Highway engineering field management method based on process inspection

Abstract

The invention provides a highway engineering site management method based on process inspection. The highway engineering site management method comprises the following steps: acquiring highway construction site data, standards and technical specifications through multiple channels; analyzing by using a big data analysis technology, constructing a quality monitoring model to perform quality monitoring on the construction process, comparing with standard data, predicting whether the quality is qualified or not, and generating a quality monitoring report at the same time; executing a corresponding strategy according to a quality prediction result, if the predicted quality is qualified, filling related data in a process inspection application form, and if the predicted quality is not qualified, performing risk assessment in combination with a quality monitoring report, scoring a risk assessment result, and triggering early warning based on a score; unqualified influence factors are analyzed and positioned based on correlation, and rectification prevention and control are carried out according to the positioned specific factors; the project quality is monitored in the process of process inspection, the project quality is improved, and meanwhile later data backtracking is facilitated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building information management, in particular to a highway engineering site management method based on process inspection. BACKGROUND

[0002] The highway engineering site management based on process inspection refers to the standardized inspection and acceptance of construction processes through digital means to ensure that the engineering quality meets the specification requirements. However, the current existing technology lacks in the management of highway engineering data, especially in the process of process inspection.

[0003] For example, the inspection application, quality inspection table, and concealed engineering image are mainly in paper or Excel, which needs to be filled in by on-site staff, photographed, printed, signed, scanned, and archived, with an average time consumption of 2-3 days / time. The process inspection process relies on manual work, which is low in efficiency. The quality inspection relies on manual inspection, the detection results are fragmented, and there is a risk of missed detection and misjudgment, making it difficult to form a full life cycle quality archive. The period from engineering inspection to acceptance node is long, and quality abnormalities cannot be found in real time. In addition, the data formats uploaded by different personnel may not be unified, and the key data cannot be associated. Once a quality defect occurs, it is difficult to locate the responsibility and trace the data.

[0004] Therefore, the technical problem to be solved by the present application is how to monitor the quality of the project during the process of process inspection, improve the quality of the project, and facilitate data backtracking in the later period. SUMMARY

[0005] To solve the above problems, the present application provides a highway engineering site management method based on process inspection, which can monitor whether the engineering quality has problems in advance by establishing a quality monitoring model, timely rectification, improve the pass rate of the project, and unify the data format to make the data tracing process simple and fast.

[0006] To achieve the above purpose, the technical solution adopted by the present application is: The present application provides a highway engineering site management method based on process inspection, comprising the following steps: S1: acquiring highway construction site data, standards and technical specifications through multiple channels, and preprocessing the collected data, wherein the construction site data includes construction data, process flow data, equipment data and personnel data; S2: analyzing the collected data using big data analysis technology, constructing a quality monitoring model to monitor the quality of the construction process, comparing with the standard data, predicting whether the quality is qualified, and generating a quality monitoring report; S3: According to the quality prediction result, a corresponding strategy is performed, if the predicted quality is qualified, the relevant data including physical parameters and process parameters in the construction process are filled in the process inspection application table, if the predicted quality is unqualified, risk assessment is combined with the quality monitoring report, and the risk assessment result is scored, and a warning is triggered based on the score; S4: Based on the correlation analysis, the unqualified influencing factors are located and rectified according to the specific factors located.

[0007] Further, a data collaboration platform based on cloud computing is constructed, the collected data are received through the network, the data are stored by using a relational database and a non-relational database, and the association between the data is established for cross-regional synchronization.

[0008] S2 includes the following steps: S21: Based on the association rule mining algorithm, the potential association between the highway construction site data is analyzed; S22: The quality data in the construction process are classified by using a clustering algorithm, and the processes with similar quality characteristics are classified into a category; S23: Taking the highway construction site data and environmental data as input and the engineering quality acceptance result as output label, a machine learning model is constructed and trained by using historical data; S24: According to the highway engineering quality acceptance specification and design requirements, a comprehensive quality monitoring index system is constructed; S25: A quality monitoring model is constructed by combining the big data analysis technology and the quality monitoring index system, and the association rules mined in S21 and the cluster features clustered in S22 are integrated; S26: The current highway construction site data and environmental data are input into the quality monitoring model, the current engineering quality acceptance result is predicted, and a quality detection report is automatically generated.

[0009] Further, step S22 includes the following steps: S221: The quality data of the process are collected and quality features are extracted, the quality features include numerical quality features and indirect quality features, and the core features with high quality correlation are reserved by using correlation analysis; S222: The elbow method is used to determine the number of clusters based on the quality feature differences of the highway engineering processes, and the K-Means algorithm is used for clustering; S223: The quality data of each cluster are statistically analyzed, common features are extracted, and common influencing factors are analyzed; S224: The clustering process is iterated and the cluster features are updated.

[0010] Further, step S223 includes the following steps: S2231: Align the quality data of all processes in the same cluster in time and space, and perform secondary screening on the data in the cluster to eliminate outliers; S2232: Calculate descriptive statistics for each quality indicator in the cluster, and based on the statistical results, extract common quality characteristics; S2233: Combine the influence factor data corresponding to the processes in the cluster to mine the common causes of common quality characteristics; S2234: Verify the identified common characteristic-influence factor correlation; S2235: Extract cluster characteristic labels, sort them by influence intensity, and generate a common influence list.

