Engineering data processing method and system based on intelligent digital supervision

Abstract

The invention provides an engineering data processing method and system based on intelligent digital supervision. The method comprises the following steps: collecting engineering field data of a target construction site through a sensor network deployed at the target construction site, and preprocessing engineering environment data and engineering equipment data; extracting environment data features and equipment data features; performing feature fusion on the environment data features and the equipment data features to obtain project supervision data features of the target construction site; mapping the project supervision data features to a preset high-dimensional feature space in a unified manner, and performing feature distribution analysis in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result; and outputting an engineering abnormal result of the target construction site by adopting a support vector machine based on a multi-dimensional distribution feature analysis result. The method has the effect of high recognition precision of recognizing the abnormal condition of the engineering field.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and specifically relates to an engineering data processing method and system based on intelligent digital supervision. Background Art

[0002] With the rapid development of information technology and the Internet of Things, traditional engineering supervision methods have begun to evolve towards digitalization and intelligence. In this context, some intelligent engineering supervision data processing methods have been developed and applied. These methods usually include using various sensor networks to collect engineering site data, using data management systems for storage and preliminary analysis, and displaying the results through a visual interface. However, the scale of engineering projects is constantly expanding, the types are becoming increasingly diverse, and the requirements for supervision are also increasing. Especially in the fields of large-scale infrastructure projects and complex industrial facility construction, traditional intelligent supervision methods have begun to show limitations. Although traditional intelligent engineering supervision data processing methods have introduced some automation and digital technologies, traditional intelligent supervision systems often focus on a single type of data and find it difficult to fully explore the complex relationships between different types of data. This makes it impossible to fully grasp the complex conditions of the engineering site, and ultimately leads to a significant reduction in the accuracy of identifying abnormal conditions on the engineering site. Summary of the invention

[0003] The present invention provides an engineering data processing method and system based on intelligent digital supervision to solve the problem of low recognition accuracy of abnormal conditions in engineering sites.

[0004] In a first aspect, the present invention provides a method for processing engineering data based on intelligent digital supervision, the method comprising the following steps:

[0005] Collecting engineering site data of the target construction site through a sensor network deployed at the target construction site, wherein the engineering site data includes engineering environment data and engineering equipment data;

[0006] Preprocessing the engineering environment data and the engineering equipment data;

[0007] Extracting environment data features and equipment data features from the preprocessed engineering environment data and engineering equipment data respectively;

[0008] Fusing the environmental data features with the equipment data features to obtain the engineering supervision data features of the target construction site;

[0009] The engineering supervision data features are uniformly mapped to a preset high-dimensional feature space, and feature distribution analysis is performed in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result;

[0010] Based on the multi-dimensional distribution feature analysis results and using a support vector machine, the engineering abnormality results of the target construction site are output.

[0011] Optionally, the engineering environment data and the engineering equipment data are both time series data, and the preprocessing of the engineering environment data and the engineering equipment data comprises the following steps:

[0012] Screening out all abnormal data values ​​in the engineering environment data and the engineering equipment data;

[0013] Performing time alignment processing on all the engineering environment data and all the engineering equipment data after outliers are screened out;

[0014] Selecting the engineering environment data with the highest sampling frequency as the first benchmark data, and using the first benchmark data as a standard to perform linear interpolation processing on all other engineering environment data;

[0015] The engineering equipment data with the highest sampling frequency is selected as the second benchmark data, and the second benchmark data is used as a standard to perform cubic spline interpolation processing on all other engineering equipment data.

[0016] Optionally, extracting environment data features and equipment data features from the pre-processed engineering environment data and engineering equipment data respectively comprises the following steps:

[0017] Calculating and obtaining the autocorrelation coefficient matrix of the preprocessed engineering environment data;

[0018] Extracting frequency domain features of environmental data of the engineering environmental data by discrete Fourier transform;

[0019] Integrating the autocorrelation coefficient matrix and the frequency domain features of the environmental data into environmental data features of the engineering environmental data;

[0020] Calculating equipment data statistical features of the engineering equipment data based on preset target statistical parameters;

[0021] Using a dynamic time warping algorithm to calculate the distance between the engineering equipment data and preset typical equipment data, and obtain equipment data pattern characteristics of the engineering equipment data;

[0022] The equipment data statistical features and the equipment data pattern features are integrated into the equipment data features of the engineering equipment data.

[0023] Optionally, the step of fusing the environmental data features with the equipment data features to obtain the engineering supervision data features of the target construction site comprises the following steps:

[0024] Taking the environment data features and the device data features as key model concepts;

[0025] Using the key model concepts as the basis of the ontology model, defining basic classes and constructing an ontology model hierarchy;

[0026] Establishing a basic class relationship between the basic classes according to the semantic association between the basic classes;

[0027] Combining the ontology model hierarchy and the basic class relationship to construct a project supervision data model;

[0028] The engineering supervision data features of the target construction site are extracted from the engineering supervision data model using feature engineering.

[0029] Optionally, the step of uniformly mapping the engineering supervision data features to a preset high-dimensional feature space, and performing feature distribution analysis in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result comprises the following steps:

[0030] Based on the radial basis function kernel, the engineering supervision data features are uniformly mapped to a preset high-dimensional feature space;

[0031] Analyzing the high-dimensional data distribution characteristics from the high-dimensional feature space by principal component analysis;

[0032] Performing cluster analysis on the high-dimensional data in the high-dimensional feature space, and using the cluster analysis results as cluster features of the high-dimensional data in the high-dimensional feature space;

[0033] The high-dimensional data distribution characteristics and the high-dimensional data clustering characteristics are combined as a multi-dimensional distribution characteristic analysis result.

[0034] In a second aspect, the present invention further provides an engineering data processing system based on intelligent digital supervision, the system comprising:

[0035] A data acquisition module, used to collect engineering site data of the target construction site through a sensor network deployed at the target construction site, wherein the engineering site data includes engineering environment data and engineering equipment data;

[0036] A data preprocessing module, used for preprocessing the engineering environment data and the engineering equipment data;

[0037] A feature extraction module, used to extract environment data features and equipment data features from the preprocessed engineering environment data and engineering equipment data respectively;

[0038] A feature fusion module, used to fuse the environmental data features with the equipment data features to obtain the engineering supervision data features of the target construction site;

[0039] A feature analysis module is used to uniformly map the features of the engineering supervision data to a preset high-dimensional feature space, and perform feature distribution analysis in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result;

[0040] The anomaly identification module is used to output the engineering anomaly results of the target construction site based on the multi-dimensional distribution feature analysis results and using a support vector machine.

[0041] Optionally, the engineering environment data and the engineering equipment data are both time series data, and the data preprocessing module includes:

[0042] A data screening unit, used to screen out all abnormal data values ​​in the engineering environment data and the engineering equipment data;

[0043] A data alignment unit, used for performing time alignment processing on all the engineering environment data and all the engineering equipment data after outliers are screened out;

[0044] A first data interpolation unit is used to select the engineering environment data with the highest sampling frequency as the first reference data, and perform linear interpolation processing on all other engineering environment data based on the first reference data;

[0045] The second data interpolation unit is used to select the engineering equipment data with the highest sampling frequency as the second reference data, and perform cubic spline interpolation processing on all other engineering equipment data based on the second reference data.

