Highway engineering management method and system based on big data analysis

Abstract

The invention relates to the field of engineering management, and particularly discloses a highway engineering management method and system based on big data analysis, and the method comprises the steps: firstly obtaining the data of a temperature sensor disposed in concrete and an image frame of a pouring region in real time, and carrying out the intelligent correlation of the data and the information of a BIM component related to the current pouring, therefore, a dynamic digital twinborn body capable of reflecting the pouring state in real time is generated. On the basis, a pre-trained intelligent analysis model is introduced, abnormal mode detection is carried out on collected temperature time sequence data, and surface defect recognition is carried out on image frames. Particularly, for temperature time sequence data, key time sequence mode features are effectively extracted and strengthened through advanced feature coding and fine-grained enhancement technologies. And meanwhile, the subjectivity and hysteresis of manual inspection are made up through automatic identification of the surface defects by the image analysis model. And finally, the intelligent analysis results are visually presented on the digital twinborn body.

Description

Technical Field

[0001] The present application relates to the field of engineering management, and more specifically, to a highway engineering management method and system based on big data analysis. Background Art

[0002] As a vital national infrastructure, the quality of highway construction is directly related to traffic safety, operational efficiency, and sustainable socioeconomic development. Quality control is crucial during highway construction, especially during concrete pouring. Temperature changes in concrete directly affect its hydration process, strength development, and ultimate durability, while surface defects may indicate internal structural problems or affect its service life. Traditional highway project management methods often rely on manual inspections, empirical judgment, and post-project sampling inspections. This approach is not only inefficient and costly, but also suffers from strong subjectivity, data lag, and difficulty in fully understanding the real-time conditions at the construction site. This leads to delayed problem discovery and difficulty in effectively preventing and controlling quality risks. Once quality problems occur, the subsequent repair costs are enormous and may even affect the overall progress and safety of the project.

[0003] Therefore, an optimized highway engineering management solution is desired. Summary of the Invention

[0004] To address the aforementioned technical problems, the present application is proposed. Embodiments of the present application provide a highway project management method and system based on big data analysis. The method first acquires real-time data from temperature sensors deployed within concrete and image frames of the pouring area, intelligently correlating these data with BIM component information related to the current pour, thereby generating a dynamic digital twin that reflects the pouring status in real time. Furthermore, a pre-trained intelligent analysis model is introduced to detect abnormal patterns in the collected temperature time series data and identify surface defects in the image frames. Specifically, advanced feature encoding and fine-grained enhancement techniques are used to effectively extract and enhance key time series pattern features from the temperature time series data, improving the accuracy and robustness of temperature anomaly detection and enabling the timely identification of potential internal quality issues. Furthermore, the automated identification of surface defects through the image analysis model mitigates the subjectivity and lag inherent in manual inspections. Ultimately, these intelligent analysis results are intuitively presented on the digital twin, enabling managers to accurately and accurately understand the project quality status in real time, achieving a shift from passive response to active early warning. This effectively addresses the shortcomings of traditional management models and enhances the overall management level and quality control capabilities of highway projects.

[0005] According to one aspect of the present application, a highway engineering management method based on big data analysis is provided, which includes: Acquire temperature time series sensor data collected by temperature sensors deployed inside the concrete and real-time image frames of the pouring area collected by cameras; Get BIM component information related to the current pour; Associating the temperature time series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a dynamic digital twin of the pouring state; Inputting the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, wherein the temperature sensor analysis result is used to indicate whether a temperature curve is abnormal; Inputting the real-time image frame of the pouring area into a pre-trained surface defect detection model to obtain an image analysis result, wherein the image analysis result is used to indicate whether there are surface defects; The temperature sensor analysis result and the image analysis result are displayed.

[0006] According to another aspect of the present application, a highway engineering management system based on big data analysis is provided, which includes: The engineering real-time data acquisition module is used to obtain the temperature time series sensor data collected by the temperature sensor deployed inside the concrete and the real-time image frames of the pouring area collected by the camera; BIM component information acquisition module, used to obtain BIM component information related to the current pouring; A pouring state dynamic digital twin construction module is used to associate the temperature time series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a pouring state dynamic digital twin; a temperature curve anomaly detection module, configured to input the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, wherein the temperature sensor analysis result is used to indicate whether the temperature curve is abnormal; a surface defect detection module, configured to input the real-time image frame of the casting area into a pre-trained surface defect detection model to obtain an image analysis result, wherein the image analysis result is used to indicate whether there are surface defects; The result display module is used to display the temperature sensor analysis result and the image analysis result.

[0007] Compared to existing technologies, this application provides a highway project management method and system based on big data analysis. This method first acquires real-time data from temperature sensors deployed within the concrete and image frames of the pouring area, intelligently associating them with BIM component information related to the current pour, thereby generating a dynamic digital twin that reflects the pouring status in real time. Furthermore, a pre-trained intelligent analysis model is introduced to detect abnormal patterns in the collected temperature time series data and identify surface defects in the image frames. Specifically, advanced feature encoding and fine-grained enhancement techniques are used to effectively extract and enhance key time series pattern features from the temperature time series data, improving the accuracy and robustness of temperature anomaly detection and enabling the timely detection of potential internal quality issues. Furthermore, the automated identification of surface defects through the image analysis model compensates for the subjectivity and lag of manual inspections. Ultimately, these intelligent analysis results are intuitively presented on the digital twin, enabling managers to accurately and accurately grasp the project quality status in real time, achieving a transition from passive response to active early warning, effectively addressing the shortcomings of traditional management models and improving the overall management level and quality control capabilities of highway projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 Flowchart of a highway engineering management method based on big data analysis according to an embodiment of the present application; Figure 2 Schematic diagram of data flow of a highway engineering management method based on big data analysis according to an embodiment of the present application; Figure 3 A flowchart of a method for highway engineering management based on big data analysis according to an embodiment of the present application for inputting the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result; Figure 4 A flowchart of performing fine-grained enhancement of temperature time series pattern features on the temperature time series feature coding vector to obtain a temperature time series enhanced feature coding vector according to a highway engineering management method based on big data analysis according to an embodiment of the present application; Figure 5 This is a block diagram of a highway engineering management system based on big data analysis according to an embodiment of the present application. DETAILED DESCRIPTION