[0011] The quality monitoring index system constructed in S24 is used as the judgment standard of the model, which specifically includes index threshold and index weight. The model needs to compare the results of machine learning prediction with the threshold of the index system, and at the same time, calculate the comprehensive quality score combined with the index weight. The calculation formula is: Z = ∑α×G; Where Z represents the comprehensive quality score, a represents the weight, and G represents the index compliance rate.

[0012] Further, when the quality monitoring model predicts that the construction quality is unqualified, combined with the quality detection report and from the three dimensions of severity, influence range and occurrence probability, the risk assessment is carried out, each dimension is quantified as a score, the weight of each dimension is determined by the analytic hierarchy process, the risk comprehensive score is calculated, and the formula is: M = Attention(S)×S+Attention(R)×R+Attention(P)×P; Where Attention(S), Attention(R), and Attention(P) represent the weights of each item, M represents the risk score, S represents the severity, R represents the influence range, and P represents the occurrence probability.

[0013] Further, the attention mechanism is introduced to dynamically calculate the weights of severity, influence range and occurrence probability, including the following steps: Collect historical risk assessment data of quality unqualified, each data contains quantized values of severity, influence range and occurrence probability, and corresponding risk consequence labels, and pre-process the features; A lightweight model containing feature embedding layer, attention calculation layer and weight output layer is constructed to map features to high-dimensional feature vectors and enhance the model's ability to capture the correlation between features; Use global features that are strongly related to risk consequences as query vectors to calculate attention scores, and normalize the scores by the Softmax function to get attention weights; The attention weight is combined with the original feature, a risk consequence is predicted through a full connection layer, and the weight value is output.

[0014] S4 comprises the following steps: S41: The potential influencing factor dimension is determined, specific indexes are refined, the specific performance of the quality unqualified is associated with the influencing factor data, and an analysis data set is formed; S42: A multiple linear regression model is constructed with the quantitative indexes of the quality unqualified as the dependent variable and each factor index as the independent variable, and the influence intensity is determined by the absolute value of the regression coefficient; S43: The correlation of each factor is calculated based on the SHAP technology and is sorted; S44: Based on the correlation analysis result, the actual on-site investigation is combined, and the problem root is accurately positioned; S45: According to the specific factor positioned, an executable and traceable rectification scheme is formulated, and the rectification elements are determined; S46: The construction unit executes according to the rectification scheme, and the progress is uploaded in real time.

[0015] Further, the method further comprises: S5, a visual platform is constructed based on the BIM technology, the overall situation and each key index of the construction are displayed in a visual manner in real time, and the data interaction and sharing function between different user roles are provided.

[0016] The beneficial effects of the present application are: The application provides a highway engineering site management method based on process inspection, which can analyze and process data from multiple aspects and collect standardized data by acquiring highway construction site data, standards and technical specifications from multiple channels, so that data comparison in the subsequent process is facilitated and the accuracy of data analysis is improved. The collected data is analyzed by using big data analysis technology, a quality monitoring model is constructed to monitor the quality of the construction process, and the quality is compared with the standard data to predict whether the quality is qualified. The quality of the project can be monitored in the construction process, quality problems can be found in time, and the project quality can be improved. According to the quality prediction result, corresponding strategies are executed, if the predicted quality is qualified, a process inspection application form is filled out, if the predicted quality is unqualified, risk assessment is combined with the quality monitoring report, the risk assessment result is scored, and early warning is triggered based on the score; the predicted quality problem is risk assessed, and the risk level is divided based on the risk assessment result, different countermeasures are taken for different risk levels, so that the risk can be treated early and the safety of the project is improved. The influence factors of unqualified based on correlation analysis are located, and the specific factors are rectified and controlled, the rectification effect is good, and resources are saved. The application realizes the control of the quality of highway engineering in the construction inspection process by the above method. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 Fig. 1 is a process flow diagram of a highway engineering site management method based on process inspection. DETAILED DESCRIPTION

[0018] It should be noted that the application is mainly for the management of highway engineering site data, which is divided into the following processes according to the process, including construction lofting, pile final hole, reinforcement cage processing and installation, pre-pouring acceptance, concrete pouring and pile acceptance. In the concrete pouring process, quality problems are most likely to occur and are relatively hidden and difficult to view, which is determined by the characteristics of the process, the complexity of the process and the site influencing factors. For example, the concrete mixing ratio on-site execution deviation (such as sandstone moisture content not adjusted in real time, resulting in too large water-cement ratio); or concrete initial setting without timely maintenance (especially in high temperature environment) will cause insufficient strength and cannot reach the design strength grade, and the pile foundation cannot bear the upper load, which exists structural safety hazard. If problems occur in this process, it will directly affect the quality of the entire project, and its importance is self-evident. In the prior art, only manual inspection is relied on, it is difficult to find internal quality problems, and there is a risk of missed detection and misjudgment. Therefore, how to realize quality inspection of concrete pouring in the construction process is the main problem to be solved by the application.

