Bridge BIM monitoring data visualization method and device

Abstract

The invention provides a bridge BIM (Building Information Modeling) monitoring data visualization method and device. The method comprises the following steps: acquiring time sequence data including strain, displacement, vibration frequency and environment temperature and humidity parameters through a bridge monitoring sensor; an improved ISABO algorithm is provided to carry out hyper-parameter optimization on a BiLSTM model, an ISABO-BiLSTM filling model is constructed, a missing value of time series data is predicted, and complete data after filling is stored in an SQL database. The method is characterized in that a Sprite elf model with a unique ID identifier is created innovatively based on a WebGL engine and a Three.js technology, user interaction is realized through a Raycaster ray detection mechanism, and associated monitoring data and filling data can be extracted by clicking the elf model. And when the data exceed a safety threshold set based on the bridge design specification, triggering a BIM model color grading alarm. The invention further relates to a visualization device, electronic equipment, a computer readable storage medium and a computer program product, and efficient visualization and safety early warning of bridge monitoring data are achieved.

Description

Technical Field

[0001] The present invention relates to monitoring technologies, and particularly to a method and device for visualizing bridge BIM monitoring data. Background Art

[0002] Currently, methods based on machine learning and deep learning are widely used in the analysis and prediction of bridge monitoring data. In existing bridge monitoring technologies, common methods include models based on machine learning and deep learning, such as LSTM, GRU, etc. However, these methods have the following problems: hyperparameter optimization relies on manual tuning, resulting in insufficient model prediction accuracy; when sensor data is missing, the existing filling methods have limited effects, affecting the reliability of structural assessment; the visualization means are single, unable to display monitoring data in real time dynamically, and the interaction response speed is slow, etc.

[0003] Therefore, there is an urgent need for a bridge monitoring method that can combine advanced optimization algorithms and deep learning models to improve data processing and prediction capabilities, and at the same time has an efficient visualization display function. This method should be able to effectively solve the problem of data missing, optimize the model performance, and help engineering personnel to grasp the structural state of the bridge in real time through intuitive visualization means, thereby improving the safety management level of the bridge. Summary of the Invention

[0004] This application provides a method and device for visualizing bridge BIM monitoring data to solve problems such as insufficient hyperparameter optimization, limited effect of missing value filling, and single visualization means in the prior art.

[0005] An embodiment of the first aspect of the present application provides a method for visualizing bridge BIM monitoring data, including the following steps: S1: Collect time series data through bridge monitoring sensors; S2: An improved ISABO algorithm is proposed, and based on the improved ISABO algorithm, hyperparameter optimization is performed on the learning rate, the number of hidden layer nodes, and the regularization coefficient of the BiLSTM model to obtain an ISABO-BiLSTM filling model; S3: Input the time series data into the filling model for missing value prediction, and store the filled complete data set in an SQL database; S4: Based on the sensor spatial coordinate data in the SQL database, an innovative Sprite sprite model with an ID identifier is created through the WebGL engine, and a mapping relationship between the ID and the database primary key is established; S5: Listen for user interaction events based on the Raycaster ray detection mechanism of Three.js. When the user clicks on the Sprite sprite, the associated monitoring data and filling data are extracted from the SQL database through the mapping relationship; S6: According to the monitoring data and filling data, compare them with a preset safety threshold range. When the data exceeds the threshold, trigger a BIM model color grading alarm, where the structural safety threshold range is set based on the material strength limit and deformation tolerance value in the bridge design specification.

[0006] Preferably, the sensors in the bridge monitoring sensors are deployed at the key monitoring points of the main beam, bridge tower, and cable stay of the bridge, and the collected time series data includes strain, displacement, vibration frequency, and environmental temperature and humidity parameters.

[0007] Preferably, the mechanism of the improved ISABO algorithm includes: S2.1.1: Randomly initialize the population individuals to obtain the initial positions of the individuals. The formula is:

[0008]

[0009] where Xi is the i-th population individual, N is the population size, m is the number of variables, ri,d is a random number between [0, 1], and ubd and lbd are the upper and lower bounds of the variables respectively. S2.1.2: Adjust the factor weights of the SABO subtraction operator to obtain the subtraction operator formula of the ISABO algorithm as:

[0010]

[0011] S2.1.3: Update the individual position through the leader learning strategy using the formula:

[0012]

[0013] where is the new position of the i-th population individual, Xbest is the optimal individual in the iteration, It is normally distributed between [0, 1];

[0014] S2.1.4: Reset the individuals that exceed the limit. The reset formula is:

[0015]

[0016] where Xi is the i-th individual in the population, ri,d are random numbers between [0, 1], and ubd and lbd are the upper and lower bounds of the variable respectively.

[0017] S2.1.5: Determine the positions of the obtained individuals to decide whether to update the positions of the individuals in the overall population in this round. The position evaluation formula is:

[0018]

[0019] where, and Fi are and the objective function values of Xi.

[0020] Preferably, the hyperparameter optimization of the BiLSTM model includes: S2.2.1: Perform CEEMDAN decomposition on the original time series data, remove the high-frequency noise components, and then reconstruct the input data; S2.2.2: Divide the reconstructed data into a training set and a test set at a ratio of 7:3, and use the root mean square error as the fitness function; S2.2.3: Search for the optimal learning rate lr ∈ [0.001, 0.1], the number of hidden layer nodes H ∈ [50, 200], and the regularization coefficient λ ∈ [0.01, 0.1] in the solution space through the ISABO algorithm.

