Cantilever bridge construction risk monitoring system based on digital twinning

Abstract

The application discloses a kind of based on digital twinning cantilever bridge construction risk monitoring system, it is related to the technical field of intelligent monitoring, including construction module and analysis module, crane construction spectrum is constructed, according to current construction data and crane construction spectrum, obtain construction parameter, analyze construction environment, and crane parameter is registered.The crane construction spectrum of the present application is generated by historical big data, which converts experience into quantitative boundaries, eliminates human subjective judgment errors, compares current operation parameters with spectrum second by second, discovers over-limit trends in advance, realizes early warning instead of post-repair, and ensures zero deviation between warning instructions and real pose through VR, AR target, point cloud and ICP.The wind and rainfall are written into the registration matrix in real time, and the crane allows the working range to automatically expand and contract with the weather, reducing the false alarm rate of downtime.The distance, rotation, pitch and lifting four key parameters are registered and locked at one time, and the driver does not need to switch between multiple systems, improving operation efficiency.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent monitoring, and in particular to a construction risk monitoring system for cantilever bridges based on digital twins. Background Technology

[0002] In recent years, the technology for risk monitoring in cantilever bridge construction has developed rapidly. Using BIM and GIS as a foundation, on-site sensor data is driven in real time by edge computing boxes to create a digital twin scene where what you see is what you measure. By introducing Bayesian networks, cloud models, and stochastic processes-entropy weighting methods, it can integrate multi-source information such as temperature, wind speed, and train window periods to achieve dynamic updates of risk probabilities. Using a hybrid framework of finite element method, machine learning, and Monte Carlo, it can predict the probability of axis deviation and unbalanced bending moment, improving the accuracy compared to purely empirical models.

[0003] Currently, Chinese invention patent CN107085644A discloses a risk assessment method for cantilever construction of concrete bridges under complex risk sources. This method treats complex risk factors as random variables, determines the random variables and their probability distributions, and establishes a critical state equation for bridge cantilever construction risk under complex risk factors. It then establishes a finite element model, designs samples for the random variables, and constructs training and testing samples. Through sample learning, a BP neural network forms a nonlinear mapping relationship between the sample input and output parameters. Based on the risk critical state equation, it uses the Monte Carlo principle to simulate random sampling and solve for the risk probability of specific risk events. The risk is evaluated based on a probability description table. However, this related technology lacks position conversion calibration based on the deviation between displayed coordinates and virtual scene coordinates, resulting in a lack of accuracy in risk detection. Furthermore, it does not construct a construction knowledge graph corresponding to different bridge parameters and equipment parameters based on historical construction data, which is detrimental to the scientific nature and rapid response of construction risk detection, thus presenting certain limitations. Summary of the Invention

[0004] The technical problem solved by this invention is that related technologies do not perform position conversion and calibration based on the deviation between the displayed coordinates and the virtual scene coordinates, lack accuracy in risk detection, and do not construct construction knowledge maps corresponding to different bridge parameters and different equipment parameters based on historical construction data, which is not conducive to the scientific nature and rapid response of construction risk detection and has certain limitations.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a construction risk monitoring system for cantilever bridges based on digital twins, comprising a construction module and an analysis module;

[0006] The construction module constructs a crane construction spectrum based on historical construction data, and obtains construction parameters based on current construction data and the crane construction spectrum.

[0007] The analysis module analyzes the construction environment and registers the crane parameters based on the analysis results to obtain the registration results.

[0008] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the historical construction data includes historical main arch clear span, historical crane model, historical environmental data, historical operation time and historical crane distance data, and the historical construction process type corresponding to each historical construction data is cantilever assembly method.

[0009] The historical crane distance data is represented as the distance between the midpoint of the line segment formed by the crane's center of gravity and the endpoint of the main arch's net span.

[0010] The historical crane model refers to the model of the crane used in the construction of the historical cantilever bridge;

[0011] The historical environmental data includes historical wind speed and historical average hourly rainfall.

[0012] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the system includes: obtaining historical crane models, retrieving a crane database, inputting historical crane models into the crane database, and matching the rated lifting capacity, rated slewing angle, rated pitch angle, rated wind force level, and rated lifting height corresponding to the historical crane models.

[0013] The rated slewing angle is the maximum clockwise rotation angle or the maximum counterclockwise rotation angle of the uppermost structure of the enclosure. The rated pitch angle is the maximum pitch angle or the maximum tilt angle of the crane boom. The rated wind force level is the dividing wind force level between continued operation and cessation of operation when the crane is operating outdoors, and the rated wind force level is the maximum wind force level for continued operation. The rated lifting height is the maximum vertical distance that the hook can be raised.