[0019] Please refer to Figure 1As shown, the present application relates to a highway engineering site management method based on process inspection, comprising the following steps: S1: Obtain highway construction site data, standards and technical specifications through multiple channels, and pretreat the collected data, the construction site data including construction data, process flow data, equipment data and personnel data; In order to fully understand the situation of the construction site, it is necessary to collect and analyze data in multiple aspects. Specifically, in the present application, highway construction site data, standards and technical specifications are collected, and the collected relevant standards are used for subsequent data comparison, so that the comparison result is more accurate. For example, the standards and construction technical specifications of "Highway Engineering Quality Inspection and Evaluation Standard", "Highway Engineering Technical Standard", "Highway Subgrade Construction Technical Specification", "Highway Pavement Base Construction Technical Specification", "Highway Bridge and Culvert Construction Technical Specification" and "Highway Tunnel Construction Technical Specification".

[0020] Among them, the highway construction site data includes construction data, process flow data, equipment data and personnel data: The construction data includes the environmental data of the construction site, and a plurality of types of sensors can be deployed at the construction site to obtain key parameters. For example, pressure sensors and temperature sensors are installed in the concrete pouring area. The pressure sensor can accurately monitor the pressure change in the pouring process, so as to reflect the compactness of the concrete. The temperature sensor can be used to detect the temperature fluctuation of the concrete during pouring and curing, and prevent cracks and other quality defects of the concrete caused by abnormal temperature.

[0021] For process flow data, a special management software such as smart construction site APP can be developed to cover the whole process from construction preparation to completion acceptance. According to the highway engineering construction specification and the actual needs of the project, the sequence, logical relationship and inspection node of each process are set in the software. After each process is completed, the construction unit submits the process inspection application online through the software, and fills in the starting time, ending time, construction team and other information of the process in detail. After receiving the application, the supervision unit records the audit time, acceptance result, rectification opinion and other contents in the software, leaving traces throughout the whole process, and realizing the complete traceability of the process flow data.

[0022] For equipment data, the mechanical equipment at the construction site is connected to the management system through Internet of Things technology, and the equipment operation data is directly collected from the control system of the equipment, such as the paving speed and thickness of the paver, the rolling speed and rolling times of the road roller, the raw material ratio and mixing time of the mixing station, etc. These data can directly reflect the running status of the equipment and the construction process parameters, and provide key basis for judging the process quality. At the same time, the fault alarm information of the equipment is collected to arrange maintenance in time and ensure the continuity of construction.

[0023] For personnel data, construction personnel and supervisors are equipped with mobile intelligent terminals integrating multiple functions, and personnel information is entered through a customized APP, including basic information such as name, job type, qualification certificate number, and contact information, as well as dynamic information such as personnel attendance records and training status. During construction, the location information of personnel is obtained in real time using the positioning function of the APP, facilitating personnel scheduling and management.

[0024] Further, a data collaboration platform based on cloud computing is constructed, which receives collected data through the network, stores data using relational and non-relational databases, and establishes associations between data for cross-regional synchronization.

[0025] The collected data is processed and stored using the powerful computing power and storage resources of cloud computing, and the platform can be flexibly expanded according to business needs. Using the distributed computing power of cloud computing, large amounts of data can be quickly analyzed and counted. Since highway engineering involves multiple parties such as the owner, the design party, the construction party, the supervisor, and the material supplier, a unified data collaboration platform is constructed to enable all parties to receive engineering information synchronously. Through this platform, data is stored uniformly and associations between data are established, facilitating data searching and tracing.

[0026] Before data storage, the collected data is preprocessed, including data cleaning, data format conversion, and data standardization. Through data cleaning, noise, outliers, and duplicate data in the data are identified and removed. For example, for abnormal pressure data generated by pressure sensors due to instantaneous interference, it is determined to be an outlier and is removed by comparing it with historical data and adjacent sensor data. For repeated transmission data caused by network fluctuations, it can also be accurately identified and deleted. For missing data, interpolation or statistical rules based on related data are used for completion. Since data collected from different data sources have different formats, format unification is required. Binary data collected by sensors and specific format data exported by device control systems are converted into standard formats that can be uniformly processed by the system, such as CSV and JSON, to facilitate subsequent data storage and analysis. Different types of data are standardized to make them comparable. Sensor data of different ranges are normalized to the [0, 1] interval to eliminate dimensional effects and lay a foundation for subsequent data analysis and model construction.

[0027] During data collection, data verification rules are set to perform real-time verification on input data. During data transmission, the SSL / TLS encryption protocol is used to encrypt the transmitted data, preventing data from being stolen or tampered with and ensuring data security. Based on the AES encryption algorithm, sensitive data stored is encrypted, and only authorized users with the correct key can decrypt and read the data.

[0028] S2: Analyze the collected data using big data analysis techniques, construct a quality monitoring model to monitor the construction process, and compare with standard data to predict whether the quality is qualified, and generate a quality monitoring report; Wherein, S2 includes the following steps: S21: Based on the association rule mining algorithm, analyze the potential association between highway construction site data; Based on the association rule mining algorithm, analyze the potential association between construction data, process flow data, equipment data and personnel data. For example, mine the relationship between concrete pouring temperature, raw material ratio and final strength of concrete, and find out the factor combination that significantly affects the strength, to provide basis for optimizing construction technology. Specifically, Apriori algorithm can be used, by analyzing historical data and combining with engineering actual demand, according to the set support threshold, the frequent item set is screened out, and the association rules greater than the confidence threshold are reserved.

