomalies in that cluster. At the same time, each cluster records the number of anomalies it contains and the distribution range of these anomalies on the concrete slab.

[0047] As an example, specifically: K outlier ground-penetrating radar (GPR) data points are randomly selected as initial cluster centers. For each outlier GPR data point, the distance between that outlier GPR data point and all initial cluster centers is calculated, and each outlier GPR data point is assigned to the nearest cluster center to form K clusters. For each cluster, the average of the feature vectors of all outlier GPR data points within the cluster is calculated, and this average is used as the new cluster center. The above steps are repeated until the number of cluster centers reaches a predetermined number of iterations, at which point the clustering process ends.

[0048] By processing ground-penetrating radar information from multiple anomalies in each anomalous region using K-means clustering, anomalies with similar characteristics can be grouped into the same cluster, thus transforming previously scattered anomaly information into a group with clear common features. This clustering method can intuitively present the distribution patterns of anomalies in key features such as depth and reflectivity, and can help quickly identify sets of regions with similar structural anomalies.

[0049] Step S6: Based on the ground-penetrating radar image of each anomalous area and the K clusters, determine multiple supplementary detection points for each anomalous area.

[0050] In some implementations, a supplementary point determination model can be used to determine multiple supplementary detection points for each anomalous region. The supplementary point determination model is a deep neural network model. The input to the supplementary point determination model is the ground-penetrating radar image of each anomalous region and the K clusters; the output of the supplementary point determination model is multiple supplementary detection points for each anomalous region.

[0051] Deep neural network models include deep neural networks (DNNs). A deep neural network is a neural network composed of multiple processing layers, each containing multiple neurons. Deep neural networks can perform classification, regression, or generation tasks by learning complex patterns in data. Through multiple layers of nonlinear transformations, deep neural networks can learn hierarchical feature representations from low to high levels from raw input and can handle highly complex nonlinear relationships.

[0052] Multiple supplementary detection points for each anomalous region are determined by supplementary points to identify specific locations within each anomalous region that require further precise detection in the model's output.

[0053] Each anomalous region's ground-penetrating radar (GPR) image may contain weak anomalous signal areas not covered by existing anomalies. These areas may not have been identified as anomalies due to low signal strength or indistinct features, but they still hold the potential for structural anomalies. The K clusters reflect the distribution patterns and characteristic types of existing anomalies. By analyzing the distribution range of the clusters, the model can identify gaps between clusters, which may contain missed anomalies. The model can analyze the characteristics of cluster boundaries on the GPR image and whether there are suspicious signals in the areas between clusters, thereby selecting the most valuable locations for supplementary detection points.

[0054] Deep neural networks can analyze radar image features across K clusters and identify core regions suspected of defects. The model can also focus on neighboring regions outside each cluster boundary, searching for areas in the radar image with signal characteristics similar to those within the cluster but not included, as well as regions where the signal shows a gradual transition from normal to abnormal. By learning the boundary features of a large number of defect samples, the model can determine which boundary locations might have a larger defect range than the current cluster indicates. Furthermore, the model can assess blank areas between different clusters and analyze the radar images of these areas for weak anomalous signals, identifying potential pathways connecting different defects. Based on a comprehensive assessment of defect boundary uncertainty and inter-regional correlations, the model can ultimately determine the location of supplementary detection points.

[0055] In some embodiments, determining multiple supplementary detection points for each anomalous region based on the ground-penetrating radar image of each anomalous region and the K clusters includes steps S21 to S23:

[0056] Step S21: Based on the ground-penetrating radar image of each abnormal region, determine the image grayscale distribution data of each abnormal region, the local area where the texture suddenly becomes denser, the gradient band where the grayscale changes from dark to light, the small area with obvious brightness difference from the surrounding area, and the local range where the ripples break.

[0057] In some embodiments, a convolutional neural network can be used to determine the image grayscale distribution data of each abnormal region, local regions where the texture suddenly becomes denser, gradient bands where the grayscale changes from dark to light, small areas with significant differences in brightness from the surrounding area, and local ranges where the ripples break.