[0046] Optionally, the feature extraction module includes:

[0047] A time domain feature extraction unit, used for calculating the autocorrelation coefficient matrix of the preprocessed engineering environment data;

[0048] A frequency domain feature extraction unit, used for extracting frequency domain features of environmental data of the engineering environment data by discrete Fourier transform;

[0049] A first feature integration unit, used to integrate the autocorrelation coefficient matrix and the frequency domain features of the environmental data into environmental data features of the engineering environmental data;

[0050] A statistical feature extraction unit, used to calculate the equipment data statistical features of the engineering equipment data based on preset target statistical parameters;

[0051] A pattern feature extraction unit, used to calculate the distance between the engineering equipment data and preset typical equipment data using a dynamic time warping algorithm to obtain equipment data pattern features of the engineering equipment data;

[0052] The second feature integration unit is used to integrate the equipment data statistical features and the equipment data pattern features into the equipment data features of the engineering equipment data.

[0053] Optionally, the feature fusion module includes:

[0054] A model concept selection unit, used for taking the environmental data features and the device data features as key model concepts;

[0055] A model structure construction unit, used to define basic classes based on the key model concepts as the basis of the ontology model and to construct a hierarchical structure of the ontology model;

[0056] A model relationship building unit, used for establishing a basic class relationship between the basic classes according to the semantic association between the basic classes;

[0057] An ontology model construction unit, used to construct a project supervision data model by combining the ontology model hierarchy and the basic class relationship;

[0058] A feature selection unit is used to extract the engineering supervision data features of the target construction site from the engineering supervision data model by using feature engineering.

[0059] Optionally, the feature analysis module includes:

[0060] A feature mapping unit, used for uniformly mapping the engineering supervision data features into a preset high-dimensional feature space based on a radial basis function kernel;

[0061] A distribution feature analysis unit, used to analyze the high-dimensional data distribution features from the high-dimensional feature space by principal component analysis;

[0062] A clustering feature analysis unit, used to perform clustering analysis on the high-dimensional data in the high-dimensional feature space, and use the clustering analysis result as the clustering feature of the high-dimensional data in the high-dimensional feature space;

[0063] The feature analysis result output unit is used to combine the high-dimensional data distribution feature and the high-dimensional data clustering feature as a multi-dimensional distribution feature analysis result.

[0064] The beneficial effects of the present invention are:

[0065] The present invention realizes the all-round and real-time collection of engineering environment and equipment data by deploying a comprehensive sensor network, greatly improving the integrity and timeliness of the data. By adopting advanced data preprocessing and feature extraction technology, the most representative and predictive information can be extracted from massive raw data, greatly improving the efficiency and accuracy of data analysis. In addition, the present invention innovatively introduces a feature fusion mechanism, organically combines environmental data features and equipment data features, and realizes the deep integration of multi-source heterogeneous data, so as to capture complex correlation patterns that are difficult to find by traditional methods. More importantly, the present invention maps the features of engineering supervision data to a high-dimensional feature space and performs distribution analysis. This method greatly enhances the system's ability to identify complex patterns and abnormal situations. Finally, based on the results of multi-dimensional distribution feature analysis and combined with the support vector machine algorithm, the present invention realizes high-precision engineering anomaly detection, which can not only detect potential risks in a timely manner, but also provide the type, severity and possible causes of anomalies, providing comprehensive and accurate decision support for engineering managers. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of an engineering data processing method based on intelligent digital supervision in one of the implementation modes of the present application.

[0067] Figure 2 This is a structural diagram of an engineering data processing system based on intelligent digital supervision in one of the implementation modes of the present application. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0069] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0070] Figure 1FIG. 1 is a flow chart of an engineering data processing method based on intelligent digital supervision in one embodiment. It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps in the above method may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps. Figure 1 As shown, the engineering data processing method based on intelligent digital supervision disclosed in the present invention specifically includes the following steps:

[0071] S101. Collect engineering site data of the target construction site through a sensor network deployed at the target construction site.

[0072] Among them, first of all, the engineering site data is collected by deploying a sensor network at the target construction site. This sensor network is a complex system composed of various types of sensors, including but not limited to temperature sensors, humidity sensors, noise sensors, dust sensors, vibration sensors, displacement sensors, etc. These sensors are arranged at various key locations on the construction site to ensure that the engineering environment data and engineering equipment data can be collected comprehensively and accurately. Engineering environment data usually includes temperature, humidity, noise level, air quality indicators (such as PM2.5 concentration), wind speed, light intensity, etc. These data reflect the overall environmental conditions of the construction site and are crucial to protecting the health of workers, ensuring construction quality, and complying with environmental protection regulations. For example, temperature and humidity sensors can monitor the temperature and humidity changes of the construction site in real time. When the temperature is too high or the humidity is abnormal, the system can promptly remind managers to take corresponding measures, such as increasing ventilation or adjusting the working time.

[0073] Engineering equipment data includes the operating parameters of various types of construction equipment, such as the load capacity and working angle of tower cranes, the digging depth and speed of excavators, and the pumping pressure and flow of concrete pump trucks. These data directly reflect the working status and efficiency of the equipment, and are of great significance for optimizing construction progress, preventing equipment failures, and ensuring construction safety. For example, by monitoring the load capacity and working angle of tower cranes, it is possible to determine in real time whether there is overloading or unsafe operation, thereby preventing possible safety accidents in a timely manner.

[0074] The process of data collection in sensor networks is continuous, and high-frequency sampling is usually used to ensure that instantaneous changes in the environment and equipment status can be captured. The frequency of data collection may vary depending on the type of data. For example, temperature and humidity may be collected every 5 minutes, while the operating parameters of the equipment may need to be collected multiple times per second. These raw data are transmitted to the central data processing unit via a wired or wireless network to provide a basis for subsequent data processing and analysis. The quality of the collected data directly affects the accuracy of subsequent analysis, so special attention should be paid to the selection, installation and maintenance of sensors. Sensors with high precision, strong anti-interference ability and good stability are preferred. At the same time, sensors need to be calibrated regularly to ensure long-term data accuracy. In addition, the dustproof, waterproof, and earthquake-resistant characteristics of the sensor need to be considered to adapt to the harsh environment of the construction site.

[0075] S102. Preprocess engineering environment data and engineering equipment data.

[0076] Among them, in the second step of engineering data processing, it is crucial to preprocess the collected engineering environment data and engineering equipment data. The main purpose of this step is to clean and standardize the raw data to lay the foundation for subsequent feature extraction and analysis. The preprocessing process includes several key sub-steps, each of which has its specific purpose and method. First, all abnormal data values ​​in the engineering environment data and engineering equipment data need to be screened out. Abnormal data may be caused by sensor failure, communication errors, or extreme environmental conditions. If these data are not processed, they will seriously affect the accuracy of subsequent analysis. The detection of outliers can be achieved through a variety of statistical methods, such as the 3sigma rule, the box plot method, or anomaly detection algorithms based on machine learning. For example, when using the 3sigma rule, the mean μ and standard deviation σ of the data can be calculated, and data outside the range of (μ-3σ, μ+3σ) can be regarded as outliers. Different outlier detection methods may be required for different types of data. Next, all engineering environment data and engineering equipment data after outliers are screened out are time-aligned. The purpose of this step is to ensure that data from different sources are consistent in the time dimension, which is convenient for subsequent comprehensive analysis. Time alignment can be achieved through methods such as interpolation or resampling.