[0029] S22: Use clustering algorithm to classify quality data in construction process, and classify similar quality characteristics into one category; S23: Take highway construction site data and environmental data as input, and engineering quality acceptance result as output label, construct machine learning model and train with historical data; Construct machine learning model, such as support vector machine (SVM), random forest (RF), etc. Train a large number of historical construction data, take construction parameters, environmental factors, personnel and equipment information as input features, and engineering quality acceptance result as output label, train the model to learn the internal pattern of quality formation, and continuously optimize the model parameters through cross validation and other methods to improve the accuracy and generalization ability of the model, so that it can accurately predict the quality status under different construction conditions.

[0030] S24: According to highway engineering quality acceptance specification and design requirements, construct a comprehensive quality monitoring index system; The index covers key quality parameters such as subgrade compaction degree, pavement flatness, concrete strength and reinforcement spacing. Set reasonable threshold range for each index as the basis for judging whether the quality is qualified or not. Quantify each index into score, compare the quantified score with the preset threshold, if it exceeds the preset threshold, it is judged as unqualified construction quality.

[0031] S25: Combine big data analysis technology and quality monitoring index system to construct quality monitoring model, and integrate the association rules mined in S21 and the cluster features clustered in S22; Specifically, the quality monitoring model takes highway construction site data and environmental data as the core input, specifically including construction data, environmental data and associated data, the associated data including factor-quality association relationships obtained through S21 association rule mining, and common characteristics of similar processes obtained through S22 clustering analysis. The model takes the engineering quality acceptance result as the output, specifically including qualitative results (the process quality is qualified or unqualified) and quantitative results (the deviation of the predicted value of each quality index from the standard threshold value in the quality monitoring index system).

[0032] The machine learning model trained in S23 is taken as the core calculation module, which has learned the mapping relationship between input parameters and quality results through historical data. The quality monitoring index system constructed in S24 is taken as the judgment standard of the model, specifically including index threshold and index weight. The model needs to compare the predicted result of machine learning with the threshold value of the index system, and at the same time, combine the index weight to calculate the comprehensive quality score, the calculation formula is: Z=∑α×G; Wherein, Z represents the comprehensive quality score, a represents the weight, and G represents the index compliance rate.

[0033] The association rule result is introduced to enhance the interpretability of the model, and the factor-quality association rules mined in S21 are embedded into the model as auxiliary judgment basis. The clustering results are used to refine the model hierarchy, and the similar process clusters obtained in S22 are taken as the hierarchical variables of the model. For different clusters of processes, the model uses targeted prediction parameters to make the prediction more in line with the process characteristics.

[0034] Specifically, after standardizing the input site data, the data is mapped to the corresponding index dimension based on the quality monitoring index system, and the predicted value of each quality index is output after calculation by the machine learning model. Each predicted value is compared with the corresponding threshold value in the index system for single-index judgment, and the comprehensive quality score is calculated combined with the index weight. If the score is less than the preset qualified line, the overall quality is judged as unqualified.

[0035] A dynamic adjustment mechanism is introduced to dynamically optimize the model parameters based on real-time collected data and historical feedback. For example, if the prediction deviation of a certain type of process is continuously >10%, the clustering results of S22 are called to redivide the similar process clusters, and the prediction parameters of the cluster are adjusted. If a new factor-quality association relationship is found, the input features of the model are updated through the association rule mining of S21.

[0036] S26: input the current highway construction site data and environmental data into the quality monitoring model, and predict the current engineering quality acceptance result, and automatically generate a quality detection report.

[0037] The prediction result includes a qualitative result of whether the quality is qualified or not and a quantitative result of comparison of key indicators. The predicted value of the key quality indicator is compared with the standard threshold value. For example, in the concrete pouring process, whether there is a crack risk is determined according to the predicted concrete strength value (for example, the actual predicted strength of C30 concrete is 28 MPa, which is lower than the design standard of 30 MPa).

[0038] The quality detection report includes basic information of the process, such as process name, construction time, construction team, supervision personnel, and data collection time; comparison and analysis of key quality indicators, that is, deviation of measured values / predicted values of physical parameters and process parameters in the construction process from corresponding standard data, if the deviation is too large, the process does not meet the specification; quality prediction result description, that is, the conclusion of quality qualification / non-qualification and the basis for obtaining the conclusion; potential risk prompt, for the predicted unqualified or fluctuating indicators, the possible quality hidden danger is analyzed; and the collected data on site, including pictures, videos, sensor data and equipment operation state records.

[0039] Further, the step S22 includes the following steps: S221: Collecting quality data of the process and extracting quality features, the quality features including numerical quality features and indirect quality features, and using correlation analysis to retain core features with high quality correlation; The numerical quality features are, for example, strength value, slump, and curing temperature of the concrete pouring, and the indirect quality features are, for example, process construction duration, equipment operation parameters (such as vibration frequency), and environmental factors (such as temperature and humidity during construction) which have potential influence on quality. If the extracted quality feature dimension is too high, principal component analysis (PCA) can be used for dimension reduction.