[0010] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0011] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0012] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0013] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0014] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0015] To enhance the refined management of highway projects, the industry has recently begun experimenting with the introduction of digital technologies. For example, Building Information Modeling (BIM) technology has been widely adopted in the engineering design and planning stages, providing rich geometric and non-geometric information, enabling visualization and collaborative management of engineering projects. Furthermore, the development of the Internet of Things (IoT) has enabled the collection of sensor data, with some projects deploying temperature sensors or using cameras for on-site monitoring. However, existing technologies suffer from widespread data silos—a lack of effective integration and correlation between BIM models, sensor data, and image data. This hinders the formation of a unified, dynamic, and real-time project management view. For example, temperature sensor data may be presented only as a curve, disconnected from the specific pouring location; camera images may be used solely for manual visual inspection, lacking intelligent analysis. This fragmented data management model makes it difficult for managers to simultaneously grasp project progress and quality status from both a macro and micro perspective, and even more difficult to implement intelligent early warning and decision support based on real-time data.

[0016] This solution aims to provide a highway project management method and system based on big data analysis to address existing issues such as inefficient highway project management, data fragmentation, and insufficient quality risk warnings. Its core concept is to deeply integrate multi-source heterogeneous data to build a dynamic digital twin, enabling real-time, visual, and intelligent monitoring of highway projects, particularly the concrete pouring process.

[0017] Specifically, this solution first acquires real-time data from temperature sensors deployed within the concrete and image frames of the pouring area. It then intelligently correlates this data with BIM component information related to the current pour, generating a dynamic digital twin that reflects the pouring status in real time. This innovative correlation mechanism effectively resolves the data silos between BIM models, sensor data, and image data in traditional management. It integrates and presents all key information in a unified, spatially contextualized virtual environment, providing managers with a comprehensive and intuitive view of the project status. Furthermore, this method incorporates a pre-trained intelligent analysis model to detect anomaly patterns in the collected temperature time series data and identify surface defects in the image frames. Specifically, for temperature time series data, this method effectively extracts and enhances key time series pattern features through advanced feature encoding and fine-grained enhancement techniques, significantly improving the accuracy and robustness of temperature anomaly detection, enabling the timely identification of potential internal quality issues. Furthermore, the automated identification of surface defects through the image analysis model mitigates the subjectivity and lag inherent in manual inspections. Ultimately, these intelligent analysis results are intuitively presented on the digital twin, allowing managers to accurately grasp the project quality status in real time and achieve a shift from passive response to active early warning, thereby effectively solving the shortcomings of the traditional management model and improving the overall management level and quality control capabilities of highway projects.

[0018] In the technical solution of this application, a highway engineering management method based on big data analysis is proposed. Figure 1 This is a flowchart of a highway engineering management method based on big data analysis according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the highway engineering management method based on big data analysis according to the embodiment of the present application. Figure 1 and Figure 2 As shown, the highway engineering management method based on big data analysis according to an embodiment of the present application includes the following steps: S100, obtaining temperature time series sensor data collected by a temperature sensor deployed inside the concrete and a real-time image frame of the pouring area collected by a camera; S200, obtaining BIM component information related to the current pouring; S300, associating the temperature time series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a dynamic digital twin of the pouring state; S400, inputting the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, and the temperature sensor analysis result is used to indicate whether there is an anomaly in the temperature curve; S500, inputting the real-time image frame of the pouring area into a pre-trained surface defect detection model to obtain an image analysis result, and the image analysis result is used to indicate whether there is a surface defect; S600, displaying the temperature sensor analysis result and the image analysis result.

[0019] Specifically, in step S100 and step S200, the temperature time series sensor data collected by the temperature sensor deployed inside the concrete and the real-time image frame of the pouring area collected by the camera are obtained, and the BIM component information related to the current pouring is obtained. It should be understood that The purpose of acquiring time-series temperature sensor data inside concrete is to monitor the dynamic changes in concrete hydration heat in real time. Concrete releases a large amount of hydration heat during its setting and hardening process. If the temperature is too high or the temperature difference is too large, it can easily lead to internal stress concentration, causing cracks and seriously affecting the durability of the structure. By continuously collecting time-series temperature data, the temperature field distribution and changing trends inside the concrete can be accurately understood, providing a scientific basis for timely adjusting maintenance measures and preventing temperature cracks, thereby ensuring the inherent quality of the concrete. Specifically, in one embodiment of this application, before concrete pouring, high-precision, corrosion-resistant temperature sensors (such as thermocouples, thermistors, or fiber optic sensors) are embedded in key locations within the concrete component to be poured and ensure a reliable connection with an external data acquisition unit (such as a wireless data transmission module or a wired data logger). These acquisition units transmit temperature data in real time to a central data server or cloud platform via wireless networks (such as LoRa, NB-IoT, Wi-Fi) or wired networks (such as Ethernet).