[0021] Preferably, creating a Sprite sprite model with a unique ID identifier through the WebGL engine includes: S4.1.1: Obtain the sensor space coordinates through Revit secondary development, and export the lightweight BIM model to the Web side using the gltfRevitExport plugin; S4.1.2: Generate a 2D sprite texture carrying the ID based on the SpriteMaterial of Three.js, and the texture content includes the sensor type icon and the status identifier.

[0022] Preferably, the Raycaster ray detection mechanism based on Three.js includes: S5.1.1: Obtain the ID of the clicked sprite through the ray collision detection with the bounding box of the Sprite sprite; S5.1.2: Send an HTTP request to the server based on the ID to query the associated monitoring historical data and the timestamp of the filled data; S5.1.3: Use ECharts to render the query result as a time series curve graph with marked points, where the filled data is highlighted in orange.

[0023] In the second aspect of the embodiments of the present application, a bridge BIM monitoring data visualization device is provided, including: a data acquisition module, S1: collecting time-series data through bridge monitoring sensors; an algorithm optimization module, S2: proposing an improved ISABO algorithm, and based on the improved ISABO algorithm, performing hyperparameter optimization on the learning rate, the number of hidden layer nodes, and the regularization coefficient of the BiLSTM model to obtain an ISABO-BiLSTM filling model; a storage module, S3: inputting the time-series data into the ISABO-BiLSTM filling model for missing value prediction, and storing the filled complete data set in an SQL database; a model construction module, S4: based on the sensor spatial coordinate data in the SQL database, creating a Sprite sprite model with a unique ID identifier through a WebGL engine, and establishing a mapping relationship between the ID and the database primary key to replace the traditional BIM family model; an interaction module, S5: listening for user interaction events based on the Raycaster ray detection mechanism of Three.js, and when the user clicks on the Sprite sprite, extracting associated monitoring data and filling data from the SQL database through the mapping relationship; a decision-making module, S6: comparing the monitoring data and the filling data with a preset structural safety threshold interval, and when the data exceeds the threshold interval, triggering a BIM model color grading alarm, where the structural safety threshold interval is set based on the material strength limit and the deformation tolerance value in the bridge design specification.

[0024] In the third aspect of the embodiments of the present application, an electronic device is provided, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the program to implement a bridge BIM monitoring data visualization method as described in the above embodiments.

[0025] In the fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and the program is executed by a processor to implement a bridge BIM monitoring data visualization method as described in the above embodiments.

[0026] In the fifth aspect of the embodiments of the present application, a computer program product is provided, including a computer program or instruction to implement a bridge BIM monitoring data visualization method as described in the above embodiments.

[0027] Therefore, the present application includes the following beneficial effects:

[0028] In the embodiments of the present application, the hyperparameters of the BiLSTM model are optimized by the improved ISABO algorithm, effectively enhancing the prediction and filling capabilities of the model when processing complex time-series data of bridges, significantly reducing the impact of data missing on the monitoring results, ensuring the integrity and reliability of the monitoring data. The learning rate, the number of hidden layer nodes, and the regularization coefficient are optimized using the ISABO-BiLSTM model, ensuring that the model can maintain excellent performance in different monitoring scenarios, enhancing the generalization ability and adaptability of the model, and being applicable to various types of bridge monitoring data. Based on the preset structural safety threshold range, when the monitoring data exceeds the safety range, the color grading alarm of the BIM model is automatically triggered, intuitively indicating potential structural safety hazards, improving the response speed and emergency handling ability of bridge safety monitoring, and ensuring the safe operation of the bridge. The complete dataset after filling is efficiently stored in the SQL database, and the mapping relationship between the ID and the database primary key is established, ensuring the orderly management and rapid retrieval of data, facilitating subsequent data analysis and decision support.

[0029] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the accompanying drawings, where:

[0031] Figure 1 is a flowchart of a method for visualizing bridge BIM monitoring data according to an embodiment of the present application.

[0032] Figure 2 is a flowchart of a method for visualizing bridge BIM monitoring data according to an embodiment provided by an embodiment of the present application.

[0033] Figure 3 is the association between the BIM monitoring area and the Sprite sensor according to an embodiment of the present application.

[0034] Figure 4 is a flowchart for constructing an ISABO-optimized BiLSTM data filling model according to an embodiment of the present application.

[0035] Figure 5 is a flowchart for applying the ISABO-BiLSTM data filling model according to an embodiment of the present application.

[0036] Figure 6 is a schematic diagram of Sprite sensor listening according to an embodiment of the present application.

[0037] Figure 7Schematic diagram of the CEEMDAN decomposition sequence provided according to an embodiment of the present application.

[0038] Figure 8 Schematic diagram of the change curves of the ISABO and SABO fitness values provided according to an embodiment of the present application.

[0039] Figure 9 Schematic diagram of the filling results of different models provided according to an embodiment of the present application.

[0040] Figure 10 Schematic diagram of the comparison of the sensor loading performance provided according to an embodiment of the present application.

[0041] Figure 11 Schematic diagram of the data interaction interface of the bridge BIM monitoring platform provided according to an embodiment of the present application.

[0042] Figure 12 Flow chart of the data visualization technical route of the monitoring platform provided according to an embodiment of the present application.

[0043] Figure 13 Schematic diagram of the structure of the bridge BIM monitoring data visualization device provided according to an embodiment of the present application.

[0044] Figure 14 Schematic diagram of the structure of the electronic device provided according to an embodiment of the present application. Figure 15 Schematic diagram of the process of the monitoring data visualization method provided according to an embodiment of the present application. Detailed implementation manners

[0045] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals indicate the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application, but should not be construed as limiting the present application.

[0046] The bridge BIM monitoring data visualization method and device according to the embodiments of the present application will be described below with reference to the accompanying drawings.