[0040] S222: Based on the quality feature differences of the highway engineering process, elbow method is used to determine the number of clusters, and K-Means algorithm is used for clustering; Specifically, K samples are randomly selected from the data set as initial cluster centers, the Euclidean distance between each sample and the K centers is calculated, the sample is assigned to the nearest cluster, the feature mean of all samples in each cluster is calculated and used as the new cluster center, the process of assigning samples to clusters and updating cluster centers is repeated until the cluster centers no longer change or the maximum iteration number is reached, and the stability of the clustering result is ensured.

[0041] S223: Statistical analysis is performed on the quality data of each cluster to extract common features and analyze common influencing factors; S224: The clustering process is iterated and the cluster features are updated.

[0042] Further, the step S223 includes the following steps: S2231: Align the quality data of all processes in the same cluster in time and space, and perform secondary screening on the data within the cluster to eliminate outliers; S2232: Calculate descriptive statistics for each quality indicator within the cluster, and based on the statistical results, extract common quality characteristics; Among them, the descriptive statistics include central tendency (mean, median), dispersion (variance / standard deviation, maximum / minimum) and frequency distribution.

[0043] S2233: Combine the influence factor data corresponding to the processes within the cluster to mine the common causes leading to common quality characteristics; Specifically, the influence factor data (including personnel, equipment, raw materials and environmental factors) is retrieved, and statistical tests and correlation analysis are used to locate the common influence factors within the cluster. The influence factors are analyzed for within-group differences, and if the values of a factor are highly concentrated within the cluster, it is considered a potential common factor. Through chi-square test (categorical variables) or Pearson correlation coefficient (continuous variables), the correlation strength between the influence factor and the common quality characteristic is calculated.

[0044] S2234: Verify the identified common characteristic-influence factor correlation; Combine the verification of the highway engineering construction specification to verify its rationality, exclude data-related but logically unrelated accidental correlations, and for key common influence factors, the authenticity can be confirmed through field tracing.

[0045] S2235: Extract cluster characteristic labels, sort them by influence strength, and generate a list of common influences.

[0046] For example, the influence factor list for the low-strength concrete cluster is: the main factors are A supplier 3-month batch cement strength not meeting standards (correlation degree 0.85), insufficient vibration time (<30 seconds, correlation degree 0.73); secondary factors are construction environment temperature > 38℃ (correlation degree 0.52).

[0047] Suppose that through K-Means, 20 concrete pouring processes of a highway project are clustered into 3 clusters, and cluster 3 contains 6 pouring processes, the analysis process is as follows: Cluster quality data statistics: 6 pouring processes with an average strength of 26MPa (design 30MPa), an average slump of 150mm (standard 180±20mm), an average curing temperature of 39℃, and 100% of the processes have slight cracks.

[0048] Common quality characteristics: low strength (average <28MPa), small slump, high curing temperature, and common slight cracks.

[0049] Influencing factor correlation: 6 times of pouring using B supplier 4 month batch cement (approach detection intensity deviation -8%), frequency of vibrating equipment <50Hz (standard 60-80Hz), construction environment temperature > 37℃. Through on-site tracing, B supplier 4 month batch cement does exist quality fluctuation, vibrating equipment frequency is insufficient due to aging, high temperature environment accelerates concrete water evaporation; finally determine the common influencing factors are cement batch quality substandard, vibrating equipment parameter insufficient, high temperature environment.

[0050] Through the above process, the quality commonness and causes of similar processes can be accurately located, providing data support for subsequent targeted rectification and cluster feature iteration update.

[0051] S3: According to the quality prediction result, if the predicted quality is qualified, fill in the relevant data in the process inspection application form, including physical parameters and process parameters in the construction process; if the predicted quality is unqualified, conduct risk assessment combined with the quality monitoring report, and score the risk assessment result, and trigger early warning based on the score; When the quality monitoring model predicts that the construction quality is qualified, the physical parameters (such as concrete pouring pressure, temperature, roadbed compaction degree, etc.) and process parameters (such as steel bar binding spacing, welding length, asphalt paving temperature, speed, etc.) in the construction process are automatically extracted and filled into the process inspection application form. Construction personnel check and confirm the data in the application form on the mobile APP, and if the check is correct, submit it to the supervision unit for audit.

[0052] After receiving the process inspection application, the supervising engineer first checks the data in the inspection application form in detail, compares the design file with the quality acceptance specification, checks whether the data is complete and accurate, and whether the construction process meets the requirements. At the same time, the supervising engineer can further understand the site construction situation by checking the photos, videos and other materials of the construction process. If the audit is passed, the supervising engineer signs the consent in the system, and the construction unit can proceed to the next process construction; if the data has problems or the construction process has defects, the supervising engineer notes the problem and returns the application form to require the construction unit to rectify.

[0053] When the quality monitoring model predicts that the construction quality is unqualified, the risk assessment is carried out from three dimensions of severity, influence range and occurrence probability, each dimension is quantified as a score, the weight of each dimension is determined by the analytic hierarchy process, the risk comprehensive score is calculated, and the formula is: M = Attention(S)×S+Attention(R)×R+Attention(P)×P; Wherein, Attention(S), Attention(R), Attention(P) represent the respective weights, M represents the risk score, S represents the severity, R represents the impact range, and P represents the occurrence probability.