[0020] At the same time, the purpose of obtaining real-time image frames of the pouring area is to conduct intuitive and real-time visual monitoring of the concrete surface quality and pouring process. Through image analysis, concrete surface defects (such as cracks, bubbles, segregation, water seepage, etc.) can be discovered in a timely manner, the uniformity and density of pouring can be evaluated, and the construction workers' operational specifications can be monitored. This makes up for the subjectivity and lag of traditional manual inspections and realizes early warning of surface quality problems. Specifically, in a specific example of the present application, a high-resolution industrial camera or network camera is set up above or around the pouring area, and the real-time video stream is transmitted to the edge computing device or the central video management system via wired (such as Ethernet) or wireless (such as Wi-Fi, 4G / 5G). These devices perform preliminary processing and compression on the image frames, and further transmit them to the cloud or local server for subsequent intelligent analysis.

[0021] In addition, obtaining BIM component information related to the current pour is the key to achieving data fusion and context association. The BIM model contains rich information such as the geometric dimensions, material properties, design parameters, and construction progress of engineering components. Associating real-time sensor data and image data with BIM component information can map abstract digital information to specific physical components, allowing managers to intuitively view the real-time temperature curve and surface defect location of specific components in the three-dimensional model, thereby achieving data visualization, problem location, and decision support. Specifically, in a specific example of this application, during the project start-up phase, BIM component data related to the current pouring task, including component ID, geometric coordinates, material type, design strength, etc., is exported from BIM design software (such as Revit, Tekla Structures, etc.) or obtained through an API interface. This BIM information is stored in the project database, and when real-time data is accessed, the sensor data and image data are bound to the corresponding BIM component through the component ID or spatial matching algorithm, laying the foundation for the subsequent digital twin construction.

[0022] Specifically, in step S300, the temperature time-series sensor data and the real-time image frames of the pouring area are associated with the BIM component information related to the current pour to generate a dynamic digital twin of the pouring status. It should be understood that existing highway project management often relies on disparate data sources. For example, BIM models provide design and planning information, sensors independently collect physical parameters, and cameras provide visual monitoring. These data are independent and lack effective integration and contextual association. This fragmentation makes it difficult for managers to fully and realistically understand the true state of the concrete pouring process. It is impossible to accurately map abstract temperature curves or images with specific project components, thus hindering the early detection and precise location of potential quality issues. By mapping real-time collected physical data (temperature, image) to BIM components with precise geometry and attribute information, the goal is to create a unified, visual, and spatially contextualized virtual model. This allows managers to intuitively view the internal temperature trends and surface defects of specific components within the 3D BIM model, transforming abstract data into concrete project status. This enables real-time, visual, and contextual monitoring of the concrete pouring process, providing a unified and accurate data foundation for subsequent intelligent analysis and decision-making.

[0023] More specifically, in an embodiment of the present application, the temperature timing sensor data and the real-time image frame of the pouring area are associated with the BIM component information related to the current pouring to obtain a dynamic digital twin of the pouring status, including: calling the corresponding digital twin based on the BIM component information related to the current pouring; updating the temperature timing sensor data and the real-time image frame of the pouring area as real-time attributes to the corresponding digital twin to obtain the dynamic digital twin of the pouring status.

[0024] Specifically, in step S400, the temperature time-series sensor data is input into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, which indicates whether the temperature curve is abnormal. It should be understood that traditional manual monitoring and empirical judgment have significant limitations when processing massive, continuous concrete internal temperature data. Temperature changes during concrete hydration are a key indicator of its strength development and potential defects (such as temperature cracks). However, continuously monitoring and accurately identifying subtle abnormal patterns in complex temperature curves is not only inefficient and time-consuming, but also highly susceptible to subjective factors, leading to misjudgments or omissions, resulting in the failure to timely detect internal quality risks and missing the optimal opportunity for intervention. By utilizing a pre-trained temperature anomaly model, the system can automatically learn and identify normal temperature patterns and, based on this, accurately capture any temperature fluctuations or abnormal trends that deviate from the normal range. This enables the system to objectively and efficiently determine whether the temperature curve is abnormal, transforming the tedious manual data analysis process into intelligent machine judgment, providing project managers with clear and timely anomaly indications.

[0025] Figure 3 The present invention provides a flow chart of inputting the temperature time series sensor data into a pre-trained temperature anomaly model according to the highway engineering management method based on big data analysis in an embodiment of the present application to obtain a temperature sensor analysis result, wherein the temperature sensor analysis result is used to indicate whether there is an anomaly in the temperature curve. Figure 3 As shown, according to the highway engineering management method based on big data analysis of an embodiment of the present application, step S400 includes: S410, inputting the temperature time series sensor data into a temperature time series encoder based on an LSTM model to obtain a temperature time series feature coding vector; S420, performing fine-grained enhancement of the temperature time series pattern features on the temperature time series feature coding vector to obtain a temperature time series enhanced feature coding vector; S430, calculating the offset between the temperature time series enhanced feature coding vector and the normal temperature time series prior feature coding vector; S440, determining the temperature sensor analysis result based on a comparison between the offset and a preset threshold.