[0047] Specifically, Figure 1 Schematic diagram of the process of the bridge BIM monitoring data visualization method provided by the embodiments of the present application.

[0048] As Figure 1 shown, the bridge BIM monitoring data visualization method includes the following steps:

[0049] In step S1, time series data is collected through bridge monitoring sensors.

[0050] Among them, the sensor deployment adopts a spatial grid layout strategy, with the spacing between the main girder monitoring points ≤ 15 m, the vertical spacing of the bridge towers ≤ 8 m, and the equal arc length point layout method is used for the stay cables. The sampling frequency of the time series data is 200 Hz, and the data missing rate is controlled within the engineering allowable range of < 5%.

[0051] It can be understood that in the embodiment of the present application, a three-level control system of the main girder - bridge tower - stay cable (32 displacement gauges are arranged on the main girder, 18 strain gauges are arranged on the bridge tower, and 24 triaxial accelerometers are arranged on the stay cables) forms a spatial monitoring network topological structure. The vibrating wire strain gauge (range ±3000 με, resolution 1 με) and GNSS (horizontal accuracy 2 mm + 1 ppm) form complementary monitoring, and the triaxial accelerometer (frequency response 0 - 100 Hz) captures the structural vibration mode. The environmental temperature and humidity sensor (accuracy ±0.5 °C / ±2% RH) establishes a temperature - strain compensation model to eliminate the measurement deviation of ±150 με caused by the daily average temperature difference. This layout scheme has been verified by an actual bridge, and the data integrity rate has been increased to 97.3%, which is 21.5% higher than the traditional random point layout method.

[0052] Specifically, in the application of a (56 + 128 + 320 + 128 + 56) m bridge, 32 strain gauges (spacing 15 m) are arranged on the main girder, 18 sets of displacement sensors (vertical spacing 8 m) are set on the bridge tower, and 24 triaxial accelerometers (equal arc length distribution) are installed on the stay cables. When the SG25 sensor fails, the system accurately reconstructs the data of the missing period through the strain gradient data (maximum difference < 8.7 με) of the adjacent SG24 and SG26 and the displacement correlation (R 2 = 0.93) of the measuring point T7 of the bridge tower.

[0053] It should be noted that the sensor adopts a temperature compensation algorithm:

[0054] ε_corrected = ε_raw - αΔT (6)

[0055] Among them, α = 11.7×10^-6 / °C is the steel expansion coefficient. When the environmental temperature difference reaches 15 °C, a measurement deviation of 175 με can be eliminated.

[0056] In the embodiment of the present application, the sensors in the bridge monitoring sensors are deployed at the key monitoring points of the main girder, bridge tower and stay cables of the bridge, and the collected time series data includes strain, displacement, vibration frequency and environmental temperature and humidity parameters.

[0057] In step S2, an improved ISABO algorithm is proposed, and based on the improved ISABO algorithm, hyperparameter optimization is performed on the learning rate, the number of hidden layer nodes and the regularization coefficient of the BiLSTM model to obtain the ISABO - BiLSTM filling model.

[0058] Among them, the normal distribution parameter is N(0.5, 0.15), the reset threshold is set to 120% of the parameter space boundary, and a dynamic weight factor is introduced into the leader learning strategy. The formula for the dynamic weight factor is:

[0059] α = 1 - (t / T)^2 (7)

[0060] where t is the current iteration number, and T = 50 is the total number of iterations.

[0061] It can be understood that the improved ISABO algorithm in the embodiments of this application promotes population diversity to avoid local optimality by introducing a non-linear convergence factor α = 1 - (t / T)^2, maintaining a large perturbation range (α > 0.91) in the initial stage of iteration (t < 0.3T); and strengthening local search (α < 0.51) in the later stage of iteration (t > 0.7T) to improve the parameter tuning accuracy. Combining with the N(0.5, 0.15) normal distribution random numbers generated by the Box-Muller transform, the coverage rate of the hyperparameter search space is increased by 42%. Compared with the standard SABO algorithm, the learning rate optimization error of BiLSTM is reduced from ±0.008 to ±0.003, and the matching accuracy of the number of hidden layer nodes is increased to 98.7%. Through actual measurement, this algorithm can converge within 50 iterations, the training time is shortened to 1 / 15 of the traditional grid search, and the RMSE index is stable below 0.023, significantly improving the generalization ability of the model under complex working conditions.

[0062] Specifically, when optimizing the number of hidden layer nodes H of BiLSTM, the ISABO algorithm locks H = 178 as the optimal solution (the theoretical optimal value is 182) in the 28th generation, while the PSO algorithm still fluctuates in the range of [153, 209] after 50 generations. The fitness curve of ISABO tends to be stable after the 15th generation, and the convergence speed is increased by 40% compared with SABO.

[0063] It should be noted that the parameter search space is set as: learning rate lr ∈ [0.001, 0.1], number of hidden layer nodes H ∈ [50, 200], regularization coefficient λ ∈ [0.01, 0.1]. This range is determined through pre-experiments. When H > 200, the model training time increases by 320% but the accuracy only increases by 0.7%.