[0054] The severity (S) is used to measure the influence level of the quality problem on the safety, function and durability of the engineering structure. A score of 1 represents a slight deviation (such as a slight deviation in surface flatness that does not affect the safety of the structure); a score of 3 represents a moderate defect (such as a 10-15% deviation in concrete strength that requires local repair); and a score of 5 represents a serious hidden danger (such as a concrete strength that is more than 20% lower than the design value, which may result in insufficient structural bearing capacity).

[0055] The impact range (R) is used to evaluate the engineering scope involved in the quality problem and its impact on subsequent processes. A score of 1 indicates that it only affects the current process locally (such as a roadbed compaction area of less than 10 square meters); a score of 3 indicates that it affects the current process as a whole and 1-2 subsequent processes (such as unqualified concrete pouring leading to delayed pile acceptance); and a score of 5 indicates that it affects multiple key processes or the entire project (such as a pile foundation quality problem leading to bridge substructure rework).

[0056] The occurrence probability (P) refers to the frequency of similar quality problems in historical data (similar projects, same process). A score of 1 indicates that it is rare (historical occurrence rate <5%); a score of 3 indicates that it is occasional (historical occurrence rate 10-30%); and a score of 5 indicates that it is highly likely (historical occurrence rate >50%, such as the probability of insufficient concrete strength due to improper curing in high temperature environments).

[0057] The attention mechanism is introduced to dynamically calculate the weights of severity, impact range and occurrence probability, including the following steps: Collect historical risk assessment data for quality non-compliance. Each data includes quantified values for severity, impact range, and occurrence probability, as well as corresponding risk consequence labels, and the features are preprocessed. Standardize the quantified values of severity, impact range, and occurrence probability to eliminate dimensional effects, and normalize the actual risk consequence labels as the target values for model training.

[0058] A lightweight model is constructed that includes a feature embedding layer, an attention calculation layer, and a weight output layer. The features are mapped to high-dimensional feature vectors to enhance the model's ability to capture the correlation between features. For example, a fully connected layer is used to map a 1-dimensional feature of each dimension to an 8-dimensional embedding vector: Embedding(S)=W1×S norm +b1; Embedding(R)=W2×Rnorm +b2; Embedding(P) = W3 x P norm +b3; wherein W1, W2, W3 are learnable weight matrices, S norm , R norm , P norm are feature vectors normalized in each dimension respectively, and b1, b2, b3 are bias terms.

[0059] The global features strongly related to the risk consequences are used as query vectors to calculate the attention scores, and the scores are normalized by the Softmax function to obtain the attention weights. The similarity between the embedding vector and the query vector, i.e. the attention score, is calculated, and the formula is: Score(S) = Q x Embedding(S) T ; Score(R) = Q x Embedding(R) T ; Score(P) = Q x Embedding(P) T ; The scores are normalized by the Softmax function to obtain the attention weights (the sum of the weights is 1), and the formula is: ; ; ; wherein, are the attention scores of each dimension respectively.

[0060] The attention weights are combined with the original features to predict the risk consequences through a fully connected layer, and the weight values are output.

[0061] wherein the model is trained based on historical data, and in each iteration of the model training process, the model calculates the predicted risk consequences based on the current attention weights, compares them with the actual consequences to adjust the parameters, so that the attention weights are inclined to the dimensions that have a greater impact on the risk consequences. The trained attention mechanism model is deployed in the risk assessment system to realize real-time dynamic weight distribution and set up a dynamic adjustment mechanism, for example, after accumulating 100 new cases, the model is retrained with the new data (incremental training) to update the attention weights, so that the weights adapt to the changes in the engineering scene. By determining the weights through the attention mechanism, the dynamic influence of different dimensions in actual risk assessment can be more accurately captured, especially in the scenario of complex quality problems and variable influencing factors in highway engineering, which improves the accuracy of risk scoring and the targeting of early warning.

[0062] According to the risk score, it is divided into three early warning thresholds, respectively triggering corresponding early warning measures: M≥4 is high risk and adopts red early warning, at this time, the project manager, chief supervising engineer and general manager of the construction unit are immediately sent early warning through SMS, APP push and email, the content needs to include quality problem type, specific location, risk level and result and emergency treatment suggestion.

[0063] 2≤M<4 is medium risk and adopts yellow early warning, at this time, the site supervising engineer and construction team leader are sent early warning, the content includes problem details and rectification time limit, and real-time uploading of rectification progress is required.

[0064] M<2 is low risk and adopts blue early warning, at this time, the construction team is sent a prompt, requiring to record the problem and supplement the rectification explanation at the next inspection.

[0065] S4: Based on correlation analysis of unqualified influencing factors and positioning, according to the specific factors positioned, rectification and prevention are carried out.

[0066] Among them, S4 includes the following steps: S41: Clear the potential influencing factor dimension, and refine the specific index, associate the specific performance of quality unqualified with the influencing factor data, and form an analysis data set; For example, the strength detection value of each batch of concrete is matched with the detection data of corresponding batch of cement, vibration equipment parameters, construction personnel information, etc. to ensure the accuracy of data correspondence. Specifically, the influencing factors include construction personnel factors (operation skill, training times, attendance records, number of illegal operations, etc.), equipment operation factors (equipment model, operation parameters, fault records, maintenance period, etc.), raw material quality factors (material batch, incoming detection data, storage conditions, etc.), environmental factors (construction temperature and humidity, rainfall, wind power, illumination time, etc.).