[0058] The image grayscale distribution data for each anomalous region is a dataset of the brightness and darkness at each location in the ground-penetrating radar image, output by a convolutional neural network.

[0059] Grayscale distribution data reflects the strength of electromagnetic signals in different regions through numerical values, and can intuitively present the differences in brightness between abnormal areas and their surroundings, as well as within abnormal areas.

[0060] A localized area where the texture suddenly becomes denser is a small area in a ground-penetrating radar image where the stripe or wavy texture suddenly changes from a sparse state to a dense state.

[0061] The gradient band where the gray level changes from dark to light is a band-shaped area in the ground-penetrating radar image of each anomaly region where the brightness gradually transitions from a dark state to a bright state.

[0062] Small areas with significantly different brightness from the surrounding area are those small areas in the ground-penetrating radar image of each anomalous region that are significantly brighter or lower than the surrounding area.

[0063] The localized area where the ripples break is a small area in a ground-penetrating radar image where the originally continuous reflected ripples are interrupted or blank.

[0064] Convolutional neural networks (CNNs) can capture pixel-level grayscale variation patterns through shallow convolutional layers, thereby statistically analyzing image grayscale distribution data. Mid-level convolutional kernels can identify abrupt changes in local texture patterns and accurately locate regions where texture suddenly becomes denser. By learning from the gradient changes of adjacent pixels, CNNs can detect the boundaries of gradient bands from dark to light grayscale and quickly pinpoint small areas with significantly different brightness from their surroundings, while also identifying localized breaks in the ripple pattern.

[0065] Step S22: Based on the image grayscale distribution data of each abnormal region, the local area where the texture suddenly becomes denser, the gradient band where the grayscale changes from dark to light, the small area with obvious brightness difference from the surrounding area, the local range where the ripples break, the K clusters determine the suspected supplementary area of ​​radar signal feature anomalies, and the suspected supplementary area of ​​cluster-like detection coverage gaps.

[0066] In some embodiments, deep neural networks can be used to determine suspected supplementary areas of radar signal feature anomalies and suspected supplementary areas of cluster-like detection coverage gaps.

[0067] Suspected supplementary areas with abnormal radar signal characteristics are specific regions that require additional detection, identified through deep neural networks.

[0068] The suspected supplementary area for radar signal feature anomalies corresponds to the blank detection area related to signal anomalies in the image.

[0069] The suspected supplementary area for cluster-like detection coverage gaps is a local area that needs to be supplemented for detection when comparing the core features of the same type of anomaly in K clusters using a deep neural network. This is because there are significant differences in the coverage range of the same type of anomaly feature among different clusters. It is used to fill the uncovered areas caused by the gap in feature coverage between these clusters.

[0070] Deep neural networks, through the collaborative computation of fully connected layers and hidden layers, can deeply integrate features such as image grayscale distribution data, abrupt texture density areas, light-dark gradient bands, small brightness difference patches, and ripple break ranges for each anomalous region. Simultaneously, by combining historical feature patterns from K clusters, the model can accurately identify detection gaps directly associated with signal anomalies, thereby determining suspected supplementary areas for radar signal feature anomalies. Furthermore, deep neural networks can learn from the feature distribution of similar anomalies across K clusters to capture the commonalities and differences in coverage of similar anomalies across different clusters, thus identifying areas not currently covered by the anomaly but generally included in other clusters.

[0071] Step S23: Based on the suspected supplementary area of ​​radar signal feature anomalies and the suspected supplementary area of ​​cluster-like detection coverage gaps, determine multiple supplementary detection points for each anomaly area.

[0072] In some embodiments, a deep neural network can be used to determine multiple supplementary detection points for each anomalous region.

