Bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data

Abstract

The present invention relates to a bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data, belonging to the field of natural language processing and data identification. The method comprises the following steps: collecting bridge load-bearing monitoring vibration data; manually labeling the bridge load-bearing monitoring vibration data; preprocessing the labeled bridge load-bearing monitoring vibration data to obtain preprocessed bridge load-bearing monitoring vibration data; constructing a bridge load-bearing monitoring vibration data anomaly identification model, the model adopts an extreme learning machine model based on a multi-objective loss function, the model is trained to obtain a trained model; the newly collected bridge load-bearing monitoring vibration data is input into the trained model to obtain the model output result, and whether the bridge load-bearing monitoring vibration data has anomalies; the anomaly identification results and monitoring indicators are fed back to the knowledge graph system to provide auxiliary decision-making. The present invention can improve the adaptability and robustness of the model under complex working conditions.

Description

Technical Field

[0001] The present invention belongs to the field of natural language processing and data recognition, and specifically relates to a bridge load-bearing monitoring data recognition method based on knowledge drive and monitoring data. Background Art

[0002] As my country's transportation infrastructure continues to expand, a large number of in-service bridge structures have entered an aging phase. Traditional manual inspection methods are unable to meet the needs of real-time, accurate safety monitoring. While existing automated monitoring systems can collect massive amounts of vibration data, they still have significant deficiencies in anomaly identification: conventional machine learning methods often focus on a single classification metric, ignoring the unique spatiotemporal correlations and manifold structure characteristics of bridge vibration data; most models lack effective modeling of the spatial topological constraints of sensor networks, resulting in poor cross-sensor data fusion; and, due to the widespread noise interference and operating condition changes in engineering environments, the robustness and adaptability of existing algorithms are insufficient to meet actual engineering requirements. These technical deficiencies have resulted in high false alarm and missed alarm rates, severely restricting the engineering application value of monitoring systems.

[0003] Existing bridge inspection methods have the following problems that need to be further addressed: Conventional bridge monitoring data processing methods often rely solely on simple classification models, such as support vector machines and decision trees, which may not be able to effectively process high-dimensional, nonlinear, and complex vibration data. These methods also often ignore the preservation of the data manifold structure, resulting in performance degradation under complex working conditions. Existing monitoring systems generally lack adaptive processing mechanisms. In particular, in anomaly detection, it is difficult to automatically adjust the weights of different types of abnormal samples, resulting in some abnormal data being ignored or misjudged. The lack of a dynamic weighting mechanism makes the system less tolerant to noise and prone to false positives and negatives. Traditional monitoring systems rely heavily on sensor data itself, ignoring additional knowledge such as environmental factors, historical data, and material fatigue properties. This leads to limited monitoring results and an inability to provide comprehensive decision support. The lack of intelligent reasoning based on knowledge graphs makes the system's decision support capabilities insufficient and unable to fully optimize maintenance strategies. Existing models are often prone to overfitting under specific working conditions and are unable to cope with the diversity and complexity of bridge conditions. Traditional models ignore the inherent connections between different data manifolds, resulting in poor adaptability to new working conditions and difficulty in ensuring stable performance output. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides a bridge load-bearing monitoring data identification method based on knowledge drive and monitoring data.

[0005] To achieve the above-mentioned purpose, the present invention is implemented through the following technical solutions:

[0006] The present invention provides a bridge load-bearing monitoring data identification method based on knowledge drive and monitoring data, comprising the following steps:

[0007] S1. Collect bridge load-bearing monitoring vibration data by installing multiple vibration sensors on the bridge structure;

[0008] S2. Manually label the bridge load-bearing monitoring vibration data;

[0009] S3. Preprocessing the labeled bridge load-bearing monitoring vibration data to obtain preprocessed bridge load-bearing monitoring vibration data; the preprocessing steps include data cleaning, denoising, and standardization;

[0010] S4. Construct a model for identifying anomalies in bridge load-bearing monitoring vibration data. This model uses an extreme learning machine model based on a multi-objective loss function. The loss function incorporates topology preservation loss and generalization loss, and dynamically adjusts the weights of abnormal samples through an adaptive sample weight allocation mechanism. The preprocessed bridge load-bearing monitoring vibration data is input into the model and trained to obtain a trained model.