[0026] Specifically, in step S410, the temperature time series sensor data is input into a temperature time series encoder based on an LSTM model to obtain a temperature time series feature encoding vector. It should be understood that raw temperature time series data is typically high-dimensional, continuous, and potentially noisy. Furthermore, the temperature curve of concrete hydration heat is not simply a collection of numerical values; it contains complex time series patterns, trend changes, and potential nonlinear relationships. These dynamic characteristics are crucial for accurately determining the health of concrete and the presence of anomalies. Traditional statistical methods or simple threshold judgments struggle to effectively capture these deep temporal dependencies and pattern characteristics, thus limiting the accuracy and robustness of anomaly detection. Models based on LSTM (Long Short-Term Memory) networks are highly favored for their superior ability to process sequential data. They can learn and memorize long-term dependencies in time series, effectively overcoming the vanishing or exploding gradient problems of traditional recurrent neural networks. Therefore, by inputting the temperature time series sensor data into the LSTM encoder, the goal is to transform it from its raw, redundant form into a more compact and representative temperature time series feature encoding vector. This vector condenses the key dynamic information of the temperature curve, such as heating rate, peak characteristics, cooling trend and fluctuation pattern, thereby converting complex time series data into a low-dimensional, high-information density representation that can be efficiently analyzed by subsequent models.

[0027] Specifically, in step S420, the temperature time series feature encoding vector is subjected to fine-grained enhancement of temperature time series pattern features to obtain a temperature time series enhanced feature encoding vector. It should be understood that although the LSTM encoder can capture the overall dynamics and long-term dependencies in temperature time series data, its output feature vector may still fail to fully highlight certain local, subtle patterns or critical instantaneous changes that are crucial for anomaly detection. Concrete temperature anomalies sometimes manifest themselves not as dramatic deviations from the overall trend, but rather as subtle features of temperature changes within a specific time period, such as local anomalies in heating / cooling rates, plateaus or fluctuations that should not occur at a certain stage, and so on. The original LSTM encoding may not adequately represent these fine-grained features or may mix important information with less important information, thereby affecting the sensitivity and accuracy of subsequent anomaly detection. Therefore, the temperature time series feature encoding vector generated by the LSTM encoder is further processed and optimized to generate a temperature time series enhanced feature encoding vector with higher information density and stronger discriminative power. Specifically for concrete temperature monitoring scenarios, this means fragmenting the one-dimensional temperature time series feature encoding vector into a series of shorter time series feature segments, and then reshaping it into a feature graph structure. This is done to enable the use of mechanisms similar to convolutional neural networks to capture the local dependencies and contextual information of the temperature time series pattern, that is, to analyze the correlation between different small segments in the temperature curve and their local characteristics. Subsequently, a spatial attention mechanism is introduced to identify significant regions, that is, to automatically learn and focus on those feature segments or patterns that contribute most to determining whether the temperature is abnormal. For example, in the early stages of concrete curing, a specific slope change in the heating stage may be more critical than the small fluctuations in the later stable stage. Therefore, the goal is to extract and strengthen these fine-grained patterns that are most indicative of abnormality judgments from the original encoded temperature time series feature encoding vector.

[0028] Figure 4 The flowchart of the highway engineering management method based on big data analysis according to the embodiment of the present application is to perform fine-grained enhancement of the temperature time series pattern feature of the temperature time series feature coding vector to obtain the temperature time series enhanced feature coding vector. Figure 4 As shown, according to the highway engineering management method based on big data analysis of an embodiment of the present application, step S420 includes: S421, performing feature segment segmentation on the temperature time series feature coding vector based on a preset window to obtain a set of temperature time series initial feature segment coding vectors; S422, performing fragment-based key content perception on the set of temperature time series initial feature segment coding vectors to obtain a temperature time series feature segment content space significant perception coding feature map; S423, performing feature shape reshaping on the temperature time series feature segment content space significant perception coding feature map to obtain the temperature time series enhanced feature coding vector.

[0029] More specifically, step S421 involves segmenting the temperature time series feature encoding vector into feature segments based on a preset window to obtain a set of initial temperature time series feature segment encoding vectors. It should be understood that although the LSTM encoder has captured the overall dynamics and long-term dependencies of the temperature time series, the single feature vector it outputs may still smooth or fuse some local, subtle, or short-term temperature patterns that are critical for abnormality judgment. Concrete temperature anomalies, such as sudden changes in the heating rate during a specific curing stage, abnormal fluctuations within a short period of time, or deviations from the expected pattern within a specific time window, may not be prominent enough in the macroscopic feature vector. In order to more deeply explore these subtle clues that may indicate early problems, it is necessary to decompose the macroscopic feature vector for more detailed local pattern analysis. Specifically, in a specific example of the present application, the temperature time series feature encoding vector can be segmented into feature segments using a preset window in a uniform manner, thereby structurally decomposing the overall "temperature time series feature encoding vector" output by the LSTM. Specifically, by applying a "preset window" to the time dimension, this long vector is split into multiple shorter "temperature time series initial feature segment encoding vectors." Each segment vector represents the local characteristic pattern of the original temperature curve within a specific, short time window. This aims to transform the originally continuous, holistic feature representation into an ordered sequence of segments that incorporate local contextual information, facilitating subsequent deep content perception and local dependency analysis using techniques such as convolution operations.