[0073] Deep neural networks, with their ability to model complex correlated features, can effectively identify multiple supplementary detection points for each anomaly region. Through a multi-layered network structure, deep neural networks can perform in-depth analysis of the spatial range, feature distribution, and correlation patterns of suspected supplementary areas for radar signal anomalies and suspected supplementary areas for gaps in cluster-like detection coverage. Furthermore, they can transform boundary information and internal feature differences of the supplementary areas into quantifiable spatial coordinate parameters. Simultaneously, deep neural networks can adaptively select the most representative supplementary detection point locations within the supplementary area by incorporating learned patterns from historical detection experience.

[0074] Step S7: Obtain ultrasonic radar data from multiple supplementary detection points.

[0075] The ultrasonic radar data from multiple supplementary detection points consists of ultrasonic propagation data related to the concrete structure at the corresponding locations of each supplementary detection point, collected after using ultrasonic radar equipment to detect multiple supplementary detection points in each anomaly area. The ultrasonic radar data from multiple supplementary detection points includes the propagation speed, propagation time, reflected wave intensity, and sound wave attenuation coefficient of the ultrasonic wave within the concrete at each supplementary detection point.

[0076] The ultrasonic radar data from the supplementary testing points can reflect structural information such as the density of the concrete at those points and the presence of internal defects. For example, in areas with high concrete density, ultrasonic waves propagate faster and have a smaller attenuation coefficient. In areas with internal defects, however, the ultrasonic wave propagation speed decreases, and the intensity of the reflected wave increases.

[0077] Step S8: Determine whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of multiple abnormal points in each abnormal area and the ultrasonic radar data of multiple supplementary detection points.

[0078] In some embodiments, Figure 4 This is a schematic flowchart illustrating the process for determining whether the strength of a large concrete slab is qualified, as provided in an embodiment of the present invention. The process for determining whether the strength of the large concrete slab is qualified includes steps S31 to S35:

[0079] Step S31: Based on the ground-penetrating radar information of multiple abnormal points in each abnormal area and the ultrasonic radar data of multiple supplementary detection points, determine N core drilling test points.

[0080] In some embodiments, Figure 5 This is a schematic flowchart illustrating the process of determining N core test points according to an embodiment of the present invention. The determination of N core test points includes steps S41 to S42:

[0081] Step S41: Based on the ground-penetrating radar information of multiple anomaly points in each anomaly region and the ultrasonic radar data of multiple supplementary detection points, a generative adversarial network is used to generate a risk distribution map of the anomaly region.

[0082] Generative Adversarial Networks (GANs) are deep learning models consisting of two competing neural networks: a generator and a discriminator. The generator learns the distribution of real data and generates new, indistinguishable fake data, while the discriminator's task is to accurately determine whether the input data is real or fake data generated by the generator. Through adversarial training, the generator can ultimately produce highly realistic data.

[0083] The risk distribution map of the anomalous area is generated by processing ground-penetrating radar information from multiple anomaly points and ultrasonic radar data from multiple supplementary detection points within each anomalous area using a generative adversarial network. This image visually illustrates the degree of risk to the concrete slab strength at different locations within the anomalous area. The risk distribution map uses color gradients to represent the level of risk: red indicates a significant impact of structural defects on the concrete slab strength, orange represents a high risk, yellow represents a medium risk, green represents a low risk, and blue represents an extremely low risk.

[0084] A higher risk level indicates a very high probability of a serious defect at that location, and the strength may be severely insufficient. A lower risk level indicates a minor defect or no defect at all.

[0085] Ground-penetrating radar (GPR) data from multiple anomaly points within each anomaly region contains key characteristics such as the location, type, size, and depth of existing defects within the region. These characteristics directly determine the risk of impact on the concrete slab strength at that location. For example, the greater the depth and the wider the area of ​​the defect, the higher the corresponding risk. Ultrasonic radar data from multiple supplementary detection points provides information on the concrete structure characteristics at other locations within the anomaly region, such as density and the presence of latent defects. This data fills the information gaps between existing anomaly points, ensuring that risk calculations cover every location within the anomaly region.