[0067] S42: Taking the quantitative index of quality unqualified as the dependent variable and each factor index as the independent variable, a multiple linear regression model is constructed, and the influence strength is judged by the absolute value of the regression coefficient; S43: Based on SHAP technology, the correlation of each factor is calculated and sorted; For example, after analyzing "unqualified concrete strength", the correlation degree ranking is obtained: cement moisture content (0.85) > vibration time (0.72) > environmental temperature (0.58) > operator qualification (0.41), which shows that cement moisture content and vibration time are the core influencing factors.

[0068] S44: Based on the correlation analysis results, combined with the actual investigation on site, the root cause of the problem is accurately positioned; For the top-ranking factors, verification is carried out through on-site inspection, data tracing, etc. For example, if the water content of cement has the highest correlation degree, the incoming record of the batch of cement, storage environment (such as whether it has been raining), and the water content of the sample are retested to confirm whether the water content exceeds the standard due to improper storage. Remove the "pseudo-correlation" factors (such as data analysis shows that "wind" is related to quality problems, but on-site inspection shows that wind does not affect construction, so this factor is excluded), and finally lock the specific cause. For example, it is confirmed that the specific factor of "unqualified concrete strength" is that the water content of a batch of cement exceeds the standard (raining during storage) and the vibration equipment failure causes insufficient vibration time.

[0069] S45: According to the specific factors of positioning, formulate an executable and traceable rectification scheme, and clearly define the rectification elements; If the quality of raw materials (such as the water content of cement exceeding the standard) is the cause, the batch of materials is immediately stopped and replaced with qualified suppliers, and the suppliers are required to provide material factory inspection reports. For the engineering parts that have used the materials, strength reinspection is carried out and reinforcement schemes are developed. If the equipment problem (such as vibration equipment failure) is the cause, the equipment is immediately repaired or replaced, the equipment parameters are calibrated (such as setting the minimum threshold for vibration time), and a "daily inspection - maintenance - calibration" account book is established. If the personnel operation (such as insufficient vibration time) is the cause, special training is carried out for the operators (only those who pass the examination can work), the vibration process standards are clearly defined (such as 30 seconds of vibration per place), and the supervision of the site is arranged by the supervisors. The scheme must include the rectification person in charge (such as the construction team leader, material staff), the completion time limit (such as replacing the cement within 48 hours), the acceptance standard (such as the water content of cement ≤0.5%), and the requirement for supporting materials (such as reinspection report, photo after rectification).

[0070] S46: The construction unit executes the rectification scheme and uploads the progress in real time.

[0071] The supervision unit regularly checks the site (such as randomly checking the vibration time and rechecking the cement test report), and signs the supervision opinion. After the rectification is completed, the construction unit self-inspects and passes the examination and submits the reinspection application (with the data after rectification, such as the concrete strength reinspection report); the supervision unit organizes the reinspection, and if it meets the standard (such as the strength meets the design value), it is confirmed that the acceptance is passed, and the quality problem process is closed; if it does not meet the standard, the cause is reanalyzed, the rectification measures are adjusted, and it is not until it is qualified. Through the above steps, the whole process management from data correlation analysis to on-site positioning and closed-loop rectification is realized, ensuring that the root cause of the quality problem is completely solved, and at the same time, the prevention and control experience is formed to prevent the same problem from recurring.

[0072] The method further comprises: S5, based on BIM technology, a visual platform is constructed to visually display the overall situation and key indicators of construction in real time, and provide data interaction and sharing functions between different user roles.

[0073] On the visualization platform, the user can view the engineering progress, process inspection state, quality detection data, safety management information, etc. in real time. Through the visualization chart and graphical display mode, the overall situation and various key indicators of the project are intuitively presented. The construction unit can submit the process inspection application and related data to the supervision unit, and the supervision unit will feed back the inspection result to the construction unit after the audit; the project management personnel can view the work progress and data of each participant in real time, and carry out unified coordination and management. Through data interaction and sharing, the information island is broken, and the project collaborative management efficiency is improved.

[0074] The above embodiments only describe the preferred embodiments of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements of the technical solutions of the present application made by ordinary engineering technicians in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A highway engineering site management method based on process reporting, characterized in that: The following steps are involved: S1: Obtain highway construction site data, standards, and technical specifications through multiple channels, and pre-process the collected data. The construction site data includes construction data, process flow data, equipment data, and personnel data; S2: Analyze the collected data using big data analysis technology, build a quality monitoring model to monitor the quality of the construction process, compare it with standard data, predict whether the quality is qualified, and generate a quality monitoring report; S3: Execute corresponding strategies based on the quality prediction results. If the predicted quality is qualified, relevant data, including physical parameters and process parameters during the construction process, are entered into the process inspection application form. If the predicted quality is unqualified, a risk assessment is conducted in conjunction with the quality monitoring report, and the risk assessment results are scored, with an alert triggered based on the score. S4: Analyze and locate the factors affecting non-compliance based on correlation, and carry out rectification and prevention and control according to the specific factors located.