[0011] S5. Preprocess the newly collected bridge load-bearing monitoring vibration data and input it into the trained model to obtain the model output and determine whether there are any abnormalities in the bridge load-bearing monitoring vibration data;

[0012] S6. Feedback the anomaly identification results and monitoring indicators to the knowledge graph system to provide auxiliary decision-making.

[0013] Furthermore, step S1 specifically includes:

[0014] The vibration sensors are arranged at key positions of the bridge, including supports, bridge deck and beam ends; the vibration sensors regularly collect acceleration or strain information of the bridge under load.

[0015] Furthermore, step S4 specifically includes:

[0016] S41 initializes the parameters in the extreme learning machine model, including the hidden layer weight matrix and bias vector of the extreme learning machine. The initialization method is random initialization, and the initialized parameters obey the normal distribution with mean 0 and variance as the unit matrix;

[0017] S42. Calculate the multi-objective loss function of the extreme learning machine: During the training process, a multi-objective collaborative optimization method is used to adjust and optimize the model parameters by optimizing multiple objective functions. The multi-objective loss function serves as the total loss function of the extreme learning machine training process, including classification loss, topology preservation loss, and generalization loss. The classification loss is defined as the cross-entropy loss. The topology preservation loss is defined by minimizing the topology preservation loss, which is expressed as follows:

[0018] ,

[0019] in, represents the topology preservation loss, and They represent the output of the hidden layer and samples, Indicates the and The neighborhood relationship matrix weight of the sample represents the and The similarity of samples; the generalization loss is defined by the regularization of the output layer weights of the model, and the formula is as follows:

[0020] ,

[0021] in, represents the generalization loss, Represents the regularization coefficient, which controls the proportion of generalization loss in the overall loss function. represents the output layer weight of the extreme learning machine, is the L2 norm, represents the adaptive sample weight loss, Represents multi-level manifold learning constraints;

[0022] S43. Perform data forward propagation calculations for the extreme learning machine: During the training process of the extreme learning machine, data forward propagation calculations involve the hidden layer output and the output layer weight update. The calculation method for the hidden layer output is expressed as:

[0023] ,

[0024] in, represents the Hadamard product, express activation function, Representation layer normalization operation, represents the hidden layer weight matrix of the extreme learning machine, represents the hidden layer input of the extreme learning machine, represents the hidden layer bias vector of the extreme learning machine, represents the Sigmoid activation function, represents the hidden layer output of the extreme learning machine, represents the dynamic adjacency matrix, represents the fusion feature; the weight update method of the output layer of the extreme learning machine is expressed as:

[0025] ,

[0026] ,

[0027] in, express The transpose of represents the adjustment matrix, represents the inverse matrix of the adjustment matrix, represents the identity matrix, represents the regularization coefficient, represents the target label matrix, represents the output layer weight of the extreme learning machine;

[0028] S44. Perform error back propagation calculation of the extreme learning machine: Update the hidden layer weight matrix and bias vector of the extreme learning machine based on error back propagation, and use the gradient descent method to calculate the multi-objective loss function of the extreme learning machine with respect to the weight matrix The gradient of and the multi-objective loss function of the extreme learning machine with respect to the bias vector gradient;

[0029] S45. The extreme learning machine gradually updates the hidden layer weight matrix and bias vector through multiple iterations until the loss function converges to the preset minimum value.

[0030] Furthermore, step S5 specifically includes:

[0031] S51. Perform the same preprocessing operations as the training data on the newly collected bridge load-bearing monitoring vibration data, including data cleaning, denoising, and standardization;

[0032] S52. Input the preprocessed data into the trained model and perform forward propagation calculation to obtain the output result of the model; based on the output result of the model and in combination with the set threshold or classification standard, determine whether there is any abnormality in the bridge load-bearing monitoring vibration data.

[0033] Furthermore, in step S6, the anomaly identification results and monitoring indicators are fed back to the knowledge graph system. The knowledge graph system obtains relevant structural health assessment knowledge, historical maintenance records, and multi-source heterogeneous data to build a decision support chain based on semantic reasoning. The specific process includes:

[0034] S61. The knowledge graph system is a natural language processing knowledge graph library; real-time abnormal features are aligned with entity relationships in the knowledge graph through graph embedding, and the similarity between abnormal events and historical cases is calculated through a graph attention network;

[0035] S62. Make decisions based on real-time anomaly features and entity relationships in the knowledge graph.