[0086] The generator in a Generative Adversarial Network (GAN) takes ground-penetrating radar (GPR) data from multiple anomaly points in each anomaly region as input and learns the spatial distribution patterns of defects implied by these points. It then attempts to generate a continuous and complete risk distribution map. The discriminator simultaneously receives the risk distribution map generated by the generator and ultrasonic radar data from multiple real supplementary detection points. The discriminator can determine whether the risk values ​​at the corresponding supplementary detection points in the generator's risk distribution map of the anomaly region match the concrete quality reflected by the real ultrasonic data; for example, slow wave velocity and high attenuation correspond to high risk. During training, if the discriminator can easily distinguish the mismatch between the generated image and the real data, it will send gradient signals to the generator to force it to adjust its generation strategy and produce a risk distribution that better reflects reality. Through repeated adversarial games, the generator eventually learns how to fuse GPR information and ultrasonic radar data to generate a high-resolution risk distribution map of the anomaly region that reflects both the macroscopic defect morphology detected by GPR and is highly consistent with the precise point detection results of ultrasonic waves.

[0087] Step S42: Determine N core test points based on the risk distribution map of the abnormal area.

[0088] In some embodiments, a core test point determination model can be used to determine N core test points. The core test point determination model is a deep neural network model. The input to the core test point determination model is the risk distribution map of the abnormal region, and the output of the core test point determination model is N core test points.

[0089] The N core test points are N specific locations on a large concrete slab determined by the core test point determination model for core sampling and testing. Each core test point has precise coordinates on the concrete slab.

[0090] Core drilling test points can reflect the strength status of concrete slab structures to the greatest extent. Core drilling test points can be selected first in areas with the most obvious abnormal characteristics and the greatest potential impact on strength, such as locations where ground penetrating radar information shows large defect depth and wide range, and locations where ultrasonic radar data shows low concrete density.

[0091] The risk distribution map of the abnormal area provides a clear screening basis for determining the model of core drilling test points. The map intuitively and comprehensively displays the degree of risk of each location within the abnormal area affecting the strength of the concrete slab. Different colors and risk values ​​clearly mark high-risk, medium-risk, and low-risk areas. High-risk areas indicate the most severe structural defects and pose the greatest threat to the strength of the concrete slab, making them key target areas for core drilling tests. Medium-risk areas may contain potential serious defects and also require appropriate selection of test points for verification.

[0092] The deep neural network first preprocesses the risk distribution map of the input anomaly region. It converts the color information of the risk distribution into a corresponding risk value matrix, where the value of each pixel represents the risk level at that location. Simultaneously, the deep neural network records the actual spatial coordinates of each pixel on the concrete slab and establishes a mapping between image pixel coordinates and actual spatial coordinates to ensure that the output test point coordinates accurately correspond to the concrete slab. Next, the deep neural network activates a spatial attention mechanism layer to process the risk value matrix. The model calculates the difference between the risk value of each pixel and the average risk value of the entire matrix. The larger the difference, the more prominent the risk at that pixel's location, and the higher the attention weight. In this way, the spatial attention mechanism layer generates an attention weight matrix, amplifying the weight of pixels in high-risk areas while reducing the weight of pixels in low-risk areas, allowing the model to focus on the features of high-risk areas during processing. The model then inputs the preprocessed and attention-mechanism-processed risk value matrix into the hidden layer of the residual network structure. Each layer of the residual network includes convolution operations, batch normalization, and the ReLU activation function. Convolution operations extract local features of the risk distribution, such as the shape, size, and continuity of high-risk areas. Batch normalization accelerates model training and prevents overfitting. The ReLU activation function introduces a non-linear transformation, enabling the model to learn complex patterns in the risk distribution, such as whether high-risk areas are concentrated or whether there are multiple independent high-risk points. Through the processing of multiple layers of the residual network, the model can gradually integrate local risk features into global risk distribution features and establish a correlation between risk features and core test point selection rules. For example, the model learns that when multiple consecutive pixels in a certain area exceed a preset high-risk threshold and the area reaches a certain range, a core test point needs to be selected at the center of that area; and when multiple scattered high-risk pixels exist, test points need to be selected near each high-risk point. The output layer of the deep neural network, through a fully connected layer combined with coordinate regression, can select N optimal core test locations from high-risk areas and key medium-risk areas. The output layer can directly output the actual spatial coordinates of each core test point and verify these coordinates to ensure that the distance between each test point is appropriate, while avoiding excessive concentration or dispersion, and excluding locations unsuitable for core testing such as the edge of concrete slabs and near embedded parts.

