Dynamic deformation monitoring data real-time processing and analysis method

Abstract

The invention provides a dynamic deformation monitoring data real-time processing and analysis method, and relates to the technical field of data processing. The dynamic deformation monitoring data real-time processing and analysis method specifically comprises the following steps: S1, deploying a multi-type sensor network, collecting displacement, acceleration and inclination angle data in real time, and performing time synchronization and data preprocessing through a wireless / wired transmission protocol; s2, mapping the multi-dimensional monitoring data into a high-dimensional space-time trajectory, and generating a visual trajectory diagram by adopting a dimension reduction algorithm; s3, segmenting the track based on a change point detection algorithm, and extracting curvature, speed and acceleration change rate characteristics of each segment; s4, constructing a dynamic behavior pattern library, and performing real-time pattern matching and switching detection in combination with a hidden Markov model and a support vector machine; and S5, identifying abnormal points through a 3 sigma criterion and a multi-model voting mechanism, evaluating a risk grade according to a mode switching frequency, and triggering graded early warning. According to the invention, multi-source cooperation is realized, and early warning precision is high.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and specifically to a method for real-time processing and analysis of dynamic deformation monitoring data. Background Art

[0002] A bridge is a large building that spans obstacles. It is erected over rivers, lakes, seas, or valleys, transportation lines, etc., enabling vehicles, pedestrians, etc. to pass smoothly. It is very necessary to conduct dynamic deformation monitoring on bridges. First of all, during the long-term use of bridges, they will be affected by various natural factors, such as earthquakes, wind loads, floods, etc. These factors will cause dynamic responses and deformations in the bridge structure. Through dynamic deformation monitoring, the structural safety status of the bridge under different environments can be grasped in real time, and potential safety hazards can be discovered in time. Secondly, the continuous change of traffic load is also an important factor. The dynamic effects such as vibration and impact force generated during vehicle driving will cause deformations and fatigue damages in the bridge structure. Monitoring can understand the performance of the bridge under different traffic flows and load patterns, providing a scientific basis for the maintenance, reinforcement, and management of the bridge. Moreover, the aging of the bridge's own materials and the degradation of its structural performance are inevitable. Dynamic deformation monitoring helps to evaluate the health status and service life of the bridge structure, and take measures in advance to ensure the safe operation of the bridge and avoid major safety accidents.

[0003] Existing methods for processing dynamic deformation monitoring data often rely on a single sensor, with limited data dimensions and difficult to comprehensively reflect the state of the monitoring object. Data processing is mostly carried out centrally in the cloud, with large transmission delays and poor real-time performance. There are lack of effective means for trajectory feature extraction and pattern recognition, making it difficult to accurately capture the dynamic changes of the structure. The anomaly detection method is vulnerable to noise interference and has a high false alarm rate. The early warning mechanism is simple and cannot accurately give early warnings according to different risk levels. The visualization degree is low, and the analysis results are not intuitive enough, which is not conducive to engineers making quick decisions. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a method for real-time processing and analysis of dynamic deformation monitoring data, which solves the problem of low detection efficiency of the prior art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for real-time processing and analysis of dynamic deformation monitoring data specifically includes the following steps:

[0006] S1. Deploy a multi-type sensor network to collect displacement, acceleration, and inclination data in real time, and perform time synchronization and data preprocessing through wireless / wired transmission protocols;

[0007] S2. Map the multi-dimensional monitoring data into a high-dimensional spatio-temporal trajectory, and use a dimensionality reduction algorithm to generate a visualized trajectory map;

[0008] S3. Segment the trajectory based on the change point detection algorithm, and extract the curvature, speed, and acceleration change rate features of each segment;

[0009] S4. Construct a dynamic behavior pattern library, and perform real-time pattern matching and switching detection by combining the hidden Markov model and the support vector machine;

[0010] S5. Identify abnormal points through the 3σ criterion and the multi-model voting mechanism, evaluate the risk level according to the pattern switching frequency, and trigger a hierarchical early warning;

[0011] S6. Use an interactive visualization tool to dynamically display the trajectory evolution and pattern switching points, and generate a structured monitoring report.

[0012] Preferably, in step S1, the sensor network uses edge computing devices for data preprocessing, including noise filtering, missing value interpolation, and normalization operations, where the normalization formula is:

[0013]

[0014] Preferably, the trajectory curvature calculation in step S3 uses the differential geometry formula:

[0015]

[0016] Realize the quantification of bending features.

[0017] Preferably, the change point detection in step S3 uses the PELT algorithm, by minimizing the cost function:

[0018]

[0019] Determine the trajectory segmentation boundary, where the penalty term β = 2log(n), and n is the total number of current data points.