[0077] By removing outliers, the reliability and representativeness of the data are greatly improved. Secondly, time alignment processing ensures the consistency of data from different sources in the time dimension, creating conditions for subsequent multi-dimensional analysis. Furthermore, different interpolation methods (linear interpolation and cubic spline interpolation) are used for different types of data, which not only ensures the continuity of the data, but also retains the characteristics and trends of the original data as much as possible. The preprocessed data set will have higher quality and consistency, which is crucial for subsequent feature extraction and analysis. For example, the preprocessed temperature and humidity data can more accurately reflect the environmental change trend of the construction site, and the preprocessed equipment operating parameters can more accurately describe the working status of the equipment. These high-quality data provide a reliable foundation for subsequent anomaly detection, efficiency analysis and predictive modeling.

[0078] S103. Extracting environmental data features and equipment data features from the preprocessed engineering environment data and engineering equipment data respectively.

[0079] Among them, by integrating the statistical features of equipment data and the pattern features of equipment data, the complete equipment data features of engineering equipment data are obtained. These features not only contain the statistical characteristics of equipment operation, but also reflect the similarity between the equipment operation mode and the typical mode, providing rich information for subsequent equipment status evaluation and anomaly detection. The implementation effect of this feature extraction method is multifaceted. First, it greatly reduces the dimension of the data and converts high-dimensional raw data into a more compact and meaningful feature representation. This not only improves the efficiency of subsequent processing, but also helps to reduce the risk of overfitting. Secondly, the extracted features are highly interpretable. For example, the frequency domain features of environmental data can intuitively reflect the periodicity of environmental changes, and the DTW distance of equipment data can reflect the degree of deviation of the current operation mode from the normal mode. Furthermore, these features provide high-quality input for subsequent model training and decision making, which helps to improve the performance and reliability of the entire system.

[0080] S104. Fusing the environmental data features and the equipment data features to obtain the engineering supervision data features of the target construction site.

[0081] Among them, through this step, the complex engineering supervision data model can be converted into a set of structured feature vectors, which not only contain the information of the original data, but also integrate the complex relationship between domain knowledge and features. The implementation effect of this feature fusion method is significant. First, the organic combination of environmental data features and equipment data features is realized, creating a unified feature space, so that subsequent analysis can consider the influence of environmental factors and equipment factors at the same time. Secondly, the ontology model-based method introduces rich semantic information, so that the features are no longer isolated values, but knowledge units with clear meanings and associations. This greatly improves the interpretability and usability of the features. Furthermore, by constructing a knowledge graph, the complex interactions between features can be better captured, which is crucial for understanding the overall situation of the engineering site and predicting potential problems.

[0082] S105. The engineering supervision data features are uniformly mapped to a preset high-dimensional feature space, and feature distribution analysis is performed in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result.

[0083] Among them, RBF kernel mapping can capture nonlinear features and relationships in the data, which are difficult to find in the original feature space. PCA analysis helps identify the most important feature combinations, which not only reduces the dimension of the data but also retains most of the information. Cluster analysis reveals natural groupings in the data, which correspond to different engineering states or operating modes. The data distribution characteristics in high-dimensional space provide an important basis for subsequent anomaly detection. For example, points far away from the cluster center are potential anomalies. Although the original data has a high dimension, PCA can project the data into two-dimensional or three-dimensional space for visualization, which helps to intuitively understand the structure and distribution of the data.

[0084] S106. Based on the multi-dimensional distribution feature analysis results and using a support vector machine to output the engineering abnormality results of the target construction site.

[0085] Among them, support vector machine is a powerful machine learning algorithm, which is particularly suitable for processing high-dimensional data and complex classification problems. In this step, SVM is used to build an anomaly detection model. The basic principle of SVM is to find an optimal hyperplane in the feature space so that the distance between sample points of different categories and this hyperplane is maximized. In the scenario of anomaly detection, one-class SVM (One-Class SVM) is usually used. Its goal is to find a hypersphere that surrounds most normal samples and excludes a few abnormal samples. The objective function of one-class SVM can be expressed as:

[0086] min(1 / 2)||w|| 2 +(1 / vn)∑ξi-ρ

[0087] st(w·Φ(xi))≥ρ-ξi,ξi≥0

[0088] Among them, w is the normal vector of the hyperplane, Φ(xi) is the mapping of sample xi in the feature space, ρ is the offset of the hyperplane, ξi is the slack variable, ν is a parameter that controls the proportion of abnormal samples, and n is the number of samples.

[0089] In practical applications, kernel techniques are often used to implicitly process high-dimensional feature spaces. Based on the RBF kernel used in the previous step, the kernel function can be directly calculated in the original feature space without explicitly calculating the coordinates in the high-dimensional space. The training process of the SVM model is as follows:

[0090] 1. Data preparation: Use the multidimensional distribution feature analysis results obtained in the previous step as input data. These features may include principal components obtained by PCA and cluster features obtained by cluster analysis.

[0091] 2. Parameter selection: Select the optimal kernel function parameter γ and anomaly ratio parameter ν through cross-validation or grid search. For example, you can try to set the value range of γ to [0.1, 1, 10] and the value range of ν to [0.01, 0.05, 0.1].

[0092] 3. Model training: Train a class of SVM models using the selected parameters and the prepared training data. The training process will find the optimal hyperplane parameters w and ρ .

[0093] 4. Model validation: Use a portion of data that was not used in training to validate the performance of the model, including indicators such as accuracy, recall, and F1 score.

[0094] 5. Model tuning: Based on the verification results, you may need to return to step 2 to adjust the parameters, or try other kernel functions (such as polynomial kernel) to achieve better performance.

[0095] After training, the SVM model can be used to detect whether new engineering data is abnormal. For each new data point x, calculate its decision function value:

[0096] f(x)=(w·Φ(x))-ρ

[0097] If f(x) ≥ 0, the data point is considered normal; if f(x) < 0, the data point is considered abnormal.

[0098] In practical applications, it may be necessary to further refine the type and degree of anomalies. For example, multiple thresholds can be defined based on the value of f(x) to classify anomalies into mild anomalies, moderate anomalies, and severe anomalies. This refinement can help engineering managers better understand the nature and urgency of anomalies.

[0099] The output of the SVM model can include the following:

[0100] 1. Abnormal flag: indicates whether the current project status is abnormal.

[0101] 2. Abnormality level: If an abnormality is detected, the severity of the abnormality is indicated.

[0102] 3. Anomaly probability: the confidence of the model in its judgment of anomalies.

[0103] 4. Contributing features: A list of features that contribute most to the anomaly judgment, which helps to understand the cause of the anomaly.

[0104] In order to improve the interpretability of the model, techniques such as SHAP (SHapley Additive exPlanations) can be combined to explain the decisions of SVM. SHAP values ​​can help understand the contribution of each feature to the final decision, thereby providing engineering managers with more insightful anomaly analysis reports.

[0105] SVM can effectively process high-dimensional nonlinear data and can capture complex abnormal patterns, thereby improving the accuracy of anomaly detection. By adjusting the kernel function and parameters, SVM can adapt to different types of engineering data and abnormal patterns. Once the model training is completed, anomaly detection of new data can be carried out quickly to meet the needs of real-time monitoring. Although SVM itself is a "black box" model, by combining technologies such as SHAP, it can provide an explanation of the cause of anomalies, which is crucial for engineering management decisions. By adjusting the anomaly ratio parameter ν, a balance can be found between detection sensitivity and false alarm rate to meet the needs of different scenarios. The SVM model can not only detect anomalies that have already occurred, but also warn of potential abnormal risks by monitoring the changing trend of the decision function value.