[0093] Step S32: Obtain experimental data from N core test points.

[0094] The experimental data from N core test points are experimental measurement data directly related to concrete strength, collected after core sampling experiments were conducted at N core test points. The experimental data for each core test point includes the compressive strength, tensile strength, elastic modulus, and appearance quality data of the concrete core sample obtained from each core test point.

[0095] The appearance quality data of the core sample includes the core sample's dimensional measurements, as well as a description of the core sample's integrity and the presence of defects such as cracks and voids.

[0096] Step S33: Construct a verification map. The verification map includes N core test point nodes and N edges between the core test point nodes. The node features of each core test point node are the experimental data of the core test point, the ground-penetrating radar information of multiple abnormal points in each abnormal region, and the ultrasonic radar data of the multiple supplementary detection points. The edges between the core test point nodes represent the distance between the core test points.

[0097] A data graph is a data structure composed of nodes and edges that can be used to represent complex relationships between data. A validation graph can integrate and associate data from N core test points. The validation graph consists of nodes and edges. Each node represents one of the N core test points. The node features include experimental data from the corresponding core test point, ground-penetrating radar information from multiple anomaly points in each anomaly region, and ultrasonic radar data from multiple supplementary detection points. These node features comprehensively reflect the concrete structure and strength condition of the core test point and its surrounding area.

[0098] In the verification diagram, the edges between the core test point nodes are used to represent the spatial distance relationship between the two core test point nodes on the concrete slab. The attribute value of the edge is the actual straight-line distance between the core test points corresponding to the two nodes.

[0099] The verification map, through this node and edge structure, can link scattered core test point data, anomaly point data, and supplementary detection point data to form a structured whole with a topology.

[0100] Step S34: Process the verification map based on the graph neural network to determine multiple core verification points.

[0101] Graph Neural Networks (GNNs) are deep learning models capable of processing graphs. GNNs can perform operations on the topological structure of a graph through message passing mechanisms between nodes, and can aggregate information from neighboring nodes to update the representation of the central node, thereby learning the structural and feature information of nodes. GNNs can effectively capture complex relationships and patterns in a graph. The input of the GNN is the verification graph, and the output of the GNN is multiple core drilling verification points.

[0102] Multiple core drilling verification points were determined by processing the verification map using a graph neural network. These points were key core drilling points used to judge the direct support strength of large concrete slabs.

[0103] By constructing a verification map, the spatial relationships between N core test points and the comprehensive characteristics of each test point can be clearly reflected. This relationship and feature information are crucial for determining core verification points, because the strength data of a single core test point needs to be combined with the conditions of surrounding test points and its own structural background information to determine whether it is representative or whether there are any anomalies. Using experimental data from core test points, ground-penetrating radar information from multiple anomaly points in each anomaly area, and ultrasonic radar data from multiple supplementary detection points as node features, while using the distance between core test points as edge features, allows for more comprehensive utilization of various data information acquired during the detection process. This helps the graph neural network better understand the relationship between the strength status of different core test points and the surrounding structural environment, as well as the spatial influence between different test points, thus providing a basis for accurately selecting core verification points. Processing the verification map based on a graph neural network can effectively learn the complex relationships and feature transmission between test points, thereby more accurately identifying detection blind spots or data anomaly areas. Compared to relying solely on single test point data to select verification points, the verification map can more comprehensively integrate multi-dimensional information and improve the rationality and targeting of core verification point selection.