[0036] Furthermore, in step S42, the neighborhood relationship matrix weight is calculated by combining the spatiotemporal correlation and local manifold density characteristics of the bridge load-bearing monitoring vibration data, and adopting an adaptive bimodal similarity measurement method for calculation. By multiplying the feature space Gaussian kernel and the spatiotemporal attenuation function, the nonlinear characteristics of the bridge load-bearing monitoring vibration data and the physical constraints of the monitoring system are simultaneously captured.

[0037] The adaptive sample weight loss automatically assigns a weight to each sample by analyzing the time domain and frequency domain characteristic attributes of the bridge load-bearing monitoring vibration data, so that the model focuses on the bridge load-bearing monitoring vibration data at abnormal moments;

[0038] The hierarchical manifold learning constraints capture the global and local structures of bridge load-bearing monitoring vibration data at different levels, allowing the model to fully understand the manifold properties of the data.

[0039] Furthermore, in step S43, the dynamic adjacency matrix explicitly models the spatial constraints of the sensor network and the time-frequency coupling characteristics of the vibration signal through the neighborhood relationship matrix weights; the fusion feature is constructed by tensor splicing the hidden layer input of the extreme learning machine and the time-frequency features;

[0040] The regularization coefficient is dynamically adjusted according to the local density output by the extreme learning machine hidden layer.

[0041] The advantages of the present invention are:

[0042] The present invention proposes an extreme learning machine model based on a multi-objective loss function. In addition to the conventional classification loss, topology preservation loss and generalization loss are also added. It can effectively maintain the manifold structure of bridge load-bearing monitoring data during training, and improve the adaptability and robustness of the model under complex working conditions; topology preservation loss helps to retain the geometric structure of the data, enhances the manifold learning ability of the model, and makes the model more stable when processing different data distributions. Moreover, generalization loss improves the generalization ability of the model by regularizing the model weights, and prevents overfitting, especially in complex bridge monitoring data environments; based on the time domain and frequency domain characteristics of bridge load-bearing monitoring vibration data, the present invention proposes an automatic The sample weight distribution mechanism is adapted to dynamically adjust the weights of abnormal samples, so that the model can pay more attention to the data at abnormal moments, thereby improving the sensitivity and accuracy of anomaly detection. The present invention combines the abnormal identification results of bridge load-bearing monitoring with the knowledge graph, and generates a decision support chain based on semantic reasoning by integrating multi-source data such as historical maintenance data, structural health assessment knowledge, and environmental corrosion maps, which helps to evaluate the safety status of bridges in real time and optimize maintenance strategies. By constructing a dynamic adjacency matrix and fusing time-frequency features, the model's ability to capture spatiotemporal characteristics and transient characteristics of vibration signals is enhanced, enabling the model to accurately identify abnormal changes in bridge vibration signals and ensure more accurate fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0044] Figure 1 is a flow chart of the steps of the method of the present invention;

[0045] Figure 2 is an example of an anomaly data point that is labeled;

[0046] Figure 3 Generate example graphs for decision making;

[0047] Figure 4 Comparison of model accuracy under different noise levels;

[0048] Figure 5 is the influence of the regularization coefficient on the model performance. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments derived by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0050] Example 1

[0051] In this embodiment, Figure 1 As shown, the present invention provides a bridge load-bearing monitoring data identification method based on knowledge drive and monitoring data, and the specific steps include:

[0052] S1. Collect bridge load-bearing monitoring vibration data by installing multiple vibration sensors on the bridge structure;

[0053] Specifically, the vibration sensors are arranged at key positions of the bridge, including supports, bridge decks and beam ends; the vibration sensors regularly collect acceleration or strain information of the bridge under load.

[0054] S2. Manually label the vibration data of bridge load-bearing monitoring; the labeling content includes whether the vibration data is normal, and examples of labeled abnormal data points are as follows: Figure 2 shown.

[0055] S3. Preprocess the labeled bridge load-bearing monitoring vibration data to obtain preprocessed bridge load-bearing monitoring vibration data; the preprocessing steps include data cleaning, denoising and standardization; data cleaning is used to remove abnormal data points caused by sensor failure or environmental interference; further, filtering technology, such as low-pass filtering or wavelet transform, is used to remove high-frequency noise signals; further, through standardization or normalization processing, the data of each feature dimension are compared under the same dimension, thereby avoiding certain features from having too much influence on model training.