[0106] In general, this step converts the previous complex data processing and analysis results into specific engineering anomaly detection results. Through the SVM model, abnormal conditions in the project can be identified in a timely and accurate manner, providing strong decision-making support for engineering managers. This method not only improves the efficiency and accuracy of engineering supervision, but also provides a data basis for preventive maintenance and risk management, and ultimately achieves the core goal of intelligent digital supervision - improving engineering quality, reducing safety risks, and optimizing resource utilization through data-driven methods.

[0107] In one embodiment, the engineering environment data and the engineering equipment data are both time series data, and preprocessing the engineering environment data and the engineering equipment data includes the following steps:

[0108] Screen out all abnormal data values ​​in engineering environment data and engineering equipment data;

[0109] Perform time alignment on all engineering environment data and all engineering equipment data after outliers are removed;

[0110] Select the engineering environment data with the highest sampling frequency as the first benchmark data, and use the first benchmark data as the standard to perform linear interpolation processing on all other engineering environment data;

[0111] The engineering equipment data with the highest sampling frequency is selected as the second benchmark data, and the second benchmark data is used as the standard to perform cubic spline interpolation processing on all other engineering equipment data.

[0112] In this embodiment, a statistical method is used to preliminarily identify outliers. Commonly used methods include the Z-score method, the interquartile range (IQR) method, and the modified Z-score method. Taking the Z-score method as an example, the Z score of each data point is calculated: Z = (X-μ) / σ, where X is the value of the data point, μ is the mean value of the data set, and σ is the standard deviation. Generally, data points with |Z|>3 are regarded as potential outliers. The implementation of outlier screening improves the quality and reliability of the data and lays a solid foundation for subsequent analysis. Secondly, it helps to identify potential equipment failures or abnormal environmental conditions, providing clues for timely maintenance and intervention. By removing outliers, the statistical characteristics of the data become more stable, which is conducive to more accurately capturing the true trends and patterns of the data. Outlier screening can also improve the efficiency of subsequent data processing and model training, because outliers often cause the algorithm to converge more slowly or fall into a local optimum.

[0113] Next, the first step in the time alignment process is to select a suitable time base. Usually, the data set with the highest sampling frequency is selected as the base, or a time interval that is divisible by all sampling frequencies is selected as a common base. For example, if there are three data sources with sampling intervals of 1 second, 2 seconds, and 5 seconds, 1 second can be selected as the base time interval. Then the timestamp issue needs to be dealt with. Different devices may use different time representation methods (such as Unix timestamp, ISO8601 format, etc.), and they need to be converted to the same format. In addition, the time zone issue needs to be considered to ensure that all data uses the same time zone. For example, all timestamps can be converted to UTC time to avoid the influence of factors such as daylight saving time. The core step of time alignment is to adjust all data to the selected time point.

[0114] For data with a sampling frequency higher than the reference frequency, downsampling is required. Common methods include the average method, the median method, the maximum method, and so on. Choosing which method also needs to consider the characteristics of the data. For example, for noisy data, the median method may be more able to retain the true signal; for peak-sensitive data, the maximum method may be more appropriate. The problem of missing data also needs to be dealt with during the time alignment process. For short-term data missing, it can be filled by interpolation methods; for long-term data missing, it may be necessary to mark it as missing values ​​or use more complex filling methods, such as pattern filling based on historical data or prediction filling based on machine learning methods.

[0115] First, it is necessary to determine which environmental parameter has the highest sampling frequency. This is usually determined by comparing the data collection intervals of different sensors. For example, suppose the temperature sensor collects data every 1 minute, the humidity sensor collects data every 5 minutes, and the light sensor collects data every 10 minutes. In this case, the temperature data will be selected as the first benchmark data because it has the highest sampling frequency (once per minute). After the benchmark data is selected, a unified timeline needs to be created. This timeline should cover the time range of all environmental data and use the sampling interval of the benchmark data. Continuing with the above example, if the entire data set covers 24 hours, the unified timeline will contain 1440 time points (24 hours * 60 minutes / hour). Next, for environmental parameters with a lower sampling frequency than the benchmark data, linear interpolation processing needs to be performed. The basic principle of linear interpolation is to assume that the values ​​between two known data points change linearly. For the interpolation at time t, the formula is:

[0116] y=y1+(y2-1)*(t-t1) / (t2-t1)

[0117] Among them, (t1,y1) and (t2,y2) are two known data points before and after time t.

[0118] Take the interpolation of humidity data as an example: Assume that the humidity is 60% at 10:00 and 65% at 10:05. Now you need to interpolate the humidity at 10:02. Use the linear interpolation formula: y(10:02) = 60% + (65% - 60%) * (2 minutes / 5 minutes) = 62%. In this way, the estimated humidity value at 10:02 is 62%. Repeat this process for each time point that needs to be interpolated. It should be noted that at the beginning and end of the data, there may be known data points on only one side. In this case, you can choose to use the nearest known value or mark it as a missing value, depending on the needs of the application scenario.

[0119] The advantage of linear interpolation is its simplicity and computational efficiency. It can provide reasonable estimates for most environmental parameters (such as temperature, humidity, air pressure, etc.) because these parameters usually change relatively slowly over a short period of time. However, for parameters that may have rapid changes or periodic fluctuations (such as light intensity suddenly turning sunny on a cloudy day), linear interpolation may not be able to accurately capture these changes.

[0120] The engineering equipment data with the highest sampling frequency is selected as the second benchmark data, and all other engineering equipment data are processed with cubic spline interpolation based on this standard. Specifically, it is first necessary to determine which equipment parameter has the highest sampling frequency. This is usually determined by comparing the data collection intervals of different equipment sensors. For example, suppose the load sensor of the tower crane collects data once per second, the position sensor of the excavator collects data every 5 seconds, and the pressure sensor of the concrete pump collects data every 10 seconds. In this case, the load data of the tower crane will be selected as the second benchmark data because it has the highest sampling frequency (once per second). After the benchmark data is selected, a unified timeline needs to be created. This timeline should cover the time range of all equipment data and use the sampling interval of the benchmark data.

[0121] Continuing with the above example, if the entire data set covers 8 hours of working time, the unified timeline will contain 28,800 time points (8 hours * 3,600 seconds / hour). Next, for equipment parameters that have a sampling frequency lower than the benchmark data, cubic spline interpolation processing is required. Cubic spline interpolation is a more complex but also more flexible interpolation method that can generate smooth curves to connect known data points and maintain continuity at these points until the second-order derivative. This makes it particularly suitable for processing equipment data that may have rapid changes or complex patterns. The basic principle of cubic spline interpolation is to construct a cubic polynomial function between every two adjacent known data points. For n data points, there will be n-1 cubic polynomials. The form of each polynomial is:

[0122] S i (x) = a i +b i (xx i )+c i (xx i ) 2 +d i (xx i ) 3

[0123] Among them, i=1,2,...,n-1, represents the i-th interval, and x_i is the x-coordinate of the i-th known data point.