[0104] Graph neural networks (Graph Neural Networks) can deeply mine the intrinsic relationships between node features and nodes in a verification graph. Nodes in the verification graph contain experimental data from core drilling test points and information about surrounding structures, while edges represent the spatial distances between test points. Graph Neural Networks can achieve message passing between nodes through graph convolutional layers, and each node can aggregate features from its neighbors based on distance weights. After multiple iterations, local features can be expanded into global features, thereby accurately locating key areas that play a decisive role in strength assessment. These include locations where experimental data exhibits abnormal fluctuations or where features are strongly correlated with surrounding nodes and directly reflect the true strength of the area. Ultimately, by integrating the overall strength distribution trend reflected by global features with the correlation features of local nodes, Graph Neural Networks can identify multiple core drilling verification points that can be used to determine whether the strength of large concrete slabs is up to standard.

[0105] Step S35: Determine whether the strength of the large concrete slab is qualified based on the experimental data of the N core test points and the experimental data of multiple core verification points.

[0106] The experimental data from multiple core drilling verification points are measurement data that directly reflect the concrete strength, obtained after core sampling experiments at multiple core drilling verification points. The experimental data for each core drilling verification point includes the compressive strength, tensile strength, elastic modulus, and appearance quality data of the concrete core sample obtained from each core drilling verification point.

[0107] In some implementations, a strength analysis model can be used to determine whether the strength of a large concrete slab is up to standard. The strength analysis model is a Transformer model. The inputs to the strength analysis model are experimental data from N core test points and experimental data from multiple core verification points. The output of the strength analysis model is whether the strength of the large concrete slab is up to standard.

[0108] Experimental data from N core drilling test points form the foundation for strength assessment and reflect the overall strength distribution trend of large concrete slabs. Experimental data from multiple core drilling verification points further supplement and verify the reliability of the test point data and fill in any blind spots in the testing. Through the experimental data from N core drilling test points and multiple core drilling verification points, a complete strength data system can be formed, thus providing a comprehensive analytical basis for the model.

[0109] Experimental data from N core test points and multiple core verification points contain key indicators such as compressive and tensile strength of different regions of the concrete slab, and are distributed in different spatial locations with complex correlations. The self-attention mechanism of the Transformer model can calculate the correlation weights between data points, and then assign different weights based on spatial distance and strength numerical differences. The model can simultaneously focus on the overall strength distribution trend and local key data, such as accurately identifying the correlation feature that multiple test points and verification points in a certain area have low strength, or the gradient change pattern of strength data in different areas. The encoder part of the Transformer model can transform the experimental data from the scattered N core test points and multiple core verification points into feature vectors containing global strength patterns through multiple layers of self-attention layers and feedforward neural networks. This not only completely preserves the accurate strength information of individual data points, but also integrates the spatial correlation and numerical consistency between data, thus clearly presenting the core characteristics of the overall strength distribution of the concrete slab. The decoder of the Transformer model can decode and analyze the feature vectors converted by the encoder based on preset concrete strength qualification standards. The model first extracts core indicators from the feature vectors, such as the overall average strength of the large concrete slab, the minimum strength value of each region, the number and distribution ratio of test points below the standard strength threshold, and the strength data of key load-bearing areas. Then, each indicator is compared with the qualification standards, for example, whether the overall average strength meets the minimum average value required by the standard, whether the minimum strength of all test points is not lower than the safety threshold specified by the standard, whether the proportion of test points with strength below the standard is controlled within the allowable range, and whether the strength of key load-bearing areas meets the specific design standards. Finally, the model can make a judgment based on the comparison results of all indicators. If all indicators meet the qualification standards, the large concrete slab is deemed to be of qualified strength; if any indicator exceeds the allowable range of the standard, the large concrete slab is deemed to be of unqualified strength.