[0056] S4. Construct a model for identifying anomalies in bridge load-bearing monitoring vibration data. This model uses an extreme learning machine model based on a multi-objective loss function. The loss function incorporates topology preservation loss and generalization loss, and dynamically adjusts the weights of abnormal samples through an adaptive sample weight allocation mechanism. The preprocessed bridge load-bearing monitoring vibration data is input into the model and trained to obtain a trained model.

[0057] Specifically, S41 initializes the parameters in the extreme learning machine model, including the hidden layer weight matrix and bias vector of the extreme learning machine, and the initialization method is random initialization. The initialized parameters obey the normal distribution with a mean of 0 and a variance of the unit matrix;

[0058] S42. Calculate the multi-objective loss function of the extreme learning machine: Since bridge load-bearing monitoring vibration data usually has the characteristics of high dimensionality and nonlinear separability, conventional machine learning models such as extreme learning machines only use classification loss as the main indicator of the loss function during the training process, ignoring the maintenance of neighborhood relationships in the manifold structure of bridge load-bearing monitoring vibration data, resulting in insufficient generalization ability under complex working conditions. During the training process, a multi-objective collaborative optimization method is adopted to adjust and optimize the parameters of the model by optimizing multiple objective functions; so that the model not only performs well in classification tasks, but also maintains the manifold structure of bridge load-bearing monitoring vibration data, and improves robustness and adaptability in complex data environments. For example, the topology preservation loss ensures that samples of the same category with adjacent relationships on the manifold of bridge load-bearing monitoring vibration data remain adjacent in the representation space of the hidden layer, thereby better preserving the geometric structure of the data of this category, making it more adaptable and robust when processing complex bridge load-bearing monitoring vibration data;

[0059] The multi-objective loss function is the total loss function of the extreme learning machine training process, including classification loss, topology preservation loss and generalization loss; the classification loss is defined as cross entropy loss; the total loss function Expressed as:

[0060] ,

[0061] in, is the multi-objective loss function of the extreme learning machine, is the classification loss, is the topology preservation loss, is the generalization loss, is the weight coefficient of the first objective function, is the weight coefficient of the second objective function, is the weight coefficient of the third objective function. Preferably, Set to 0.6, Set to 0.2, Set to 0.3. The topology preservation loss adopts the method of minimizing the topology preservation loss, which is expressed as follows:

[0062] ,

[0063] in, and They represent the output of the hidden layer and samples, Indicates the and The neighborhood relationship matrix weight of the sample represents the and The similarity of samples (the premise is that and samples are of the same category, if and The weight of the neighborhood relationship matrix is set to 0 if the samples are not of the same category. The weight of the neighborhood relationship matrix of the samples is calculated by combining the spatiotemporal correlation and local manifold density characteristics of the bridge load-bearing monitoring vibration data, and adopting an adaptive bimodal similarity measurement method for calculation. The product of the feature space Gaussian kernel and the spatiotemporal attenuation function is used to capture the nonlinear characteristics of the bridge load-bearing monitoring vibration data and the physical constraints of the monitoring system at the same time, which is expressed as:

[0064]

[0065] Where, is an exponential function with the natural constant e as the base, For samples The local density adaptive bandwidth enables sparse area samples to obtain a larger local range of action and avoid overfitting in high density areas, which is defined as sample To the front The average distance between similar neighbors, that is, ; For samples The local density adaptive bandwidth is calculated in the same way as same; For samples Tags, For samples 's label; For the samples, representing the first bridge load-bearing monitoring vibration data; For the samples, representing the first bridge load-bearing monitoring vibration data; is the sensor spatial distance (measured in meters), is the number of neighbors of the sample; For samples mutual Neighbor set, that is, simultaneously satisfying and samples, enhance neighborhood robustness, mutual Neighbor screening preserves bidirectional neighborhood relationships, eliminates unidirectional pseudo-adjacencies caused by noise, and improves the reliability of topology preservation. for Nearest neighbor function, Indicates the The sample is The samples are Neighbor samples, Indicates the The sample is The samples are Neighbor samples, is a positive integer, is a positive integer, and the “ "Condition means and When the samples are not of the same category, the purpose is to build adjacency weights only between samples of the same category to avoid cross-category manifold confusion.