[0064] In the embodiments of this application, the mechanism of the improved ISABO algorithm includes: S2.1.1: Randomly initialize the population individuals to obtain the initial positions of the individuals. The formula is:

[0065] X i,d = lb d + r i,d ·(ub d - lb d ), i = 1, ···, N, d = 1, ···, m (1)

[0066] Among them, $X_i$ is the $i$-th individual in the population, $N$ is the population size, $m$ is the number of variables, $r_{i,d}$ is a random number between $[0, 1]$, and $ub_d$ and $lb_d$ are the upper and lower bounds of the variables respectively. S2.1.2: Adjust the factor weights of the SABO subtraction operator, and the subtraction operator formula of the ISABO algorithm is obtained as follows:

[0067]

[0068] S2.1.3: The individual position update formula through the leader learning strategy is:

[0069]

[0070] Among them, is the new position of the $i$-th individual in the population, $X_{best}$ is the optimal individual in the iteration, is a normal distribution between $[0, 1]$; S2.1.4: Reset the individuals that exceed the bounds, and the reset formula is:

[0071]

[0072] Among them, $X_i$ is the $i$-th individual in the population, $r_{i,d}$ is a random number between $[0, 1]$, and $ub_d$ and $lb_d$ are the upper and lower bounds of the variables respectively. S2.1.5: Determine the obtained individual positions to decide whether to update the positions of the individuals in the overall population in this round. The position evaluation formula is:

[0073]

[0074] Among them, and $F_i$ are and the objective function values of $X_i$.

[0075] In the embodiment of the present application, the hyperparameter optimization of the BiLSTM model includes: S2.2.1: Perform CEEMDAN decomposition on the original time series data, and reconstruct the input data after removing the high-frequency noise components; S2.2.2: Divide the reconstructed data into a training set and a test set according to a ratio of 7:3, and use the root mean square error as the fitness function; S2.2.3: Search for the optimal learning rate $lr\in[0.001, 0.1]$, the number of hidden layer nodes $H\in[50, 200]$, and the regularization coefficient $\lambda\in[0.01, 0.1]$ in the solution space through the ISABO algorithm.

[0076] In step S3, input the time series data into the filling model for missing value prediction, and store the filled complete data set in the SQL database.

[0077] Among them, the SQL database uses the PostgreSQL 14.5 version and configures the TimescaleDB 2.8 extension module. The data sharding strategy is hash partitioning by sensor ID, and the time dimension chunk granularity is set to 1 hour per chunk.

[0078] It can be understood that after adopting the TimescaleDB extension module of PostgreSQL in the embodiments of the present application, the database uses the adaptive time sharding technology (chunk size is 7 days) to increase the write throughput of billions of data points to 120,000 records per second, which is 8 times higher than that of MySQL. The columnar storage is combined with the ZSTD compression algorithm (compression ratio up to 10:1), reducing the storage space occupancy by 78%. In terms of query optimization, continuous aggregate pre-computation is used to calculate the hourly average value, reducing the historical data query response time from the minute level to within 300 ms. At the same time, the real-time data writing and the synchronous update of the filling results are realized through the streaming processing interface (PipelineDB), ensuring that the front-end visualization display latency is less than 500 ms, meeting the real-time requirements of engineering monitoring.

[0079] Specifically, a certain data query test shows that it takes 283 ms to retrieve 30 days of raw data (about 5.18 million records) of the SG25 sensor, while a traditional relational database takes 8.7 s. When the number of concurrent users reaches 2000, the P95 value of the query response time remains below 420 ms.

[0080] It should be noted that the database establishes a triple index: the timestamp BRIN index, the sensor ID hash index, and the data type GIN index. This structure increases the range query speed by 17 times and reduces the space complexity by 63%.

[0081] In step S4, based on the sensor spatial coordinate data in the SQL database, an innovative Sprite sprite model with an ID identifier is created through the WebGL engine, and the mapping relationship between the ID and the database primary key is established.

[0082] Among them, the mapping relationship is implemented by a Redis 7.0 cluster, and the key-value structure is "SensorID:{coordinates (x, y, z), data type, latest value}". The LRU cache policy retains the most recent 10,000 active records

[0083] It can be understood that in the embodiments of the present application, a key-value pair index table is established through the Redis in-memory database, and the LRU eviction policy is adopted to manage the ID mapping relationships with high-frequency access (cache hit rate > 95%), reducing the single ID parsing time from 120 ms to 8 ms. Combined with the Redis Cluster cluster deployment (3 masters and 3 slaves), it supports 200,000 concurrent queries per second, which is 2 orders of magnitude higher than directly accessing the SQL database. In addition, by periodically persisting the RDB snapshot (interval of 15 minutes) and the AOF log (synchronized per second), it is ensured that the integrity loss of the mapping relationship data during fault recovery does not exceed 5 seconds. This design enables the system to maintain a request response time of no more than 50 ms for 98.5% of requests even in the scenario of 2000+ user concurrent clicks, greatly improving the fluency of the interaction experience.

[0084] Specifically, in the stress test, when simulating 5000 concurrent click events: the throughput of the Redis cluster reaches 18,500 times per second, and the CPU load rate is 72%, while the throughput of directly querying the SQL database is only 920 times per second and 54% of the requests time out.

[0085] It should be noted that the key-value storage structure is designed as:

[0086] {

[0087] "SensorID":"SG25",

[0088] "Coordinates":[125.73,38.92,56.4],

[0089] "DataType":"Strain",

[0090] "LatestValue":1523 με,

[0091] "Status":"Normal"

[0092] }

[0093] In the embodiments of the present application, creating a Sprite sprite model with a unique ID identifier through the WebGL engine includes: S4.1.1: Obtaining the sensor spatial coordinates through Revit secondary development, and exporting the lightweight BIM model to the Web side using the gltfRevitExport plugin; S4.1.2: Generating a 2D sprite texture carrying the ID based on the SpriteMaterial of Three.js, and the texture content includes the sensor type icon and the status identifier.

[0094] In step S5, the Raycaster ray detection mechanism based on Three.js listens for user interaction events. When the user clicks on a Sprite, the associated monitoring data and filling data are extracted from the SQL database through the mapping relationship.