[0030] Accordingly, according to an embodiment of the present application, step S422 performs fragment-based key content perception on the set of the temperature time series initial feature segment coding vectors to obtain a temperature time series feature segment content space-significant perception coding feature map, including: reshaping the set feature shape of the temperature time series initial feature segment coding vectors into a temperature time series feature segment shape reshaping feature map, and then performing convolutional coding-based content perception to obtain a temperature time series feature segment content-sensing coding feature map; inputting the temperature time series feature segment content-sensing coding feature map into a spatial attention layer to obtain the temperature time series feature segment content-sensing coding feature map.

[0031] More specifically, after the set characteristic shape of the temperature time series initial feature segment encoding vector is reshaped into a temperature time series feature segment shape reshaped feature map, content perception based on convolutional coding is performed on it to obtain a temperature time series feature segment content perception encoding feature map, which is expressed as follows: , , in, is the set of encoding vectors of initial feature segments of temperature time series, The first, second, and third in the set of the temperature time series initial feature segment encoding vectors and The initial feature fragment encoding vector of the temperature time series, Reshape the feature shape, is convolutional coding, Content-aware encoding feature map for temperature temporal feature segments.

[0032] It should be understood that although the previous step of feature segmentation decomposes the macroscopic temperature time series feature encoding vector into a sequence of locally focused segments, these initial temperature time series feature segment encoding vectors themselves remain one-dimensional, and the potential correlations and more subtle internal structural patterns between them have not yet been fully explored. Simply analyzing these segments independently, or understanding their relationships solely by their order in the original sequence, may fail to capture the more complex local temperature time series feature patterns composed of multiple adjacent segments, or the finer dynamic changes within the segments. Therefore, in the technical solution of this application, the one-dimensional "set of initial temperature time series feature segment encoding vectors" is further organized into a two-dimensional "temperature time series feature segment reshape feature map." The purpose of this step is to create an image-like data structure that can leverage mature feature extraction paradigms in the field of computer vision, particularly the powerful capabilities of convolutional neural networks (CNNs). Subsequently, this temperature time series feature segment reshape feature map is subjected to content-aware convolutional coding. This aims to leverage the inherent local connectivity and weight sharing properties of convolutional layers to automatically learn and extract local correlations, temporal dependencies, or other potential structural patterns between and within these temperature feature segments. In concrete temperature monitoring, this means that the convolution kernel can scan this feature map and identify, for example, specific combined patterns of temperature changes within several consecutive time segments (such as a rapid temperature rise followed by an abnormal plateau period), or subtle features of temperature changes within a segment (such as small fluctuations or irregularities), which may indicate abnormalities in the concrete hydration process.

[0033] More specifically, the temperature time series feature segment content perceptual coding feature map is input into the spatial attention layer to obtain the temperature time series feature segment content spatial saliency perceptual coding feature map, which is expressed as: , , , , in, is the global average pooling process, is the local system distribution representation vector between temperature time series segments, is a multi-layer perceptron, for activation function, and are the first trainable weight matrix and the first trainable bias vector, respectively. To activate the modulation temperature timing distribution vector, and are the second trainable weight matrix and the second trainable bias vector, respectively, is the static characteristic vector of temperature time series, is the characteristic diagram of the temperature time series dynamic convolution kernel parameters, is the convolution operation, for activation function, is the temperature time series spatial significance feature map, is the point product by position, Content-space saliency perceptual encoding feature map for temperature temporal feature segments.

[0034] It should be understood that although convolutional coding effectively extracts local correlations and underlying patterns between temperature feature segments, not all extracted features or all regions on the feature map are equally important for ultimately determining whether a temperature anomaly exists. In complex temperature variations, certain time segments or combinations thereof (represented as specific spatial regions on the feature map) may contain key clues indicating an anomaly, while other regions may only contain routine fluctuations or redundant information. If all features are treated indiscriminately, key signals may be buried in a large amount of secondary information, thereby affecting the sensitivity and accuracy of anomaly detection. Therefore, in the technical solution of this application, the content-aware encoded feature map of the temperature time series feature segments is further input into a spatial attention layer to selectively enhance the information on the content-aware encoded feature map of the temperature time series feature segments. The spatial attention layer, through a learned or preset mechanism, assigns different importance weights to different spatial locations (i.e., different feature segments or combinations thereof) on the content-aware encoded feature map of the temperature time series feature segments. In the context of concrete temperature monitoring, this means that the system can automatically identify and focus on those feature segment combinations or patterns that best reflect concrete hydration heat anomalies, potential crack risks, or other quality issues. For example, if a sudden drop in temperature or sustained high temperature during a specific time period is a key anomaly indicator, the areas in the feature map corresponding to these segments will be given a higher attention weight. After spatial attention processing, the parts of the feature map that contribute most to the judgment of temperature anomalies are highlighted, while those interfering or unimportant information is weakened. This enables subsequent analysis or classification tasks to more effectively utilize this key information, thereby improving the detection accuracy and robustness of concrete temperature anomalies (especially those subtle but critical anomalies). For example, the system can more accurately capture local temperature anomalies that may accompany the development of early micro-cracks, or local overcooling / overheating caused by improper covering during the maintenance process. Even if these signals are not very obvious in the overall temperature curve, they can be effectively amplified through the attention mechanism, ultimately improving the level of intelligence in highway engineering quality monitoring.

[0035] More specifically, in step S423, the temperature time series feature segment content space salient perceptual coding feature map is reshaped to obtain the temperature time series enhanced feature coding vector, which is expressed as: , in, It is the spatially significant perceptual coding feature map of the temperature temporal feature segment content, Reshape the feature shape, A feature encoding vector is enhanced for the temperature time series.