[0020] Preferably, the pattern matching in step S4 uses an incremental learning method to update the pattern library, and the model parameter update formula is:

[0021]

[0022] Realize dynamic adaptation to new monitoring scenarios.

[0023] Preferably, the multi-level early warning mechanism in step S5 includes:

[0024] Trigger a yellow early warning when the gradual state lasts for more than the preset duration threshold;

[0025] Trigger a red early warning when the mutation state frequency exceeds the preset critical value or the pattern switches disorderly.

[0026] Preferably, the visualization tool in step S6 supports three-dimensional dynamic playback of trajectories and maps the curvature distribution characteristics through a heat map, where the heat map is generated using a kernel density estimation algorithm, preferably a Gaussian kernel function:

[0027]

[0028] Perform curvature distribution rendering.

[0029] Preferably, the method is deployed in an edge-cloud collaborative architecture, where the edge side executes steps S1 - S3, and the cloud side executes in-depth analysis and pattern training of steps S4 - S6.

[0030] Preferably, the hidden Markov model in step S4 passes through a state transition probability matrix:

[0031] A = [a ij

[0032] Observation probability:

[0033] b j (o t )

[0034] The present invention provides a method for real-time processing and analysis of dynamic deformation monitoring data. It has the following beneficial effects:

[0035] The present invention provides a method for real-time processing and analysis of dynamic deformation monitoring data. This technology deploys multiple types of sensor networks to collect multi-dimensional data, which can comprehensively reflect the state of the monitoring object. It adopts an edge-cloud collaborative architecture, where the edge side preprocesses the data, reducing the transmission volume, reducing latency, and improving real-time performance. Through change point detection and feature quantification, it can accurately extract trajectory features, combines a hidden Markov model with a support vector machine for pattern matching, has a high recognition accuracy, and a multi-level early warning mechanism can accurately give early warnings according to the risk level. The visualization tool dynamically displays the trajectory evolution and pattern switching, and the analysis results are intuitive, facilitating engineers to make quick decisions. Specific embodiments

[0036] Next, in combination with the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0037] The embodiments of the present invention provide a method for real-time processing and analysis of dynamic deformation monitoring data, which specifically includes the following steps:

[0038] Multi-sensor data collection and preprocessing

[0039] Deploy accelerometers, inclinometers and laser displacement sensors at key positions of the bridge, and achieve time synchronization through the ZigBee protocol.

[0040] Edge computing preprocessing:

[0041] y t = median(x t-2 , x t-1 , x t , x t+1 , x t+2

[0042] For data with less than 3 consecutive missing points, use linear interpolation to complete it, scale the data to the [0, 1] interval according to the weight 2 formula, and perform normalization.

[0043] Space-time trajectory mapping and dimensionality reduction

[0044] Trajectory modeling:

[0045] Fuse displacement, acceleration, and inclination data into a six-dimensional space-time trajectory

[0046] T(t) = [x(t), y(t), z(t), a x (t), a y (t), θ(t)

[0047] Dimensionality reduction visualization:

[0048] Use the t-SNE algorithm (learning rate 200, number of iterations 1000) to project the six-dimensional data onto a two-dimensional plane and generate a dynamic trajectory map.

[0049] Change point detection and feature quantization

[0050] Change point segmentation:

[0051] Adopt the PELT algorithm with weight 4, set the penalty term β = 2log(n) (n is the current number of data points), and segment the trajectory into three types of segments: stable state, gradual change state, and mutation state.

[0052] Curvature calculation:

[0053] Calculate the curvature of each segment according to the weight 3 formula, and extract the mean and variance of the curvature as bending features.

[0054] Dynamic pattern recognition and model update

[0055] Pattern library construction:

[0056] The initial pattern library contains three types of behavior patterns (normal vibration, structural micro-damage, local instability), and each type stores 100 groups of historical trajectory feature vectors.

[0057] HMM pattern matching

[0058] The state transition matrix A = [a ij is initialized to a uniform distribution, and the observation probability b j (o t ) follows a Gaussian distribution N(μ j , σ j 2 ).

[0059] The hidden state sequence corresponding to the current trajectory is decoded in real time through the Viterbi algorithm.

[0060] Incremental learning

[0061] New data is collected weekly, and the model parameters are updated according to Equation 5. The misreported data labeled by the user is retrained through an SVM classifier (kernel function RBF, C = 1.0).

[0062] Multi-level warning trigger

[0063] Anomaly detection:

[0064] Apply the 3σ criterion to the feature vector. If it exceeds the range of the mean ± 3 standard deviations, it is determined as an anomaly. Combining the voting results of HMM and SVM, the anomaly is confirmed when the two models are consistent.