2. A highway engineering site management method based on process reporting according to claim 1, characterized in that: Build a data collaboration platform based on cloud computing, receive collected data through the network, use relational databases and non-relational databases to store data, establish associations between data, and perform cross-regional synchronization.

3. A highway engineering site management method based on process reporting according to claim 1, characterized in that: Among them, S2 includes the following steps: S21: Analyze the potential associations between highway construction site data based on association rule mining algorithm; S22: Use clustering algorithms to classify the quality data during the construction process and group processes with similar quality characteristics into one category; S23: Using highway construction site data and environmental data as input and project quality acceptance results as output labels, a machine learning model is constructed and trained using historical data. S24: Build a comprehensive quality monitoring indicator system based on highway project quality acceptance standards and design requirements; S25: Combine big data analysis technology with the quality monitoring indicator system to build a quality monitoring model, and incorporate the association rules mined in S21 and the cluster features clustered in S22; S26: Input the current highway construction site data and environmental data into the quality monitoring model, predict the current project quality acceptance result, and automatically generate a quality inspection report.

4. A highway engineering site management method based on process reporting according to claim 3, characterized in that: in, Step S22 includes the following steps: S221: Collecting process quality data and extracting quality features, including numerical quality features and indirect quality features, and using correlation analysis to retain core features with a high degree of correlation with quality; S222: Combined with the quality characteristics differences of highway engineering processes, the elbow method is used to determine the number of clusters, and clustering is performed based on the K-Means algorithm; S223: Perform statistical analysis on the quality data of each cluster, extract common features, and analyze common influencing factors; S224: The clustering process is continuously iterated, and cluster features are updated.

5. A highway engineering site management method based on process reporting according to claim 4, characterized in that: Step S223 includes the following steps: S2231: Align the quality data of all processes in the same cluster in time and space, and perform secondary screening on the data within the cluster to eliminate outliers; S2232: Calculate descriptive statistics for each quality indicator within the cluster and extract common quality features based on the statistical results; S2233: Combine the influencing factor data corresponding to the processes within the cluster to explore the common causes leading to common quality characteristics; S2234: Verify the identified common characteristics-influencing factors relationship; S2235: Extract cluster feature labels, sort them by impact intensity, and generate a list of common impacts.

6. A highway engineering site management method based on process reporting according to claim 3, characterized in that: The quality monitoring indicator system constructed by S24 is used as the judgment standard of the model, which specifically includes indicator thresholds and indicator weights. The model needs to compare the results of machine learning predictions with the thresholds of the indicator system, and calculate the comprehensive quality score based on the indicator weights. The calculation formula is: Z = ∑α × G; Among them, Z represents the comprehensive quality score, α represents the weight, and G represents the indicator compliance rate.

7. A highway engineering site management method based on process reporting according to claim 1, characterized in that: When the quality monitoring model predicts that the construction quality is unqualified, the quality inspection report is combined with the three dimensions of severity, scope of impact, and probability of occurrence to conduct a risk assessment. Each dimension is quantified into a score, and the hierarchical analysis method is used to determine the weight of each dimension to calculate the comprehensive risk score. The formula is: M = Attention(S)×S+Attention(R)×R+Attention(P)×P; Among them, Attention(S), Attention(R), and Attention(P) represent the weights of each item respectively, M represents the risk score, S represents the severity, R represents the scope of impact, and P represents the probability of occurrence.

8. A highway engineering site management method based on process reporting according to claim 7, characterized in that: The attention mechanism is introduced to dynamically calculate the weights of severity, impact range, and probability of occurrence, including the following steps: Collect historical risk assessment data on quality failures. Each piece of data contains quantitative values ​​for severity, impact scope, and probability of occurrence, along with corresponding risk consequence labels. Feature preprocessing is also performed. Build a lightweight model consisting of a feature embedding layer, an attention calculation layer, and a weight output layer, mapping features into high-dimensional feature vectors to enhance the model's ability to capture the correlation between features; The global features that are strongly correlated with risk consequences are used as query vectors to calculate attention scores, which are normalized by the Softmax function to obtain attention weights. The attention weights are combined with the original features, the risk consequences are predicted through the fully connected layer, and the weight values ​​are output.

9. A highway engineering site management method based on process reporting according to claim 1, characterized in that: Among them, S4 includes the following steps: S41: Clarify the dimensions of potential influencing factors and refine specific indicators, associate the specific manifestations of quality failure with the influencing factor data, and form an analysis data set; S42: Using the quantitative indicators of unqualified quality as dependent variables and the indicators of various factors as independent variables, a multivariate linear regression model was constructed, and the impact intensity was judged by the absolute value of the regression coefficient; S43: Calculate the correlation of each factor and rank them based on SHAP technology; S44: Based on the correlation analysis results and combined with the actual on-site investigation, the root cause of the problem is accurately located; S45: Develop an executable and traceable rectification plan based on the specific factors identified, and clearly define the rectification elements; S46: The construction unit implements the rectification plan and uploads the progress in real time.

10. A highway engineering site management method based on process reporting according to claim 1, characterized in that: The method further comprises: S5. Build a visualization platform based on BIM technology to display the overall construction situation and key indicators in real time in a visual way, and provide data interaction and sharing functions between different user roles.

Images