[0036] It should be understood that although the spatially salient perceptual encoding feature map of the temperature time series feature segments has extracted and enhanced the most relevant local patterns and salient regions for temperature anomaly judgment through convolutional coding and spatial attention mechanisms, it still represents a two-dimensional (or higher-dimensional) feature map structure. Subsequent temperature anomaly models (such as a fully connected neural network classifier or a traditional machine learning model) typically expect a one-dimensional feature vector as input. Therefore, to effectively convey this deeply processed and optimized feature information to the final decision module, the feature map must be converted back into a flattened one-dimensional vector. Therefore, the spatially salient perceptual encoding feature map of the temperature time series feature segments is further reshaped to integrate and compress the two-dimensional spatially salient perceptual encoding feature map containing spatial saliency information, generating a final one-dimensional enhanced temperature time series feature encoding vector. This process is more than a simple dimensional transformation; it condenses the essential information extracted from all previous steps (fragmentation, convolutional content perception, and spatial attention) into a unified vector representation that can be directly used by downstream models.

[0037] Preferably, when generating the temperature time series dynamic convolution kernel parameter characteristic map In the process of , the global mean pooling on the channel dimension is performed through the content-aware encoding feature map of the temperature time series feature fragment to obtain the local system distribution representation vector between the temperature time series fragments. Then, in the weight matrix and and the bias vector and Pair Vector During the modulation process, when the distribution of the local system of the fragment unit depends on the parameter distribution modulation of the weight matrix and the bias vector, it is also expected to form a closed path through the parameter distribution modulation, so that the temperature time series dynamic convolution kernel parameter characteristic map Can improve the accuracy of feature extraction.

[0038] Therefore, if and As a surface state adjustment in the weight parameter space, the bias vector and Undoubtedly, additional phase parameters will be introduced. In order to maintain the surface state from phase localization defects or chaotic disturbances, the surface plane state is first introduced as follows: , , in, is the surface plane state representation vector of the first temperature time series feature fragment content, It is the surface plane state characterization vector of the second temperature time series feature segment content.

[0039] Then, considering that the weight matrix constitutes the transformation matrix component of the surface state adjustment transformation, the transformation operator vector is weighed as: , in, is the temperature time series feature fragment content conversion operator vector.

[0040] In this way, the symmetry characteristics under the influence of the matrix components can be determined according to the polarization selectivity of the bias vector that introduces the additional phase: , , in, is matrix multiplication, is the matrix two norm, is the first trainable bias vector after optimization, is the second trainable bias vector after optimization.

[0041] That is, by biasing the vector and Iterative optimization can be performed to introduce the rotation phase perception of the local connection through the bias vector. For example, the defect of the additional phase causes the surface state of the weight modulation to be disturbed, thereby making the weight matrix and and the bias vector and Parameter distribution modulation for the temperature time series dynamic convolution kernel parameter characteristic diagram The generation forms a closed path, improving the temperature time series dynamic convolution kernel parameter characteristic map The fidelity of feature extraction.

[0042] Specifically, in step S430, the degree of deviation between the temperature time series enhanced feature coding vector and the normal temperature time series prior feature coding vector is calculated. It should be understood that the temperature time series enhanced feature coding vector has highly concentrated the key dynamic information and fine-grained patterns in the current concrete temperature evolution process. However, to determine whether the current state is abnormal, a clear reference benchmark is required. This benchmark is the normal temperature time series prior feature coding vector, which represents the characteristic representation of the temperature change pattern that concrete should have under ideal or standard conditions in similar environments, mix proportions and construction processes. It is obtained by learning and encoding a large amount of historical normal maintenance data. Simply observing the temperature time series enhanced feature coding vector itself is difficult to directly determine whether it deviates from normal, so a quantitative comparison method is required. Based on this, in the technical solution of the present application, the degree of deviation between the temperature time series enhanced feature coding vector and the normal temperature time series prior feature coding vector is further calculated.

[0043] More specifically, in a specific example of the present application, calculating the offset between the temperature time series enhanced feature coding vector and the normal temperature time series prior feature coding vector includes: calculating the Euclidean distance between the temperature time series enhanced feature coding vector and the normal temperature time series prior feature coding vector as the offset. By calculating the Euclidean distance between these two high-dimensional feature vectors, the complex, multi-dimensional feature differences can be converted into a single, intuitive scalar value - "offset". This "offset" directly reflects the "distance" of the current temperature time series pattern from the "ideal" or "standard" pattern in the feature space. As a commonly used measurement method, Euclidean distance can effectively measure the similarity or difference between two vectors in the feature space: the smaller the distance, the closer the current temperature pattern is to the normal pattern; the larger the distance, the greater the deviation and the higher the possibility of potential anomaly. This quantified offset provides a key and interpretable input for subsequent temperature anomaly model judgment.

[0044] Specifically, in step S440, the temperature sensor analysis result is determined based on the comparison between the offset and a preset threshold. It should be understood that although the "offset" (i.e., the Euclidean distance between the "temperature time series enhanced feature encoding vector" and the "normal temperature time series prior feature encoding vector") has quantified the degree of difference between the current concrete temperature state and the normal state, this continuous numerical value itself cannot directly give a clear "normal" or "abnormal" conclusion. Engineering management and decision-making require a clear, binary judgment result to facilitate the implementation of appropriate measures. Therefore, introducing a "preset threshold" as a decision boundary and mapping the continuous offset to a discrete abnormal state judgment is a key step in achieving automated decision-making.