[0066] The generalization loss is defined by regularizing the output layer weights of the model, and is expressed as follows:

[0067] ,

[0068] in, Represents the regularization coefficient, which controls the proportion of generalization loss in the overall loss function. represents the output layer weight of the extreme learning machine, is the L2 norm, represents the adaptive sample weight loss, Represents multi-level manifold learning constraints;

[0069] The adaptive sample weight loss automatically assigns a weight to each sample by analyzing the time domain and frequency domain characteristic attributes of the bridge load-bearing monitoring vibration data, so that the model can pay more attention to the bridge load-bearing monitoring vibration data at abnormal moments. The calculation method is expressed as:

[0070] ,

[0071] in, is the total number of batch samples input to the extreme learning machine; is the time-frequency feature fusion coefficient, and Determined by cross-validation; is the adaptive adjustment factor, and , is the covariance matrix of the training set, It is the Mahalanobis distance calculation function, which automatically increases the weight of abnormal samples in sparse data distribution areas. It is the adjustment factor for abnormal samples, which is used to adjust the weight of samples, especially the weighting in the abnormal data area; For The frequency domain energy entropy obtained after discrete wavelet transform, and , For the Layer wavelet coefficients, It refers to the number of wavelet transform layers, which is preset manually. Wavelet transform extracts frequency domain features by decomposing the signal into sub-signals of multiple scales. The frequency domain energy distribution features are extracted by wavelet transform and combined with the time domain sliding window statistics to construct a bimodal anomaly indicator. For The time domain sliding window standard deviation is calculated as ( is the binary index corresponding to the data sampling frequency), For the The data segment of the time domain signal of samples in the sliding window; is the mean vector of the batch samples input to the extreme learning machine, is the variance estimator of the batch samples input to the extreme learning machine, is the abnormal sensitivity parameter, 3 hours can be detected Preferably, Set to 2.

[0072] The hierarchical manifold learning constraint captures the global and local structures of bridge load-bearing monitoring vibration data at different levels, enabling the model to more comprehensively understand the manifold characteristics of the data, thereby improving its adaptability to complex bridge load-bearing monitoring vibration data. The calculation method is expressed as:

[0073] ,

[0074] in, represents the matrix trace operation, The output layer weights of the extreme learning machine are used to project the feature representation of the hidden layer into the target space. By imposing global, local, and spatial constraints on it, the manifold structure of the data is maintained. is the global graph Laplacian matrix, constructed using the Isometric Feature Mapping (ISOMAP) algorithm, which preserves the global geodesic distance structure of the dataset, and , is the global adjacency matrix, is the diagonal matrix; is the local graph Laplacian matrix, constructed by Local Linear Embedding (LLE), maintaining local linear relationships, and , is the local reconstruction weight matrix, is the identity matrix; is the first layer constraint coefficient, is the second layer constraint coefficient, is the third layer constraint coefficient, satisfying ; For the and The spatial distance (in meters) between the sensors corresponding to each sample. is the sensor network topology adjustment factor, and .

[0075] S43. Perform data forward propagation calculations for the extreme learning machine: During the training process of the extreme learning machine, data forward propagation calculations involve the hidden layer output and the output layer weight update. The calculation method for the hidden layer output is expressed as:

[0076] ,

[0077] in, represents the Hadamard product, express activation function, is the layer normalization operation, is the hidden layer weight matrix of the extreme learning machine, is the hidden layer input of the extreme learning machine, is the hidden layer bias vector of the extreme learning machine, is the Sigmoid activation function, is the hidden layer output of the extreme learning machine, is the dynamic adjacency matrix, is a fusion feature; the fusion feature is constructed by tensor splicing the hidden layer input of the extreme learning machine and the time-frequency feature to enhance the model's ability to capture the transient characteristics of the vibration signal, and is expressed as:

[0078] ,

[0079] in, For The frequency domain energy entropy obtained after discrete wavelet transform is For The time domain sliding window standard deviation, Represents a tensor concatenation operation. The dynamic adjacency matrix explicitly models the spatial constraints of the sensor network and the time-frequency coupling characteristics of the vibration signal through the neighborhood relationship matrix weights. The calculation method of its elements is expressed as:

[0080] ,

[0081] in, is the first Rank Column elements, It is by and The neighborhood relationship matrix weight of samples, For the and The spatial distance (in meters) between the sensors corresponding to each sample. is the sensor network topology adjustment factor, and .