[0124] To determine the coefficients of these polynomials, the following conditions need to be met:

[0125] 1. Every polynomial must pass through both endpoints of its interval.

[0126] 2. The first-order derivatives of adjacent polynomials at common endpoints must be equal (to ensure smoothness).

[0127] 3. The second-order derivatives of adjacent polynomials at common endpoints must be equal (to ensure curvature continuity).

[0128] 4. Additional boundary conditions (such as natural boundary conditions, i.e., the second-order derivative at the endpoints is 0) can be set at the two end points of the entire interpolation range.

[0129] These conditions form a linear system of equations, and by solving this system of equations, the coefficients of all polynomials can be obtained. Once these coefficients are determined, the corresponding polynomials can be used to calculate the interpolation at any time point. Take the interpolation of excavator position data as an example: suppose that the position of the excavator is (0,0) at t=0 seconds, (2,3) at t=5 seconds, and (5,4) at t=10 seconds. Now you need to interpolate the position at t=2 seconds. Using cubic spline interpolation, you first need to construct a spline function over the entire time range, and then use the polynomial corresponding to t=2 to calculate the interpolation. The advantage of cubic spline interpolation is that it can generate smooth curves, avoiding the sudden changes that may be caused by linear interpolation, while maintaining sensitivity to local features. This is particularly important for equipment data, because equipment parameters (such as position, speed, pressure, etc.) usually show complex nonlinear changes over time.

[0130] Cubic spline interpolation of engineering equipment data is an advanced method that balances data continuity, smoothness, and local feature retention. It provides high-quality equipment operation data for the intelligent digital supervision system, supporting deeper and more accurate equipment performance analysis and predictive maintenance. Through this processing, the dynamic behavior of engineering equipment can be understood more comprehensively and finely, thereby supporting more optimized equipment management and risk control strategies.

[0131] In one embodiment, extracting the environment data features and the equipment data features from the preprocessed engineering environment data and engineering equipment data respectively comprises the following steps:

[0132] Calculate and obtain the autocorrelation coefficient matrix of the preprocessed engineering environment data;

[0133] Extract the frequency domain characteristics of environmental data of engineering environmental data through discrete Fourier transform;

[0134] Integrate the autocorrelation coefficient matrix and the frequency domain characteristics of environmental data into environmental data characteristics of engineering environmental data;

[0135] Calculate the equipment data statistical characteristics of the engineering equipment data based on the preset target statistical parameters;

[0136] Use a dynamic time warping algorithm to calculate the distance between engineering equipment data and preset typical equipment data, and obtain equipment data pattern characteristics of the engineering equipment data;

[0137] The equipment data statistical features and equipment data pattern features are integrated into the equipment data features of the engineering equipment data.

[0138] In this embodiment, for engineering environment data, the feature extraction process includes two main parts: autocorrelation analysis and frequency domain analysis. First, the autocorrelation coefficient matrix of the engineering environment data is calculated. Autocorrelation analysis can reveal the correlation of data at different time delays and help identify periodic patterns and long-term trends in the data. The calculation formula of the autocorrelation coefficient ρ(k) is:

[0139] ρ(k)=∑[(x[t]-μ)(x[t+k]-μ)] / ∑[(x[t]-μ) 2 ]

[0140] Among them, x[t] represents the time series, μ is the mean value of the series, and k is the time delay. By calculating the autocorrelation coefficients under different k values, we can get an autocorrelation coefficient matrix, which contains the important characteristics of environmental data in the time dimension.

[0141] Next, the frequency domain features of the engineering environment data are extracted through discrete Fourier transform (DFT). DFT can convert the time domain signal into frequency domain representation and reveal the periodic components in the data. The basic formula of DFT is:

[0142] X[k]=∑[x[n]*e (-j2πkn / N) ]

[0143] Among them, x[n] is a discrete time signal of length N, and X[k] is its frequency domain representation. By analyzing the amplitude and phase of the spectrum, the main frequency components in the environmental data can be identified, which is very useful for understanding the periodic characteristics of environmental changes. By integrating the autocorrelation coefficient matrix and the frequency domain characteristics of the environmental data, the complete environmental data characteristics of the engineering environmental data are obtained. These characteristics can fully describe the time correlation and frequency characteristics of the environmental data, providing an important basis for subsequent environmental status assessment and prediction.

[0144] For engineering equipment data, the feature extraction process includes two parts: statistical feature extraction and pattern feature extraction. First, the equipment data statistical characteristics of the engineering equipment data are statistically calculated based on the preset target statistical parameters. These statistical parameters may include mean, standard deviation, skewness, kurtosis, quantiles, etc. These statistical characteristics can summarize the overall distribution and variation characteristics of the equipment data. For example, the standard deviation can reflect the degree of fluctuation of the equipment parameters, and the skewness can reflect the asymmetry of the data distribution. Next, the dynamic time warping (DTW) algorithm is used to calculate the distance between the engineering equipment data and the preset typical equipment data to obtain the equipment data pattern characteristics of the engineering equipment data. The core idea of ​​the DTW algorithm is to find the best alignment between two time series, even if the two series differ in time or speed. The calculation process of the DTW distance is as follows:

[0145] 1. Construct a distance matrix D, where D(i,j) represents the distance between the i-th point in sequence A and the j-th point in sequence B.

[0146] 2. Construct a cumulative distance matrix M, where M(i,j) represents the minimum cumulative distance from (1,1) to (i,j).

[0147] 3. Starting from M(n,m), find the optimal path by backtracking.

[0148] The final DTW distance is the value of M(n,m), which represents the similarity between the two sequences. By calculating the DTW distance between the current device data and multiple typical device data patterns, a set of feature vectors describing the current device operation mode can be obtained.

[0149] By integrating the statistical features of equipment data and the pattern features of equipment data, we can obtain the complete equipment data features of engineering equipment data. These features not only contain the statistical characteristics of equipment operation, but also reflect the similarity between the equipment operation mode and the typical mode, providing rich information for subsequent equipment status evaluation and anomaly detection. The implementation effect of this feature extraction method is multifaceted. First, it greatly reduces the dimension of the data and converts high-dimensional raw data into a more compact and meaningful feature representation. This not only improves the efficiency of subsequent processing, but also helps to reduce the risk of overfitting. Secondly, the extracted features are highly interpretable. For example, the frequency domain features of environmental data can intuitively reflect the periodicity of environmental changes, and the DTW distance of equipment data can reflect the degree of deviation of the current operation mode from the normal mode. Furthermore, these features provide high-quality input for subsequent model training and decision making, which helps to improve the performance and reliability of the entire system.

[0150] In one implementation, fusing the environmental data features and the equipment data features to obtain the engineering supervision data features of the target construction site includes the following steps:

[0151] Consider environmental data features and device data features as key model concepts;

[0152] Define basic classes and construct the ontology model hierarchy based on key model concepts as the foundation of the ontology model;

[0153] Establish basic class relationships between basic classes based on semantic associations between basic classes;

[0154] The engineering supervision data model is constructed by combining the ontology model hierarchy and basic class relationships;

[0155] Feature engineering is used to extract the engineering supervision data features of the target construction site from the engineering supervision data model.