[0045] More specifically, in one embodiment of the present application, the temperature sensor analysis result is determined based on a comparison between the deviation and a preset threshold. The results include: determining that the temperature curve is abnormal if the deviation is greater than the preset threshold; and determining that the temperature curve is normal if the deviation is less than or equal to the preset threshold. By comparing the calculated deviation with a preset threshold (typically based on historical data analysis, engineering experience, or risk assessment), the system can automatically determine the health of the current temperature curve without manual intervention. If the deviation exceeds the threshold, it means that the current temperature pattern has deviated from the normal pattern to an unacceptable degree, sufficient to be considered abnormal. Conversely, if the deviation does not exceed the threshold, the current temperature variation is considered to be within the acceptable normal range. This process transforms complex feature space distances into a simple, clear "yes / no" anomaly signal, namely the "temperature sensor analysis result."

[0046] Specifically, in step S500, the real-time image frames of the pouring area are input into a pre-trained surface defect detection model to obtain image analysis results, and the image analysis results are used to indicate whether there are surface defects. It should be understood that in highway engineering management, especially in the quality control of concrete pouring, traditional manual visual inspection of concrete surface quality has significant limitations. Manual inspection is not only labor-intensive and inefficient, but also easily affected by subjective factors, experience level, and environmental factors such as lighting conditions, resulting in the stability and consistency of the detection results being difficult to ensure, and it is even more difficult to achieve comprehensive, continuous and real-time monitoring of a large pouring area, which may result in missing early minor defects and missing the best time for treatment. Therefore, in the technical solution of the present application, the real-time image frames of the pouring area are further input into a pre-trained surface defect detection model to obtain image analysis results. In particular, in a specific example of the present application, the pre-trained surface defect detection model is a pre-trained YOLOv5 model. By utilizing advanced computer vision technology, especially efficient target detection models like YOLOv5, it aims to overcome the shortcomings of manual inspection and can quickly and accurately automatically identify and locate various defects that may appear on the concrete surface from real-time images, such as cracks, pitting, holes, exposed rebar, cold joints, etc., and output structured analysis results to provide project managers with immediate and reliable surface quality information.

[0047] Specifically, in step S600, the temperature sensor analysis results and the image analysis results are displayed. It should be understood that presenting internal temperature data or surface defect information in isolation makes it difficult for project managers to form a comprehensive, three-dimensional understanding of concrete pouring quality. Traditional management methods often result in data fragmentation: temperature monitoring is one system, and appearance inspection is another process. The two processes are not well integrated, resulting in managers lacking a unified, intuitive view that accurately corresponds to the actual project location when determining the root cause of problems, assessing risks, and formulating response strategies. Therefore, the temperature sensor analysis results and the image analysis results are further displayed in the dynamic digital twin of the pouring state. By simultaneously displaying the internal temperature analysis results and the external surface defect analysis results in the dynamic digital twin, the goal is to closely link the abstract analysis data with the actual three-dimensional spatial position of the concrete component. As a virtual mapping of the physical entity, the digital twin can carry multi-source heterogeneous data and integrate and present this data in a unified virtual environment. In this way, managers can not only see whether there are temperature abnormalities or surface defects, but also clearly understand where these abnormalities or defects occur in which components and whether there are potential correlations between them (for example, whether the internal temperature abnormality area corresponds to the external crack).

[0048] In summary, the highway engineering management method based on big data analysis according to the embodiment of the present application is explained. It first obtains the temperature sensor data deployed inside the concrete and the image frames of the pouring area in real time, and intelligently associates them with the BIM component information related to the current pouring, thereby generating a dynamic digital twin that can reflect the pouring status in real time. On this basis, a pre-trained intelligent analysis model is introduced to detect abnormal patterns in the collected temperature time series data and identify surface defects in the image frames. In particular, for the temperature time series data, advanced feature encoding and fine-grained enhancement technology are used to effectively extract and enhance key time series pattern features, improve the accuracy and robustness of temperature anomaly detection, and thus enable the timely detection of potential internal quality problems. At the same time, the automatic identification of surface defects through the image analysis model compensates for the subjectivity and lag of manual inspections. Finally, these intelligent analysis results are intuitively presented on the digital twin, allowing managers to grasp the project quality status in real time and accurately, realizing the transition from passive response to active early warning, thereby effectively solving the drawbacks of the traditional management model and improving the overall management level and quality control capabilities of highway projects.

[0049] Furthermore, a highway engineering management system based on big data analysis is also provided.

[0050] Figure 5 FIG is a block diagram of a highway engineering management system based on big data analysis according to an embodiment of the present application. Figure 5As shown, according to an embodiment of the present application, a highway engineering management system 500 based on big data analysis includes: a real-time engineering data acquisition module 510, which is used to obtain temperature time-series sensor data collected by a temperature sensor deployed inside the concrete and a real-time image frame of the pouring area collected by a camera; a BIM component information acquisition module 520, which is used to obtain BIM component information related to the current pouring; a pouring state dynamic digital twin construction module 530, which is used to associate the temperature time-series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a dynamic digital twin of the pouring state; a temperature curve anomaly detection module 540, which is used to input the temperature time-series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, and the temperature sensor analysis result is used to indicate whether there is an anomaly in the temperature curve; a surface defect detection module 550, which is used to input the real-time image frame of the pouring area into a pre-trained surface defect detection model to obtain an image analysis result, and the image analysis result is used to indicate whether there is a surface defect; a result display module 560, which is used to display the temperature sensor analysis result and the image analysis result.