[0082] Furthermore, the weight update method of the output layer of the extreme learning machine is expressed as:

[0083] ,

[0084] ,

[0085] in, express The transpose of represents the adjustment matrix, represents the inverse matrix of the adjustment matrix, represents the identity matrix, represents the regularization coefficient, represents the target label matrix, represents the output layer weight of the extreme learning machine; the regularization coefficient is dynamically adjusted according to the local density of the output of the hidden layer of the extreme learning machine to balance the generalization requirements of regions with different manifold complexity and avoid overfitting. The calculation method is expressed as:

[0086] ,

[0087] Where, is the information entropy of the hidden layer output, is the maximum entropy value of the training set samples, is the initial regularization coefficient. Preferably, Set to 0.2.

[0088] S44. Perform error back propagation calculation of the extreme learning machine: Update the hidden layer weight matrix and bias vector of the extreme learning machine based on error back propagation, and use the gradient descent method to calculate the multi-objective loss function of the extreme learning machine with respect to the weight matrix The gradient of and the multi-objective loss function of the extreme learning machine with respect to the bias vector The gradient of , the learning rate of extreme learning machine training is set to 0.01;

[0089] S45. The extreme learning machine gradually updates the hidden layer weight matrix and bias vector through multiple iterations until the loss function converges to the preset minimum value.

[0090] S5. Preprocess the newly collected bridge load-bearing monitoring vibration data and input it into the trained model to obtain the model output and determine whether there are any abnormalities in the bridge load-bearing monitoring vibration data;

[0091] Specifically, S51. performing the same preprocessing operations as the training data on the newly collected bridge load-bearing monitoring vibration data, including data cleaning, denoising, and standardization;

[0092] S52. Input the preprocessed data into the trained model and perform forward propagation calculations to obtain the model output. Based on the model output and the set threshold or classification criteria, determine whether the bridge load-bearing monitoring vibration data contains anomalies. For example, if the probability value output by the model (e.g., the probability value for abnormal data is 0.8) is greater than the set threshold (e.g., 0.5), the data is considered abnormal and may contain a fault or safety hazard. Otherwise, the data is considered normal.

[0093] S6. Feedback the abnormal identification results and monitoring indicators to the knowledge graph system to provide auxiliary decision-making, such as Figure 3 shown.

[0094] Specifically, the anomaly identification results and monitoring indicators are fed back to the knowledge graph system (natural language processing knowledge graph library). Through the knowledge graph system, relevant structural health assessment knowledge, historical maintenance records, and multi-source heterogeneous data (such as bridge design parameters, material fatigue properties, and environmental corrosion maps) are obtained to build a decision support chain based on semantic reasoning. The specific process includes:

[0095] S61. Knowledge Fusion: Align the real-time abnormal features with the entity relationships in the knowledge graph through graph embedding, and calculate the similarity between abnormal events and historical cases through the graph attention network.

[0096] S62. Decision generation: Decision generation is performed based on real-time abnormal features and entity relationships in the knowledge graph.

[0097] For example, to monitor the health of a bridge, sensors installed at key locations on the bridge collect vibration data. This data can then be analyzed using an extreme learning machine model to perform fault diagnosis.

[0098] Assume that bridge vibration data has been obtained over a period of time. After training with an extreme learning machine model, the model outputs a probability value to indicate the degree of abnormality in the data. For example, a probability value of 0.8 indicates that the data is likely abnormal (for example, there may be problems such as cracks or corrosion, which can be determined based on manual experience or by classifying different fault types according to different abnormality probabilities).

[0099] Furthermore, real-time abnormal features (such as vibration amplitude and classification results of the extreme learning machine) are aligned with the entity relationships in the knowledge graph through graph embedding; for example: real-time abnormal feature: vibration amplitude (such as 10mm), classification result of the extreme learning machine (0.8 indicates abnormality), the vibration amplitude is calculated by signal processing.

[0100] Furthermore, it is assumed that the entity relationships in the knowledge graph are: component-stress association (such as the relationship between beam components and the stress state of the beam); sensor-location topology (such as the relationship between sensors and different locations of the bridge).