[0110] Based on the same inventive concept Figure 6 A schematic diagram of a process monitoring system provided in an embodiment of the present invention is shown. The process monitoring system includes:

[0111] Acquisition module 51 is used to acquire infrared images of large concrete slabs;

[0112] Anomaly region determination module 52 is used to determine multiple anomaly regions of the concrete slab based on the infrared image of the large concrete slab using an anomaly region determination model.

[0113] Detection module 53 is used to obtain a ground-penetrating radar image of each abnormal area based on multiple abnormal areas of the concrete slab using ground-penetrating radar detection.

[0114] Anomaly point information determination module 54 is used to determine multiple anomaly point ground-penetrating radar information for each anomaly region based on the ground-penetrating radar image of each anomaly region.

[0115] Clustering module 55 is used to cluster K clusters based on ground-penetrating radar information of multiple anomaly points in each anomaly region;

[0116] The supplementary detection point determination module 56 is used to determine multiple supplementary detection points for each abnormal region based on the ground-penetrating radar image of each abnormal region and the K clusters.

[0117] The ultrasonic data acquisition module 57 is used to acquire ultrasonic radar data from multiple supplementary detection points;

[0118] The strength determination module 58 is used to determine whether the strength of the large concrete slab is qualified based on the ground-penetrating radar information of multiple abnormal points in each abnormal area and the ultrasonic radar data of multiple supplementary detection points.

[0119] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0120] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A process supervision method, characterized by, The method comprises the following steps: acquiring an infrared image of a large concrete slab; determining a plurality of abnormal areas of the concrete slab based on the infrared image of the large concrete slab using an abnormal area determination model; obtaining a ground penetrating radar image of each abnormal area based on the plurality of abnormal areas of the concrete slab using ground penetrating radar detection; determining a plurality of abnormal point ground penetrating radar information of each abnormal area based on the ground penetrating radar image of each abnormal area; performing clustering based on the plurality of abnormal point ground penetrating radar information of each abnormal area to obtain K clusters; determining a plurality of supplementary detection points of each abnormal area based on the ground penetrating radar image of each abnormal area and the K clusters, which comprises the following steps: determining image gray scale distribution data, a local area with suddenly dense texture, a gradual band with gray scale changing from deep to shallow, a small block area with obvious brightness difference from the surrounding, and a local range with broken corrugation of each abnormal area based on the ground penetrating radar image of each abnormal area; determining a radar signal feature abnormal suspected supplementary area and a cluster similar detection coverage gap suspected supplementary area based on the image gray scale distribution data, the local area with suddenly dense texture, the gradual band with gray scale changing from deep to shallow, the small block area with obvious brightness difference from the surrounding, the local range with broken corrugation of each abnormal area, and the K clusters; determining a plurality of supplementary detection points of each abnormal area based on the radar signal feature abnormal suspected supplementary area and the cluster similar detection coverage gap suspected supplementary area; acquiring ultrasonic radar data of the plurality of supplementary detection points; determining whether the strength of the large concrete slab is qualified based on the plurality of abnormal point ground penetrating radar information of each abnormal area and the ultrasonic radar data of the plurality of supplementary detection points, which comprises the following steps: determining N core drilling test points based on the plurality of abnormal point ground penetrating radar information of each abnormal area and the ultrasonic radar data of the plurality of supplementary detection points; acquiring experimental data of the N core drilling test points; constructing a verification graph, wherein the verification graph comprises N core drilling test point nodes and edges between the N core drilling test point nodes, and the node features of each core drilling test point node are the experimental data of the core drilling test point, the plurality of abnormal point ground penetrating radar information of each abnormal area, and the ultrasonic radar data of the plurality of supplementary detection points, and the edges between the core drilling test point nodes represent the distances between the core drilling test points; determining a plurality of core drilling verification points based on processing the verification graph using a graph neural network; determining whether the strength of the large concrete slab is qualified based on the experimental data of the N core drilling test points and the experimental data of the plurality of core drilling verification points.