[0156] In this embodiment, the feature fusion process adopts a method based on the ontology model, which can effectively capture the semantic association between different features and construct a structured knowledge representation. The specific implementation steps are as follows: First, the environmental data features and equipment data features are used as key model concepts. These features may include the autocorrelation coefficient, frequency domain features, statistical features, and pattern features extracted in the previous step. Each feature is regarded as an independent concept, and they together constitute the basic semantic unit describing the state of the engineering site. Next, these key model concepts are used as the basis of the ontology model to define basic classes and construct the ontology model hierarchy. The ontology model is a formalized knowledge representation method that describes the knowledge system of a specific field by defining concepts, attributes, and relationships. In this step, the environmental data features and equipment data features can be defined as two top-level classes, and then further subdivided into subclasses under each top-level class. For example, the environmental data feature class can contain subclasses such as temperature features, humidity features, and noise features, while the equipment data feature class can contain subclasses such as load features, vibration features, and energy consumption features.

[0157] When constructing the ontology model hierarchy, it is necessary to consider the subordination and inheritance relationships between concepts. For example, the daily temperature change feature can be a subclass of the temperature feature, and the temperature feature is a subclass of the environmental data feature. This hierarchical structure can not only clearly organize various features, but also reflect the logical relationship between features.

[0158] Then, basic class relationships are established based on the semantic associations between basic classes. These relationships can be varied, for example:

[0159] 1. "Influence" relationship: Temperature characteristics may affect device performance characteristics.

[0160] 2. "Dependency" relationship: Certain equipment operating characteristics may depend on specific environmental conditions.

[0161] 3. "Correlation" relationship: There may be correlation between the operating characteristics of different devices.

[0162] 4. "Composition" relationship: multiple low-level features may together constitute a high-level feature.

[0163] The definition of these relationships needs to be based on the knowledge of domain experts and the results of data analysis. For example, the influence relationship between features can be determined through correlation analysis or causal inference, and the correlation between features can be discovered through cluster analysis. The purpose of defining these relationships is to capture the complex interactions between features, which is crucial for subsequent comprehensive analysis and decision-making. For example, by establishing the "influence" relationship between temperature features and equipment performance features, the impact of environmental changes on equipment operation can be better understood and predicted. Next, the engineering supervision data model is constructed by combining the ontology model hierarchy and basic class relationships. This model is a comprehensive knowledge graph that not only contains various features and their hierarchical relationships, but also various semantic associations between features. This model can be represented in a graphical way, where nodes represent features or concepts and edges represent relationships between features.

[0164] The process of building a project supervision data model may require multiple iterations and optimizations. After the initial construction of the model, it may be necessary to verify the rationality of the model through actual data, and adjust and improve the model based on the verification results. This process may involve adding new features, adjusting the relationship between features, redefining certain concepts, etc. Finally, feature engineering is used to extract the project supervision data features of the target construction site from the project supervision data model. The purpose of this step is to convert the constructed knowledge model into a feature vector that can be used for machine learning and data analysis. Feature engineering methods may include:

[0165] 1. Path feature: Generate features based on the paths in the knowledge graph. For example, the shortest path length between two features can be calculated as a new feature.

[0166] 2. Subgraph features: Extract subgraphs with specific patterns as features. For example, the number of neighbors or the distribution of neighbor types of a feature node can be counted as features.

[0167] 3. Embedded features: Use graph embedding algorithms (such as Node2Vec, DeepWalk, etc.) to map nodes in the knowledge graph into low-dimensional vector space, which can be used as features.

[0168] 4. Relationship features: Generate features based on the type and strength of the relationship between nodes. For example, you can calculate the connection strength between a feature node and different types of nodes.

[0169] 5. Semantic features: Generate features using the concept hierarchy and semantic relationships defined in the ontology model. For example, features can be generated based on the hierarchical depth of concepts or the semantic distance between concepts.

[0170] Through these methods, the complex engineering supervision data model can be transformed into a set of structured feature vectors, which not only contain the information of the original data, but also integrate the complex relationship between domain knowledge and features. The implementation effect of this feature fusion method is remarkable. First, it realizes the organic combination of environmental data features and equipment data features, creating a unified feature space, so that subsequent analysis can consider the influence of environmental factors and equipment factors at the same time. Secondly, the method based on the ontology model introduces rich semantic information, so that the features are no longer isolated values, but knowledge units with clear meanings and associations. This greatly improves the interpretability and usability of the features. Furthermore, by constructing a knowledge graph, the complex interactions between features can be better captured, which is crucial for understanding the overall situation of the engineering site and predicting potential problems. Finally, this method has strong flexibility and scalability. With the addition of new data sources or the introduction of new domain knowledge, the ontology model can be easily updated and expanded, thereby continuously improving the expression ability and coverage of features. This is of great significance for the continuous optimization and upgrading of the intelligent digital supervision system.

[0171] In one implementation, the engineering supervision data features are uniformly mapped to a preset high-dimensional feature space, and feature distribution analysis is performed in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result, which includes the following steps:

[0172] Based on the radial basis function kernel, the engineering supervision data features are uniformly mapped to the preset high-dimensional feature space;

[0173] Analyze the high-dimensional data distribution characteristics from the high-dimensional feature space through principal component analysis;

[0174] Perform cluster analysis on high-dimensional data in the high-dimensional feature space, and use the cluster analysis results as cluster features of the high-dimensional data in the high-dimensional feature space;

[0175] The high-dimensional data distribution characteristics and high-dimensional data clustering characteristics are combined as the multidimensional distribution feature analysis results.

[0176] In this embodiment, first, the engineering supervision data features are uniformly mapped to a preset high-dimensional feature space based on the radial basis function (RBF) kernel. The RBF kernel is a commonly used kernel function that can map data points in the original feature space to an infinite-dimensional feature space. The expression of the RBF kernel is:

[0177] K(x,y)=exp(-γ||xy||2 )

[0178] Among them, x and y are two data points in the original feature space, and γ is an adjustable parameter used to control the "tightness" of the mapping. Through this mapping, data that was originally linearly inseparable in low-dimensional space becomes linearly separable in high-dimensional space. An important feature of RBF kernel mapping is that it does not require explicit calculation of coordinates in high-dimensional space, but directly calculates the inner product in high-dimensional space through kernel techniques. This greatly reduces the computational complexity and makes the analysis of high-dimensional space. In practical applications, the optimal γ value can be selected through methods such as grid search or cross-validation. For example, you can try a series of γ values ​​(such as 0.1, 1, 10, etc.) and select the one that performs best on the validation set.

[0179] Next, the principal component analysis (PCA) method is used to parse the high-dimensional data distribution characteristics from the high-dimensional feature space. Although the RBF kernel maps the data to an infinite-dimensional space, PCA can help find the main direction of change of the data in this high-dimensional space. The basic steps of PCA are as follows:

[0180] 1. Calculate the covariance matrix of the data.

[0181] 2. Calculate the eigenvalues ​​and eigenvectors of the covariance matrix.

[0182] 3. Sort the eigenvectors by the size of the corresponding eigenvalues.

[0183] 4. Select the first k eigenvectors to form a projection matrix.

[0184] 5. Use this projection matrix to project the original data into a k-dimensional subspace.

[0185] In kernel PCA, these calculations are performed in the feature space, but through the kernel technique, all calculations can be completed in the original space. Ultimately, the distribution characteristics of the data in the main change direction can be obtained, which can reflect the main structure and change pattern of the data. For example, suppose that through PCA analysis, it is found that the first three principal components explain 95% of the data variance. Then these three principal components can be used as new features, which represent certain comprehensive status indicators of the engineering site. By analyzing the physical meaning of these principal components, a deeper understanding of the engineering supervision data can be obtained.