[0095] Among them, the ray detection adopts a hierarchical detection strategy: first, perform a rough AABB bounding box detection (taking 0.3 ms), and then execute an accurate triangle face collision detection (taking 1.8 ms).

[0096] It can be understood that in the embodiment of this application, based on the octree space partitioning algorithm, the system divides the three-dimensional scene into hierarchical space grids (with a depth of 8 layers). Through the fast intersection test of the ray and the scene bounding box (AABB), the computational complexity of collision detection is reduced from O(n) to O(logn). In a typical scene containing more than 5000 Sprites, the time-consuming of a single ray detection is reduced from 35 ms to 2.1 ms, making the WebGL rendering frame rate stable at 60 fps. At the same time, the frustum culling technology is adopted to only detect the Sprites within the visible area, further reducing the CPU occupancy rate to less than 12%. Combining with the WebWorker multi-threaded rendering mechanism of ECharts, the drawing delay of the time series curve graph is shortened from 220 ms to 45 ms, supporting the simultaneous display of 20 sensor curves without jamming, greatly enhancing the multi-parameter collaborative analysis ability.

[0097] Specifically, when the user clicks on the main span SG25 sensor Sprite, the system completes the collision detection within 2.3 ms, obtains the data of the most recent 24 hours (19,200 sampling points) through an HTTP GET request. After downsampling to 192 key points by the LTTB algorithm, the curve rendering is completed within 38 ms. The time-domain continuity error between the filling data points (orange) and the measured data (blue) is <0.8%.

[0098] It should be noted that the ray detection accuracy is set to 0.01 meters, and Sprites with a size of 0.5×0.5 meters can be accurately identified. When the distance between Sprites is less than 0.3 meters, the system automatically enables the multi-selection mode.

[0099] In the embodiment of this application, the Raycaster ray detection mechanism based on Three.js includes: S5.1.1: Through the collision detection of the ray and the bounding box of the Sprite, obtain the ID of the clicked Sprite; S5.1.2: Based on the ID, send an HTTP request to the server to query the time stamps of the associated monitoring historical data and filling data; S5.1.3: Use ECharts to render the query result as a time series curve graph with marked points, where the filling data is highlighted in orange.

[0100] In step S6, based on the monitoring data and the filled data, compare them with the preset safety threshold range. When the data exceeds the threshold, trigger the BIM model color grading alarm.

[0101] Among them, the structural safety threshold range is set based on the material strength limit and deformation tolerance value in the bridge design specification. The color level alarm includes red (≥120% threshold), orange (100%-120%), yellow (80%-100%), and green (<80%).

[0102] It can be understood that the HSV color space interpolation algorithm in the embodiment of the present application maps the data exceeding standard ratio to hue (H: 0°-120° corresponding to green-red gradient), saturation (S: 20%-100% representing the severity), and lightness (V: 80% constant), generating a gradient color temperature that conforms to the human eye perception characteristics. Compared with the RGB linear interpolation, this solution can improve the abnormal recognition accuracy from 69.7% to 98.5% at the same color level (16 levels), especially increasing the discrimination for the color-blind user group by 43%. The alarm color is rendered onto the BIM model surface in real time through the ShaderMaterial of Three.js (update frequency 30Hz), and Gaussian blur edge processing is used to eliminate pixel jaggedness, making the visualization positioning error of the structural abnormal area less than 0.5 meters. Through actual measurement, the average fault recognition time of the operation and maintenance personnel is reduced from 8.2 minutes to 1.5 minutes, and the emergency response efficiency is increased by 82%.

[0103] Specifically, when the main girder strain value reaches 28.7MPa (107% of the threshold 26.8MPa), corresponding to H = 102° (orange-red), S = 85%, trigger the secondary alarm. The system synchronously pushes a text message warning to 3 responsible persons and highlights the L12-L14 segments in the BIM model.

[0104] A bridge BIM monitoring data visualization method proposed according to the embodiment of the present application optimizes the hyperparameters of the bidirectional long short-term memory network through an improved leader learning strategy optimization algorithm, constructs a high-precision missing data filling model, and uses the lightweight Sprite sprite technology to establish a spatial mapping mechanism between the sensor data and the BIM model, realizing the real-time dynamic visualization of the monitoring data and the structural safety grading warning. Thus, it solves the problems in the prior art such as the prediction accuracy being insufficient due to the hyperparameters of the bridge monitoring model depending on manual tuning, the reliability of the structural assessment being affected by the missing sensor data, the rendering efficiency of the traditional BIM model being low, restricting the interaction response speed, and the lack of a quantitative grading standard for the abnormal state alarm mechanism, significantly improving the intelligent level and engineering practical value of the bridge safety monitoring system.

[0105] Next, a specific embodiment will be used to elaborate on the bridge BIM monitoring data visualization method, asFigure 2 As shown in the figure, it includes:

[0106] Step 1: Sensor deployment and data acquisition.

[0107] As Figure 3 shown in the figure, 32 FBG strain sensors (with a spacing of 15 m) are arranged on the main girder of the bridge, 18 GNSS displacement monitoring points (with a vertical spacing of 8 m) are installed on the bridge tower, and 24 three-axis accelerometers are set on the stay cables. As Figure 7 shown in the figure, it can be seen from the CEEMDAN decomposition results that after removing the first two high-frequency IMF components (IMF1 - IMF2) from the original strain signal, the signal-to-noise ratio is increased from 14.2 dB to 28.7 dB.

[0108] Specifically, during a certain data missing period (3 hours), the maximum deviation between the reconstructed data and the measured value is 2.3 με, and RMSE = 0.1254.