[0051] As described above, the highway engineering management system 500 based on big data analysis according to the embodiment of the present application can be implemented in various wireless terminals, such as a server having a highway engineering management algorithm based on big data analysis. In one possible implementation, the highway engineering management system 500 based on big data analysis according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the highway engineering management system 500 based on big data analysis can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the highway engineering management system 500 based on big data analysis can also be one of the many hardware modules of the wireless terminal.

[0052] While various embodiments of the present disclosure have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A highway engineering management method based on big data analysis, characterized in that: include: Acquire temperature time series sensor data collected by temperature sensors deployed inside the concrete and real-time image frames of the pouring area collected by cameras; Get BIM component information related to the current pour; Associating the temperature time series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a dynamic digital twin of the pouring state; Inputting the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, wherein the temperature sensor analysis result is used to indicate whether a temperature curve is abnormal; Inputting the real-time image frame of the pouring area into a pre-trained surface defect detection model to obtain an image analysis result, wherein the image analysis result is used to indicate whether there are surface defects; The temperature sensor analysis result and the image analysis result are displayed.

2. The highway engineering management method based on big data analysis according to claim 1 is characterized in that: Associating the temperature time series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a dynamic digital twin of the pouring state, including: Based on the BIM component information related to the current pouring, calling the corresponding digital twin; The temperature time series sensor data and the real-time image frame of the casting area are updated to the corresponding digital twin as real-time attributes to obtain the casting state dynamic digital twin.

3. The highway engineering management method based on big data analysis according to claim 1 is characterized in that: Inputting the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, including: Inputting the temperature time series sensor data into a temperature time series encoder based on an LSTM model to obtain a temperature time series feature encoding vector; Performing fine-grained enhancement of temperature time series pattern features on the temperature time series feature coding vector to obtain a temperature time series enhanced feature coding vector; Calculating the offset between the temperature time series enhanced feature coding vector and the normal temperature time series priori feature coding vector; The temperature sensor analysis result is determined based on a comparison between the deviation and a preset threshold.

4. The highway engineering management method based on big data analysis according to claim 3 is characterized in that: Calculating the offset between the temperature time series enhanced feature coding vector and the normal temperature time series prior feature coding vector includes: The Euclidean distance between the temperature time series enhanced feature coding vector and the normal temperature time series priori feature coding vector is calculated as the offset.

5. The highway engineering management method based on big data analysis according to claim 4 is characterized in that: Determining the temperature sensor analysis result based on a comparison between the offset and a preset threshold value includes: In response to the deviation being greater than the preset threshold, determining that the temperature sensor analysis result indicates that an abnormality exists in the temperature curve; In response to the deviation being less than or equal to the preset threshold, it is determined that the temperature sensor analysis result is that there is no abnormality in the temperature curve.

6. The highway engineering management method based on big data analysis according to claim 1 is characterized in that: The pre-trained surface defect detection model is a pre-trained YOLOv5 model.

7. The highway engineering management method based on big data analysis according to claim 6 is characterized in that: Performing fine-grained enhancement of temperature time series pattern features on the temperature time series feature coding vector to obtain a temperature time series enhanced feature coding vector, including: Performing feature segment segmentation on the temperature time series feature coding vector based on a preset window to obtain a set of temperature time series initial feature segment coding vectors; Performing segment-based key content perception on the set of the temperature time series initial feature segment encoding vectors to obtain a temperature time series feature segment content space salient perception encoding feature map; The temperature time series feature segment content space salient perceptual coding feature map is reshaped to obtain the temperature time series enhanced feature coding vector.

8. The highway engineering management method based on big data analysis according to claim 7 is characterized in that: The set of the temperature time series initial feature segment encoding vectors is subjected to segment-based key content perception to obtain a temperature time series feature segment content spatially significant perceptual encoding feature map, including: After reshaping the set characteristic shape of the temperature time series initial feature segment encoding vector into a temperature time series feature segment shape reshaping feature map, performing content perception based on convolution coding on the reshaped feature map to obtain a temperature time series feature segment content perception coding feature map; The content-perceptual encoding feature map of the temperature time series feature segment is input into the spatial attention layer to obtain the content-saliency-perceptual encoding feature map of the temperature time series feature segment.

9. The highway engineering management method based on big data analysis according to claim 8 is characterized in that: Displaying the temperature sensor analysis result and the image analysis result includes: displaying the temperature sensor analysis result and the image analysis result in the casting state dynamic digital twin.

10. A highway engineering management system based on big data analysis, characterized in that: include: The engineering real-time data acquisition module is used to obtain the temperature time series sensor data collected by the temperature sensor deployed inside the concrete and the real-time image frames of the pouring area collected by the camera; BIM component information acquisition module, used to obtain BIM component information related to the current pouring; A pouring state dynamic digital twin construction module is used to associate the temperature time series sensor data and the real-time image frame of the pouring area with the BIM component information related to the current pouring to obtain a pouring state dynamic digital twin; a temperature curve anomaly detection module, configured to input the temperature time series sensor data into a pre-trained temperature anomaly model to obtain a temperature sensor analysis result, wherein the temperature sensor analysis result is used to indicate whether the temperature curve is abnormal; a surface defect detection module, configured to input the real-time image frame of the casting area into a pre-trained surface defect detection model to obtain an image analysis result, wherein the image analysis result is used to indicate whether there are surface defects; The result display module is used to display the temperature sensor analysis result and the image analysis result.

Images