[0101] The graph attention network calculates similarity between real-time data (vibration amplitude, classification results) and similar events in historical cases (such as vibration data from bridge cracks or corrosion). For example, if a historical case study shows that a crack has grown when the vibration amplitude reaches 10mm at a certain location, the graph attention network can calculate the similarity based on the characteristics of the historical case and the current abnormal event (vibration amplitude and classification results). A high similarity indicates that the current data may be similar to a historical crack event, suggesting that the current bridge structure may have cracks or other faults.

[0102] Furthermore, decisions are made based on the calculated similarity. For example, for crack-type anomalies, the remaining service life can be predicted by calling the material stress-life curve database and combining it with the current load spectrum, and a graded warning (for example, yellow / orange level warning) can be generated.

[0103] Example 2

[0104] In this embodiment, the performance changes of the model under noise interference of different intensities are analyzed experimentally, and the noise resistance of traditional extreme learning machines, support vector machines and other methods are compared. Figure 4 As shown in the figure, as the noise level increases, this technology effectively suppresses the feature distortion caused by noise by integrating the adaptive weight mechanism of time-frequency features with the manifold topology maintenance strategy. The decrease in accuracy is significantly smaller than that of other methods, which shows the advantage of the dual denoising mechanism in this technology. It not only filters high-frequency noise through time-frequency analysis at the signal processing level, but also uses manifold learning to maintain the essential structure of the data in the feature space, thereby maintaining reliable recognition capabilities in strong interference environments.

[0105] Example 3

[0106] In this embodiment, Figure 5 As shown in the figure, the impact of the regularization coefficient on the model's generalization ability was experimentally analyzed. By observing the changes in training loss and validation accuracy under different regularization strengths, it was shown that the dynamic adjustment strategy adopted by this technology enables the model to fully learn data characteristics while avoiding overfitting. Compared with the traditional method of fixed regularization coefficients, this technology adaptively adjusts the constraint strength by using the hidden layer information entropy, allowing the model to automatically balance fitting ability and generalization requirements in different data density areas, thereby achieving better generalization performance in complex vibration signal processing.

[0107] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data, characterized in that: The following steps are involved: S1. Collect bridge load-bearing monitoring vibration data by installing multiple vibration sensors on the bridge structure; S2. Manually label bridge load-bearing monitoring vibration data; S3. Preprocessing the labeled bridge load-bearing monitoring vibration data to obtain preprocessed bridge load-bearing monitoring vibration data; the preprocessing steps include data cleaning, denoising, and standardization; S4. Construct a model for identifying anomalies in bridge load-bearing monitoring vibration data. This model uses an extreme learning machine model based on a multi-objective loss function. The loss function incorporates topology preservation loss and generalization loss, and dynamically adjusts the weights of abnormal samples through an adaptive sample weight allocation mechanism. The preprocessed bridge load-bearing monitoring vibration data is input into the model and trained to obtain a trained model. The multi-objective loss function serves as the total loss function of the extreme learning machine training process, including classification loss, topology preservation loss, and generalization loss. The classification loss is defined as the cross entropy loss; the topology preservation loss is minimized by minimizing the topology preservation loss, which is expressed as follows: ， in, represents the topology preservation loss, and They represent the output of the hidden layer and samples, Indicates the and The neighborhood relationship matrix weight of the sample represents the and The neighborhood relationship matrix weight is calculated by combining the spatiotemporal correlation and local manifold density characteristics of the bridge load-bearing monitoring vibration data, and an adaptive bimodal similarity measurement method is used for calculation. The product of the feature space Gaussian kernel and the spatiotemporal attenuation function is used to simultaneously capture the nonlinear characteristics of the bridge load-bearing monitoring vibration data and the physical constraints of the monitoring system. The generalization loss is defined by regularizing the output layer weights of the model, and is expressed as follows: ， in, represents the generalization loss, Represents the regularization coefficient, which controls the proportion of generalization loss in the overall loss function. represents the output layer weight of the extreme learning machine, represents the L2 norm, represents the adaptive sample weight loss, which automatically assigns a weight to each sample by analyzing the time domain and frequency domain characteristic attributes of the bridge load-bearing monitoring vibration data, so that the model focuses on the bridge load-bearing monitoring vibration data at abnormal moments; Representing a multi-level manifold learning constraint, the hierarchical manifold learning constraint captures the global and local structure of bridge load-bearing monitoring vibration data at different levels, enabling the model to fully understand the manifold characteristics of the data; S5. Preprocess the newly collected bridge load-bearing monitoring vibration data and input it into the trained model to obtain the model output and determine whether there are any abnormalities in the bridge load-bearing monitoring vibration data; S6. Feedback the anomaly identification results and monitoring indicators to the knowledge graph system to provide auxiliary decision-making.