[0065] Warning logic:

[0066] Yellow warning: The gradual change state lasts for more than the preset threshold, which is 30 minutes by default;

[0067] Red warning: The mutant state is triggered more than 5 times per hour, or the disorder degree of mode switching H = -∑p i logp i > 1.5.

[0068] Visualization and report generation

[0069] The three-dimensional dynamic playback of the trajectory is realized by using WebGL technology, supporting zooming, rotation, and time-axis dragging. According to the Gaussian kernel density estimation formula of Equation 7. Render the heat map of curvature distribution (the color scale is mapped to blue-yellow-red). Finally, a PDF report is automatically generated, including the anomaly event timestamp, mode switching statistics, and risk level assessment conclusion.

[0070] At the data acquisition level, this invention deploys multi-type sensor networks, which can collect multi-dimensional data such as displacement, acceleration, and inclination in real time. Compared with a single sensor, it can comprehensively and accurately reflect the actual state of the monitoring object. In terms of the processing architecture, an edge-cloud collaborative architecture is adopted. The edge side performs data preprocessing, greatly reducing the data transmission volume, effectively reducing latency, improving the real-time performance of the system, and enabling a faster response to the monitoring situation. In terms of feature extraction and pattern recognition, the trajectory is segmented through a change point detection algorithm, and features such as curvature, speed, and acceleration change rate are extracted. Combining the hidden Markov model and the support vector machine for pattern matching and switching detection can accurately capture the dynamic changes of the structure. Anomaly detection uses the 3σ criterion and a multi-model voting mechanism, effectively reducing the false alarm rate and ensuring the reliability of early warnings. The multi-level early warning mechanism triggers hierarchical early warnings according to different risk levels. For example, a yellow early warning is triggered when the gradual change state continuously exceeds the threshold, and a red early warning is triggered when the mutation state has a too high frequency or the pattern switches disorderly, enabling accurate early warnings. The visualization tool supports three-dimensional dynamic playback of the trajectory and the display of the curvature heat map, making the analysis results intuitive and easy to understand, facilitating engineers to quickly locate the risk source and make decisions. In addition, the incremental learning mechanism can dynamically adapt to new monitoring scenarios, enabling the model to maintain a high accuracy rate in different environments.

[0071] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for real-time processing and analysis of dynamic deformation monitoring data, characterized in that: The specific steps include: S1. Deploy multi-type sensor networks to collect displacement, acceleration and inclination data in real time, and perform time synchronization and data preprocessing through wireless / wired transmission protocols; S2. Map the multi-dimensional monitoring data into high-dimensional spatiotemporal trajectories and generate a visual trajectory map using a dimensionality reduction algorithm; S3. Segment the trajectory based on the change point detection algorithm and extract the curvature, velocity and acceleration change rate characteristics of each segment; S4. Build a dynamic behavior pattern library and combine hidden Markov model and support vector machine for real-time pattern matching and switching detection; S5. Identify outliers through the 3σ criterion and multi-model voting mechanism, assess risk levels based on mode switching frequency, and trigger graded warnings; S6. Use interactive visualization tools to dynamically display trajectory evolution and mode switching points and generate structured monitoring reports.

2. A method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: In step S1, the sensor network uses edge computing devices to perform data preprocessing, including noise filtering, missing value interpolation and normalization operations, where the normalization formula is:

3. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The trajectory curvature calculation in step S3 adopts the differential geometry formula: Realizes quantification of curved features.

4. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The change point detection in step S3 adopts the PELT algorithm by minimizing the cost function: Determine the trajectory segmentation boundary, where the penalty term β = 2log(n), n is the total number of current data points.

5. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The pattern matching in step S4 adopts an incremental learning method to update the pattern library, and the model parameter update formula is: Achieve dynamic adaptation to new monitoring scenarios.

6. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The multi-level early warning mechanism of step S5 includes: When the gradual change lasts longer than the preset time threshold, a yellow warning is triggered; A red warning is triggered when the mutation state frequency exceeds the preset critical value or the mode switches disorderly.

7. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The visualization tool of step S6 supports three-dimensional dynamic playback of the trajectory and maps the curvature distribution characteristics through a heat map, wherein the heat map is generated using a kernel density estimation algorithm, preferably a Gaussian kernel function: Perform curvature distribution rendering.

8. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The method is deployed in an edge-cloud collaborative architecture, with the edge executing steps S1-S3 and the cloud executing deep analysis and pattern training of steps S4-S6.

9. The method for real-time processing and analysis of dynamic deformation monitoring data according to claim 1, characterized in that: The hidden Markov model of step S4 is provided by the state transition probability matrix: A=[a ij Observation probability: b j (o t )。

Images