[0186] Next, cluster analysis is performed on the high-dimensional data in the high-dimensional feature space, and the cluster analysis results are used as the clustering features of the high-dimensional data in the high-dimensional feature space. Commonly used clustering algorithms include K-means, DBSCAN, Gaussian mixture model, etc. Here, the K-means algorithm is used as an example to illustrate the clustering process:

[0187] 1. Select K initial cluster centers.

[0188] 2. Assign each data point to the nearest cluster center.

[0189] 3. Recalculate the center of each cluster.

[0190] 4. Repeat steps 2 and 3 until the cluster center no longer changes significantly or the maximum number of iterations is reached.

[0191] When performing K-means clustering in kernel space, the distance calculation needs to be completed through the kernel function. Specifically, the distance between two points x and y in the feature space can be calculated by the following formula:

[0192] d 2 (x,y)=K(x,x)+K(y,y)-2K(x,y)

[0193] Where K is the RBF kernel function.

[0194] The results of cluster analysis can reveal natural groups in the data, which represent different engineering states or operating modes. For example, it is found that the data is naturally divided into three categories, corresponding to normal operating state, mild abnormal state and severe abnormal state. By combining the results of PCA and cluster analysis, a multi-dimensional feature distribution analysis result can be obtained. This result not only includes the main direction of change of the data, but also includes the natural grouping of the data. This multi-dimensional analysis result provides a rich information basis for subsequent anomaly detection and decision support.

[0195] The implementation effects of this high-dimensional feature space mapping and distribution analysis method are multifaceted:

[0196] 1. Nonlinear feature extraction: Through RBF kernel mapping, it is possible to capture nonlinear features and relationships in the data, which are difficult to find in the original feature space.

[0197] 2. Dimensionality reduction and feature selection: PCA analysis helps identify the most important feature combinations, which not only reduces the dimensionality of the data but also retains most of the information.

[0198] 3. Pattern discovery: Cluster analysis reveals natural groupings in the data that correspond to different engineering states or operating modes.

[0199] 4. Anomaly detection basis: The data distribution characteristics in high-dimensional space provide an important basis for subsequent anomaly detection. For example, points far away from the cluster center are potential anomalies.

[0200] 5. Visualization support: Although the original data has a high dimension, PCA can project the data into a two-dimensional or three-dimensional space for visualization, which helps to intuitively understand the structure and distribution of the data.

[0201] In general, this step maps the characteristics of engineering supervision data to a more expressive high-dimensional space and conducts a comprehensive distribution feature analysis in this space. This not only provides a deeper understanding of the data, but also lays a solid mathematical foundation for subsequent anomaly detection and decision support. In this way, valuable patterns and insights can be extracted from complex engineering supervision data, providing strong support for the efficient operation of the intelligent digital supervision system.

[0202] The present invention also discloses an engineering data processing system based on intelligent digital supervision, referring to Figure 2 , the system includes:

[0203] The data acquisition module is used to collect the engineering site data of the target construction site through the sensor network deployed at the target construction site.

[0204] Engineering site data includes engineering environment data and engineering equipment data;

[0205] Data preprocessing module, used for preprocessing engineering environment data and engineering equipment data;

[0206] A feature extraction module, used to extract environmental data features and equipment data features from the preprocessed engineering environment data and engineering equipment data respectively;

[0207] The feature fusion module is used to fuse the environmental data features and the equipment data features to obtain the engineering supervision data features of the target construction site;

[0208] The feature analysis module is used to uniformly map the features of the engineering supervision data to a preset high-dimensional feature space, and perform feature distribution analysis in the high-dimensional feature space to obtain multi-dimensional distribution feature analysis results;

[0209] The anomaly identification module is used to output the engineering anomaly results of the target construction site based on the multi-dimensional distribution feature analysis results and using a support vector machine.

[0210] In one embodiment, the engineering environment data and the engineering equipment data are both time series data, and the data preprocessing module includes:

[0211] A data screening unit, used to screen out all abnormal data values ​​in engineering environment data and engineering equipment data;

[0212] A data alignment unit is used to perform time alignment processing on all engineering environment data and all engineering equipment data after outliers are screened out;

[0213] A first data interpolation unit is used to select the engineering environment data with the highest sampling frequency as the first reference data, and perform linear interpolation processing on all other engineering environment data based on the first reference data;

[0214] The second data interpolation unit is used to select the engineering equipment data with the highest sampling frequency as the second reference data, and perform cubic spline interpolation processing on all other engineering equipment data based on the second reference data.

[0215] In one embodiment, the feature extraction module includes:

[0216] A time domain feature extraction unit, used for calculating the autocorrelation coefficient matrix of the pre-processed engineering environment data;

[0217] A frequency domain feature extraction unit, used for extracting frequency domain features of environmental data of engineering environmental data by discrete Fourier transform;

[0218] A first feature integration unit is used to integrate the autocorrelation coefficient matrix and the frequency domain features of the environmental data into environmental data features of the engineering environmental data;

[0219] A statistical feature extraction unit, used to calculate the equipment data statistical features of the engineering equipment data based on preset target statistical parameters;

[0220] A pattern feature extraction unit, used to calculate the distance between the engineering equipment data and the preset typical equipment data using a dynamic time warping algorithm, and obtain the equipment data pattern feature of the engineering equipment data;

[0221] The second feature integration unit is used to integrate the equipment data statistical features and the equipment data pattern features into the equipment data features of the engineering equipment data.

[0222] In one embodiment, the feature fusion module includes:

[0223] A model concept selection unit, used to select environmental data features and equipment data features as key model concepts;

[0224] A model structure building unit is used to define basic classes and build a hierarchical structure of the ontology model based on key model concepts;

[0225] A model relationship building unit, used for establishing basic class relationships between basic classes according to semantic associations between basic classes;

[0226] An ontology model building unit is used to build a project supervision data model by combining the ontology model hierarchy and basic class relationships;

[0227] The feature selection unit is used to extract the engineering supervision data features of the target construction site from the engineering supervision data model by using feature engineering.

[0228] In one embodiment, the feature analysis module includes:

[0229] A feature mapping unit, used to uniformly map the features of the engineering supervision data into a preset high-dimensional feature space based on a radial basis function kernel;

[0230] A distribution feature analysis unit, used to analyze the high-dimensional data distribution features from the high-dimensional feature space by principal component analysis;

[0231] A clustering feature analysis unit, used for performing clustering analysis on high-dimensional data in the high-dimensional feature space, and using the clustering analysis result as a clustering feature of the high-dimensional data in the high-dimensional feature space;

[0232] The feature analysis result output unit is used to combine the high-dimensional data distribution characteristics and the high-dimensional data clustering characteristics as a multi-dimensional distribution feature analysis result.

[0233] A person skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of protection of the present application is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of different aspects of one or more embodiments of the present application as above, which are not provided in detail for the sake of simplicity.

[0234] One or more embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the present application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present application should be included in the protection scope of the present application.