2. The bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data according to claim 1 is characterized in that: Step S1 specifically includes: The vibration sensors are arranged at key positions of the bridge, including supports, bridge deck and beam ends; the vibration sensors regularly collect acceleration or strain information of the bridge under load.

3. The bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data according to claim 2 is characterized in that: Step S4 specifically includes: S41 initializes the parameters in the extreme learning machine model, including the hidden layer weight matrix and bias vector of the extreme learning machine. The initialization method is random initialization, and the initialized parameters obey the normal distribution with mean 0 and variance as the unit matrix; S42. Calculate the multi-objective loss function of the extreme learning machine: During the training process, a multi-objective collaborative optimization method is used to adjust and optimize the model parameters by optimizing multiple objective functions; S43. Perform data forward propagation calculations for the extreme learning machine: During the training process of the extreme learning machine, data forward propagation calculations involve the hidden layer output and the output layer weight update. The calculation method for the hidden layer output is expressed as: ， in, represents the Hadamard product, express activation function, is the layer normalization operation, represents the hidden layer weight matrix of the extreme learning machine, represents the hidden layer input of the extreme learning machine, represents the hidden layer bias vector of the extreme learning machine, represents the Sigmoid activation function, represents the hidden layer output of the extreme learning machine, represents the dynamic adjacency matrix, represents the fusion feature; the weight update method of the output layer of the extreme learning machine is expressed as: ， ， in, express The transpose of represents the adjustment matrix, represents the inverse matrix of the adjustment matrix, represents the identity matrix, represents the regularization coefficient, represents the target label matrix, represents the output layer weight of the extreme learning machine; S44. Perform error back propagation calculation of the extreme learning machine: Update the hidden layer weight matrix and bias vector of the extreme learning machine based on error back propagation, and use the gradient descent method to calculate the multi-objective loss function of the extreme learning machine with respect to the weight matrix The gradient of and the multi-objective loss function of the extreme learning machine with respect to the bias vector gradient; S45. The extreme learning machine gradually updates the hidden layer weight matrix and bias vector through multiple iterations until the loss function converges to the preset minimum value.

4. The bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data according to claim 3 is characterized in that: Step S5 specifically includes: S51. Perform the same preprocessing operations as the training data on the newly collected bridge load-bearing monitoring vibration data, including data cleaning, denoising, and standardization; S52. Input the preprocessed data into the trained model and perform forward propagation calculation to obtain the output result of the model; based on the output result of the model and in combination with the set threshold or classification standard, determine whether there is any abnormality in the bridge load-bearing monitoring vibration data.

5. The bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data according to claim 4 is characterized in that: In step S6, the anomaly identification results and monitoring indicators are fed back to the knowledge graph system. The knowledge graph system obtains relevant structural health assessment knowledge, historical maintenance records, and multi-source heterogeneous data to build a decision support chain based on semantic reasoning. The specific process includes: S61. The knowledge graph system is a natural language processing knowledge graph library; real-time abnormal features are aligned with entity relationships in the knowledge graph through graph embedding, and the similarity between abnormal events and historical cases is calculated through a graph attention network; S62. Make decisions based on real-time anomaly features and entity relationships in the knowledge graph.

6. The bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data according to claim 5 is characterized in that: The dynamic adjacency matrix explicitly models the spatial constraints of the sensor network and the time-frequency coupling characteristics of the vibration signal through the neighborhood relationship matrix weights; the fusion feature is constructed by tensor splicing the hidden layer input of the extreme learning machine and the time-frequency features.

7. The bridge load-bearing monitoring data identification method based on knowledge-driven and monitoring data according to claim 6 is characterized in that: The regularization coefficient is dynamically adjusted according to the local density output by the extreme learning machine hidden layer.

Images