[0109] It should be noted that the sensors use a temperature compensation algorithm to eliminate the deviation of 175 με caused by the daily average temperature difference of 15 °C.

[0110] Step 2: An improved ISABO algorithm is proposed.

[0111] As Figure 4 shown in the figure, the improved ISABO algorithm optimizes the BiLSTM parameters through the leader learning strategy. As Figure 8 shown in the figure, the comparative experiment results show that the fitness value of ISABO converges to 0.023 after 35 iterations, and the convergence speed is increased by 40% compared with the standard SABO algorithm. The optimized hyperparameters are: learning rate 0.032, number of hidden layer nodes 178, and regularization coefficient 0.047.

[0112] Among them, the formula of the improved ISABO algorithm includes the leader learning strategy update and reset formulas.

[0113] Among them, the leader learning strategy update formula:

[0114]

[0115] Among them, is the new position of the i-th individual in the population, X best is the optimal individual of the iteration, is a normal distribution between [0, 1].

[0116] Reset formula:

[0117]

[0118] Step 3: Based on the improved ISABO algorithm, optimize the hyperparameters of the BiLSTM model, including the learning rate, the number of hidden layer nodes, and the regularization coefficient, to obtain the ISABO-BiLSTM filling model.

[0119] Specifically, the search range of BiLSTM hyperparameters is set as shown in Code Block 1:

[0120] Code Block 1:

[0121] # Parameter space definition

[0122]

[0123] # ISABO optimization process (iterate 50 times)

[0124] best_params = isabo_optimize(bilstm_model, train_data, param_space, max_iter = 50)

[0125] It should be noted that when H > 200, the model training time increases by 320% while the accuracy only improves by 0.7%. Therefore, the upper limit is set to 200.

[0126] Step 4: Data filling and storage

[0127] As Figure 9 shown, use the ISABO-BiLSTM filling model to fill the data, and evaluate the model using the root mean square error RMSE and the mean absolute error MAE.

[0128] Among them, the calculation formulas of RMSE and MAE are:

[0129]

[0130] Among them, N is the number of data, ytp(i) is the filled value, and yact(i) is the actual value.

[0131] Among them, the database uses PostgreSQL 14.5 + TimescaleDB 2.8 extension, and the configuration is as follows: Sharding strategy: Hash partitioning by sensor ID, time chunk granularity 1 hour, compression algorithm: ZSTD (compression ratio 10:1).

[0132] The index structure is shown in Table 1 below.

[0133] Table 1

[0134] Index type Field Performance improvement BRIN index Timestamp 17 times faster for range queries Hash index Sensor ID 23 times faster for point queries GIN index Data type 9 times faster for text retrieval

[0135] Step 5: Sprite model construction and mapping

[0136] As Figure 10 shown, based on the spatial coordinate data in the database, a lightweight Sprite model is created through the WebGL engine (performance comparison. Each Sprite forms a spatial mapping relationship with the BIM model segment. When 5000 sprites are loaded, the rendering frame rate of Three.js remains at 58fps

[0137] Among them, the implementation code of Three.js is shown in Code Block 2.

[0138] Code Block 2:

[0139]

[0140]

[0141] Specifically, the Revit model export process: Obtain the sensor spatial coordinates through the Revit API, and use the gltfRevitExport plug-in to export a lightweight model (the volume is reduced from 1.2GB to 217MB).

[0142] It should be noted that the Sprite texture uses the Signed Distance Fields (SDF) anti-aliasing technology and still maintains a clear outline when scaled to 10% of the size.

[0143] Step Six: Interactive Data Visualization

[0144] As Figure 6 shown by the ray detection mechanism, the Raycaster of Three.js identifies the clicked sprite through bounding box collision detection (which takes 2.3ms). As Figure 5 shown, when the user clicks on the SG25 sensor, the system extracts the monitoring historical data from the database according to the data stream and displays the time series curve through the right panel of the interface.

[0145] Among them, the Raycaster ray detection and data rendering code are shown in Code Block 3.

[0146]

[0147] Specifically, for a certain user click on the SG25 sensor: the ray detection takes 2.3ms; the HTTP request response time is 38ms; the curve rendering time is 45ms.

[0148] Step Seven: Safety Threshold Classification Alarm

[0149] As Figure 11As shown, when the strain value reaches the threshold of 107%, the corresponding area of the BIM model triggers an orange - red alarm. The application flow of the alarm logic execution process controls the end - to - end response delay within 45 ms, meeting the requirements of real - time monitoring.

[0150] Among them, the system dynamically adjusts the color of the BIM model for hierarchical early warning according to the comparison result between the monitoring data and the preset safety threshold. When the monitoring value is within the safe range (less than 80% of the allowable value), the model shows green; when it reaches 80% - 100% of the allowable value, it gradually turns yellow; when it exceeds the allowable value but does not exceed 120%, it shows orange; when it exceeds 120%, it triggers a red alarm. The color - gradient process uses the HSV color - space mapping technology, where the hue smoothly transitions from green (0°) to red (120°), and the saturation increases with the degree of exceeding the standard, ensuring that the color change intuitively reflects the structural safety status.

[0151] Specifically, in the main - girder strain monitoring, when the measured value reaches 28.7 MPa (107% of the allowable value of 26.8 MPa), the system automatically colors the corresponding segment of the BIM model orange - red and pushes the warning message by text message to 3 responsible persons. The operation and maintenance personnel can locate the abnormal position within an average of 1.5 minutes, with an 82% efficiency improvement compared to the traditional manual inspection method.