Claims

1. A method for processing engineering data based on intelligent digital supervision, characterized in that: The steps include: Collecting engineering site data of the target construction site through a sensor network deployed at the target construction site, wherein the engineering site data includes engineering environment data and engineering equipment data; Preprocessing the engineering environment data and the engineering equipment data; Extracting environment data features and equipment data features from the preprocessed engineering environment data and engineering equipment data respectively; Fusing the environmental data features with the equipment data features to obtain the engineering supervision data features of the target construction site; The engineering supervision data features are uniformly mapped to a preset high-dimensional feature space, and feature distribution analysis is performed in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result; Based on the multi-dimensional distribution feature analysis results and using a support vector machine, the engineering abnormality results of the target construction site are output.

2. The engineering data processing method based on intelligent digital supervision according to claim 1 is characterized in that: The engineering environment data and the engineering equipment data are both time series data, and the preprocessing of the engineering environment data and the engineering equipment data includes the following steps: Screening out all abnormal data values ​​in the engineering environment data and the engineering equipment data; Performing time alignment processing on all the engineering environment data and all the engineering equipment data after outliers are screened out; Selecting the engineering environment data with the highest sampling frequency as the first benchmark data, and using the first benchmark data as a standard to perform linear interpolation processing on all other engineering environment data; The engineering equipment data with the highest sampling frequency is selected as the second benchmark data, and the second benchmark data is used as a standard to perform cubic spline interpolation processing on all other engineering equipment data.

3. The engineering data processing method based on intelligent digital supervision according to claim 1 is characterized in that: The extracting of environment data features and equipment data features from the pre-processed engineering environment data and engineering equipment data respectively comprises the following steps: Calculating and obtaining the autocorrelation coefficient matrix of the preprocessed engineering environment data; Extracting frequency domain features of environmental data of the engineering environmental data by discrete Fourier transform; Integrating the autocorrelation coefficient matrix and the frequency domain features of the environmental data into environmental data features of the engineering environmental data; Calculating equipment data statistical features of the engineering equipment data based on preset target statistical parameters; Using a dynamic time warping algorithm to calculate the distance between the engineering equipment data and preset typical equipment data, and obtain equipment data pattern characteristics of the engineering equipment data; The equipment data statistical features and the equipment data pattern features are integrated into the equipment data features of the engineering equipment data.

4. The engineering data processing method based on intelligent digital supervision according to claim 3 is characterized in that: The step of fusing the environmental data features with the equipment data features to obtain the engineering supervision data features of the target construction site comprises the following steps: Taking the environment data features and the device data features as key model concepts; Using the key model concepts as the basis of the ontology model, defining basic classes and constructing an ontology model hierarchy; Establishing a basic class relationship between the basic classes according to the semantic association between the basic classes; Combining the ontology model hierarchy and the basic class relationship to construct a project supervision data model; The engineering supervision data features of the target construction site are extracted from the engineering supervision data model using feature engineering.

5. The engineering data processing method based on intelligent digital supervision according to claim 4 is characterized in that: The step of uniformly mapping the engineering supervision data features to a preset high-dimensional feature space, and performing feature distribution analysis in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result comprises the following steps: Based on the radial basis function kernel, the engineering supervision data features are uniformly mapped to a preset high-dimensional feature space; Analyzing the high-dimensional data distribution characteristics from the high-dimensional feature space by principal component analysis; Performing cluster analysis on the high-dimensional data in the high-dimensional feature space, and using the cluster analysis results as cluster features of the high-dimensional data in the high-dimensional feature space; The high-dimensional data distribution characteristics and the high-dimensional data clustering characteristics are combined as a multi-dimensional distribution characteristic analysis result.

6. An engineering data processing system based on intelligent digital supervision, characterized in that: The system comprises: A data acquisition module, used to collect engineering site data of the target construction site through a sensor network deployed at the target construction site, wherein the engineering site data includes engineering environment data and engineering equipment data; A data preprocessing module, used for preprocessing the engineering environment data and the engineering equipment data; A feature extraction module, used to extract environment data features and equipment data features from the preprocessed engineering environment data and engineering equipment data respectively; A feature fusion module, used to fuse the environmental data features with the equipment data features to obtain the engineering supervision data features of the target construction site; A feature analysis module is used to uniformly map the features of the engineering supervision data to a preset high-dimensional feature space, and perform feature distribution analysis in the high-dimensional feature space to obtain a multi-dimensional distribution feature analysis result; The anomaly identification module is used to output the engineering anomaly results of the target construction site based on the multi-dimensional distribution feature analysis results and using a support vector machine.

7. The engineering data processing system based on intelligent digital supervision according to claim 6 is characterized in that: The engineering environment data and the engineering equipment data are both time series data, and the data preprocessing module includes: A data screening unit, used to screen out all abnormal data values ​​in the engineering environment data and the engineering equipment data; A data alignment unit, used for performing time alignment processing on all the engineering environment data and all the engineering equipment data after outliers are screened out; A first data interpolation unit is used to select the engineering environment data with the highest sampling frequency as the first reference data, and perform linear interpolation processing on all other engineering environment data based on the first reference data; The second data interpolation unit is used to select the engineering equipment data with the highest sampling frequency as the second reference data, and perform cubic spline interpolation processing on all other engineering equipment data based on the second reference data.

8. The engineering data processing system based on intelligent digital supervision according to claim 6 is characterized in that: The feature extraction module comprises: A time domain feature extraction unit, used for calculating the autocorrelation coefficient matrix of the preprocessed engineering environment data; A frequency domain feature extraction unit, used for extracting frequency domain features of environmental data of the engineering environment data by discrete Fourier transform; A first feature integration unit, used for integrating the autocorrelation coefficient matrix and the frequency domain features of the environmental data into environmental data features of the engineering environmental data; A statistical feature extraction unit, used to calculate the equipment data statistical features of the engineering equipment data based on preset target statistical parameters; A pattern feature extraction unit, used to calculate the distance between the engineering equipment data and preset typical equipment data using a dynamic time warping algorithm to obtain equipment data pattern features of the engineering equipment data; The second feature integration unit is used to integrate the equipment data statistical features and the equipment data pattern features into the equipment data features of the engineering equipment data.

9. The engineering data processing system based on intelligent digital supervision according to claim 8 is characterized in that: The feature fusion module includes: A model concept selection unit, used for taking the environmental data features and the device data features as key model concepts; A model structure construction unit, used to define basic classes based on the key model concepts as the basis of the ontology model and to construct a hierarchical structure of the ontology model; A model relationship building unit, used for establishing a basic class relationship between the basic classes according to the semantic association between the basic classes; An ontology model construction unit, used to construct a project supervision data model by combining the ontology model hierarchy and the basic class relationship; A feature selection unit is used to extract the engineering supervision data features of the target construction site from the engineering supervision data model by using feature engineering.

10. The engineering data processing system based on intelligent digital supervision according to claim 9 is characterized in that: The feature analysis module comprises: A feature mapping unit, used for uniformly mapping the engineering supervision data features into a preset high-dimensional feature space based on a radial basis function kernel; A distribution feature analysis unit, used to analyze the high-dimensional data distribution features from the high-dimensional feature space by principal component analysis; A clustering feature analysis unit, used to perform clustering analysis on the high-dimensional data in the high-dimensional feature space, and use the clustering analysis result as the clustering feature of the high-dimensional data in the high-dimensional feature space; The feature analysis result output unit is used to combine the high-dimensional data distribution feature and the high-dimensional data clustering feature as a multi-dimensional distribution feature analysis result.

Images