2. The process supervision method according to claim 1, characterized by, The step of determining N core drilling test points based on the plurality of abnormal point ground penetrating radar information of each abnormal area and the ultrasonic radar data of the plurality of supplementary detection points comprises the following steps: The risk degree distribution map of the abnormal area is generated based on the ground penetrating radar information of multiple abnormal points of each abnormal area and the ultrasonic radar data of the multiple supplementary detection points; N core drilling test points are determined based on the risk degree distribution map of the abnormal area.

3. The process supervision method according to claim 1, characterized by, The abnormal area determination model is a convolutional neural network model.

4. A process supervisory system characterized by, The method comprises the steps of: an acquisition module configured to acquire an infrared image of a large concrete slab; an abnormal area determination module configured to determine multiple abnormal areas of the concrete slab based on the infrared image of the large concrete slab using an abnormal area determination model; a detection module configured to obtain a ground penetrating radar image of each abnormal area based on the multiple abnormal areas of the concrete slab using ground penetrating radar detection; an abnormal point information determination module configured to determine multiple abnormal point ground penetrating radar information of each abnormal area based on the ground penetrating radar image of each abnormal area; a clustering module configured to cluster the multiple abnormal point ground penetrating radar information of each abnormal area to obtain K clusters; a supplementary detection point determination module configured to determine multiple supplementary detection points of each abnormal area based on the ground penetrating radar image of each abnormal area and the K clusters, wherein the determination of the multiple supplementary detection points of each abnormal area based on the ground penetrating radar image of each abnormal area and the K clusters comprises the following steps: determining image gray scale distribution data, a local area with suddenly dense texture, a gradual band with gray scale changing from deep to light, a small area with obvious brightness difference from the surrounding, and a local range with broken corrugation of each abnormal area based on the ground penetrating radar image of each abnormal area; determining radar signal feature abnormal suspected supplementary areas and cluster similar detection coverage gap suspected supplementary areas of each abnormal area based on the image gray scale distribution data, the local area with suddenly dense texture, the gradual band with gray scale changing from deep to light, the small area with obvious brightness difference from the surrounding, the local range with broken corrugation of each abnormal area, and the K clusters; determining the multiple supplementary detection points of each abnormal area based on the radar signal feature abnormal suspected supplementary areas and the cluster similar detection coverage gap suspected supplementary areas; an ultrasonic data acquisition module configured to acquire ultrasonic radar data of the multiple supplementary detection points; a strength determination module configured to determine whether the strength of the large concrete slab is qualified based on the multiple abnormal point ground penetrating radar information of each abnormal area and the ultrasonic radar data of the multiple supplementary detection points, wherein the strength determination module is further configured to: determine N core drilling test points based on the multiple abnormal point ground penetrating radar information of each abnormal area and the ultrasonic radar data of the multiple supplementary detection points; acquire experimental data of the N core drilling test points; construct a verification graph, wherein the verification graph comprises N core drilling test point nodes and edges between the N core drilling test point nodes, and the node features of each core drilling test point node are the experimental data of the core drilling test point, the multiple abnormal point ground penetrating radar information of each abnormal area, and the ultrasonic radar data of the multiple supplementary detection points, and the edges between the core drilling test point nodes represent the distances between the core drilling test points; determine multiple core drilling verification points based on processing of the verification graph by a graph neural network. Determine whether the strength of the large concrete slab is qualified based on experimental data of the N core drilling test points and experimental data of a plurality of core drilling verification points.

5. The process supervisory system of claim 4, wherein, The strength determination module is further configured to: Generate a risk degree distribution map of the abnormal area based on the ground penetrating radar information of a plurality of abnormal points of each abnormal area and the ultrasonic radar data of the plurality of supplementary detection points; Determine N core drilling test points based on the risk degree distribution map of the abnormal area.

6. The process supervisory system of claim 4, wherein, The abnormal area determination model is a convolutional neural network model.

7. An electronic device, comprising: Comprise: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the process supervision method according to any one of claims 1 to 3.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the process supervision method according to any one of claims 1 to 3.

Images