[0152] It should be noted that the color - grading standard strictly follows the "Code for Design of Highway Bridges" (JTG / T D65 - 01 - 2023), and different thresholds are set for key parameters such as concrete compressive stress and cable vibration frequency. For example, the allowable deviation of the cable vibration frequency is ±0.25 Hz, and when it exceeds this range, a yellow early warning is triggered. The system also uses anti - aliasing rendering technology to make the color boundary transition natural, and the recognition accuracy of color - blind users is increased to 91.3%.

[0153] In summary, the present invention proposes an intelligent processing and visualization method for bridge monitoring data based on the deep integration of ISABO-BiLSTM and WebGL, and constructs a complete "data acquisition - missing value filling - dynamic mapping - interactive warning" technology chain through multi-dimensional technological innovation. Specifically, the improved ISABO algorithm reduces the BiLSTM model parameter optimization error by 66.09% through the leader learning strategy and the dynamic reset mechanism, significantly improving the data filling accuracy; the lightweight Sprite technology combined with the octree spatial index increases the Web-side BIM model rendering efficiency by 18 times and controls the interactive latency within 50 ms; the collaborative architecture of the time series database and the Redis cluster realizes millisecond-level response query for billions of monitoring data and compresses the storage space by 78%; the HSV color gamut dynamic mapping technology combined with the hierarchical alarm mechanism enables the abnormal recognition accuracy to reach 98.5% and improves the operation and maintenance response efficiency by 82%. The organic integration of these technologies not only overcomes the pain points of traditional monitoring systems such as high data missing rate, visualization lag, and rough alarm, but also realizes millimeter-level real-time perception and centimeter-level spatial positioning of the bridge structure state. By reducing the manual review workload by more than 70% and extending the service life of sensors by 30%, this method provides an intelligent solution for the whole life cycle health management of bridges, with significant economic benefits and social safety values.

[0154] Next, a bridge BIM monitoring data visualization device according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0155] Figure 13 It is a block diagram of the bridge BIM monitoring data visualization device according to an embodiment of the present application.

[0156] As Figure 13 shown, the intelligent cache device 10 for front-end resource on-demand loading includes: a data acquisition module 100, an algorithm optimization module 200, a storage module 300, a model construction module 400, an interaction module 500, and a decision module 600.

[0157] Among them, the data acquisition module 100 is used to collect time series data through bridge monitoring sensors; the construction module 200 is used to propose an improved ISABO algorithm, and based on the improved ISABO algorithm, optimize the hyperparameters of the learning rate, the number of hidden layer nodes and the regularization coefficient of the BiLSTM model to obtain the ISABO-BiLSTM filling model; the calculation module 300 is used to input the time series data into the ISABO-BiLSTM filling model for missing value prediction, and store the filled complete data set in the SQL database; the selection module 400 is used to create a Sprite sprite model with a unique ID identifier through the WebGL engine based on the sensor spatial coordinate data in the SQL database, and establish a mapping relationship between the ID and the database primary key to replace the traditional BIM family model; the loading module 500 is used to listen for user interaction events based on the Raycaster ray detection mechanism of Three.js. When the user clicks on the Sprite sprite, extract the associated monitoring data and filling data from the SQL database through the mapping relationship; the adjustment module 600 is used to compare the monitoring data and filling data with a preset structural safety threshold interval. When the data exceeds the threshold interval, trigger the BIM model color grading alarm, where the structural safety threshold interval is set based on the material strength limit and deformation tolerance value in the bridge design specification.

[0158] It should be noted that the foregoing explanation of the embodiments of the bridge BIM monitoring data visualization method also applies to the bridge BIM monitoring data visualization device of this embodiment, and will not be repeated here.

[0159] Figure 14 The following is a schematic structural diagram of the electronic device provided by the embodiment of the present application. The electronic device may include:

[0160] A memory 1401, a processor 1402, and a computer program stored on the memory 1401 and executable on the processor 1402.

[0161] When the processor 1402 executes the program, it implements the bridge BIM monitoring data visualization method provided in the foregoing embodiments.

[0162] Furthermore, the electronic device further includes:

[0163] A communication interface 1403 for communication between the memory 1401 and the processor 1402.

[0164] The memory 1401 is used to store a computer program executable on the processor 1402.

[0165] The memory 1401 may include a high-speed RAM (Random Access Memory) memory and may also include a non-volatile memory, such as at least one disk memory.

[0166] If the memory 1401, the processor 1402, and the communication interface 1403 are implemented independently, the communication interface 1403, the memory 1401, and the processor 1402 can be interconnected through a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component) bus, an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 6 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0167] Optionally, in a specific implementation, if the memory 1401, the processor 1402, and the communication interface 1403 are integrated on a single chip, the memory 1401, the processor 1402, and the communication interface 1403 can communicate with each other through an internal interface.

[0168] The processor 1402 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0169] The embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned method for visualizing bridge BIM monitoring data is implemented.

[0170] In addition, the embodiments of the present application also provide a computer program product, including a computer program or instruction, and when the computer program or instruction is executed, the above-mentioned method for visualizing bridge BIM monitoring data is implemented.

[0171] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0172] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of these features. In the description of this application, "a plurality of" means at least two, such as two, three, etc., unless otherwise specifically and clearly defined.

[0173] Any process or method description shown in a flowchart or described in other ways herein can be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a customized logical function or process. And the scope of the preferred embodiments of this application includes additional implementations, where the functions can be executed in a manner that is not in the order shown or discussed, including in a substantially simultaneous manner according to the involved functions or in a reverse order, which should be understood by those skilled in the art to which the embodiments of this application pertain.

[0174] It should be understood that the various parts of this application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware as in another embodiment, any one of the following techniques well known in the art or a combination thereof can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc.

[0175] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0176] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for visualizing bridge BIM monitoring data, characterized in that, The content of the method includes: S1: Collect time series data through bridge monitoring sensors; S2: An improved ISABO algorithm is proposed, and based on the improved ISABO algorithm, hyperparameter optimization is performed on the learning rate, the number of hidden layer nodes, and the regularization coefficient of the BiLSTM model to obtain the ISABO-BiLSTM filling model; S3: Input the time series data into the filling model for missing value prediction, and store the complete data set after filling in the SQL database; S4: Based on the sensor spatial coordinate data in the SQL database, a Sprite sprite model with a unique ID is innovatively created through the WebGL engine, and a mapping relationship between the ID and the database primary key is established; S5: Listen for user interaction events based on the Raycaster ray detection mechanism of Three.js. When the user clicks on the Sprite sprite, the associated monitoring data and filling data are extracted from the SQL database through the mapping relationship; S6: Compare the monitoring data and filling data with a preset safety threshold range. When the data exceeds the threshold, trigger a BIM model color grading alarm, where the structural safety threshold range is set based on the material strength limit and deformation tolerance value in the bridge design specification.

2. The visualization method of bridge BIM monitoring data according to claim 1, characterized in that In the bridge monitoring sensors, the sensors are deployed at the key monitoring points of the main beam, bridge tower, and stay cables of the bridge, and the collected time series data includes strain, displacement, vibration frequency, and environmental temperature and humidity parameters.

3. The method for visualizing bridge BIM monitoring data according to claim 1, wherein The mechanism of the improved ISABO algorithm includes: S2.1.1: Randomly initialize the population individuals to obtain the initial positions of the individuals. The formula is: X i,d = lb d + r i,d ·(ub d - lb d ), i = 1, ···, N, d = 1, ···, m (1) where Xi is the i-th population individual, N is the population size, m is the number of variables, ri,d is a random number between [0,1], and ubd and lbd are the upper and lower bounds of the variables respectively. S2.1.2: Adjust the factor weight of the SABO subtraction operator to obtain the subtraction operator formula of the ISABO algorithm as: S2.1.3: Update the individual positions through the leader learning strategy with the formula: Among them, is the new position of the i-th population individual, X best is the optimal individual of the iteration, is a normal distribution between [0, 1]; S2.1.4: Reset the exceeded individuals, and the reset formula is: Among them, X i is the i-th population individual, r i,d is a random number between [0, 1], ub d and lb d are the upper and lower bounds of the variable respectively. S2.1.5: Determine the obtained individual positions to decide whether to update the positions of the individuals in the overall population in this round. The position evaluation formula is: Among them, and Fi is and the objective function value of Xi.

4. The method for visualizing bridge BIM monitoring data according to claim 1, characterized in that, The hyperparameter optimization of the BiLSTM model includes: S2.2.1: Perform CEEMDAN decomposition on the original time series data, remove the high-frequency noise components, and then reconstruct the input data; S2.2.2: Divide the reconstructed data into a training set and a test set in a ratio of 7:3, and use the root mean square error as the fitness function; S2.2.3: Search for the optimal learning rate lr∈[0.001,0.1], the number of hidden layer nodes H∈[50,200], and the regularization coefficient λ∈[0.01,0.1] in the solution space through the ISABO algorithm.

5. The visualization method of bridge BIM monitoring data according to claim 1, wherein Creating a Sprite sprite model with a unique ID through the WebGL engine includes: S4.1.1: Obtain the spatial coordinates of the sensor through Revit secondary development, and use the gltfRevitExport plug-in to export the lightweight BIM model to the Web end; S4.1.2: Generate a 2D sprite texture with an ID based on Three.js's SpriteMaterial. The texture content includes the sensor type icon and status indicator.

6. The method for visualizing bridge BIM monitoring data according to claim 1, wherein, The Raycaster ray detection mechanism based on Three.js includes: S5.1.1: Get the ID of the clicked sprite through collision detection between the ray and the bounding box of the sprite; S5.1.2: Send an HTTP request to the server based on the ID to query the associated monitoring history data and fill in the data timestamp; S5.1.3: Use ECharts to render the query results as a time series curve chart with annotated points, where the filled data is highlighted in orange.

7. A bridge BIM monitoring data visualization device, characterized in that, include: Data acquisition module, S1: collects time series data through bridge monitoring sensors; Algorithm optimization module, S2: an improved ISABO algorithm is proposed, and based on the improved ISABO algorithm, the learning rate, number of hidden layer nodes and regularization coefficient of the BiLSTM model are optimized for hyperparameters to obtain the ISABO-BiLSTM filling model; Storage module, S3: input the time series data into the ISABO-BiLSTM filling model to predict missing values, and store the filled complete data set into the SQL database; Model building module, S4: based on the sensor spatial coordinate data in the SQL database, a Sprite model with a unique ID is created through a WebGL engine, and a mapping relationship between the ID and the database primary key is established to replace the traditional BIM family model; Interaction module, S5: Based on the Raycaster ray detection mechanism of Three.js, monitor user interaction events. When the user clicks the Sprite, the associated monitoring data and fill-in data are extracted from the SQL database through the mapping relationship. Decision module, S6: Based on the monitoring data and the filled data, compare them with the preset structural safety threshold interval. When the data exceeds the threshold interval, trigger the BIM model color grading alarm, wherein the structural safety threshold interval is set based on the material strength limit and deformation allowable value in the bridge design specification.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the bridge BIM monitoring data visualization method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instruction is executed, the bridge BIM monitoring data visualization method described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instruction is executed, the bridge BIM monitoring data visualization method described in any one of claims 1 to 6 is implemented.

Images