[0014] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, wherein: any historical construction data is extracted to construct historical sample data;

[0015] A five-dimensional grid is established with the first value being the distance step size, the second value being the lifting weight step size, the third value being the angle step size, and the fourth value being the wind force level step size.

[0016] Historical sample data of the same main arch ring net span are selected, and the 90% and 10% values ​​of the corresponding historical crane distance data are statistically analyzed and denoted as D90 and D10 respectively. A two-dimensional safety zone is generated with D10 as the inner boundary and D90 as the outer boundary, which is denoted as the first zone.

[0017] Within each five-dimensional grid corresponding to the net span of the main arch ring, the total number of historical sample data is counted and recorded as the first quantity. The historical working time corresponding to each historical sample data is sorted in descending order, and the historical working time with the largest value is selected to calculate the unit confidence.

[0018] Using the unit confidence level as the pixel value, a two-dimensional safety heat map of historical lifting weight and historical rotation angle is obtained, which is denoted as the first image;

[0019] Obtain the rated wind force level corresponding to each crane model in the clear span of the main arch ring. When the wind force level of the historical sample data is less than or equal to the corresponding rated wind force level, set the first safety factor to 0.8; otherwise, set the first safety factor to 0.4. Obtain the historical average hourly rainfall in the clear span of the main arch ring. Set the first rainfall amount as the rainfall threshold. Calculate the first difference between the historical average hourly rainfall amount and the first rainfall amount. Set the fifth value as the first difference threshold. Compare the first difference with the fifth value. When the first difference is less than or equal to the fifth value, set the second safety factor to 0.8; otherwise, set the second safety factor to 0.4. Calculate the first average value of the first safety factor and the second safety factor.

[0020] The net span of the main arch ring and the crane model in the historical sample data of the net span of the main arch ring are combined to obtain the first combination. The sample points of each corresponding pitch angle and lifting height in any first combination are extracted and convex hull operation is performed to obtain the boundary polygon. The upper boundary function of the boundary polygon is fitted by the least squares method.

[0021] The value of the first proportion of the boundary function is set as the upper limit of the allowable range, and the first proportion is distributed between 0 and 1.

[0022] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the expression for historical sample data is:

[0023] R = {L, M, D, W, ...} , , H, V, T};

[0024] Where R represents historical sample data, L represents historical net span of the main arch ring, M represents historical crane model, D represents historical crane distance data, and W represents rated wind force level. This represents the historical average rainfall. For the rated rotation angle, H is the rated pitch angle, V is the rated lifting height, and T is the historical operating time.

[0025] The expression for calculating the unit confidence level is as follows:

[0026] ;

[0027] Where C is the unit confidence level, N is the first quantity, and T max The longest historical working time is represented by the largest value, where e is a natural number.

[0028] The expression for the upper boundary function is:

[0029] ;

[0030] Where a, b, and c are constants.

[0031] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the logic for constructing a crane construction spectrum based on historical construction data includes:

[0032] Acquire the distance data of the first region and the current crane. If the current crane distance data is not distributed in the corresponding first region, send a distance over-limit warning. Do not send a distance over-limit warning until the current crane distance data is distributed in the corresponding first region.

[0033] The system acquires the first image, the current lifting weight, and the current slewing angle. Based on the first image and the current lifting weight, it obtains the corresponding theoretical slewing angle. The system compares the current slewing angle with the theoretical slewing angle. If the current slewing angle is different from the theoretical slewing angle, a slewing angle over-limit warning is sent. The system continues to send a slewing angle over-limit warning until the current slewing angle is the same as the theoretical slewing angle.

[0034] Obtain the first average value. If the first average value is not 0.8, send a weather over-limit warning. Continue to send a weather over-limit warning until the weather improves and the first average value is 0.8.

[0035] Obtain the upper boundary function, current pitch angle, and current lift height. Input the current pitch angle into the upper boundary function to obtain the current function result. Calculate the first product of the current function result and the first ratio, and record it as the allowable upper limit. Compare the current lift height with the allowable upper limit. If the current lift height is greater than the allowable upper limit, send a lift height over-limit message. Continue until the lift height is less than or equal to the allowable upper limit, then do not send a lift height over-limit message.

[0036] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, construction parameters are obtained based on current construction data and crane construction data. The current construction data includes the current main arch clear span, current crane model, current pitch angle, current lifting capacity, current wind force level, and current hourly average rainfall. The construction parameters include the current crane distance data, current slewing angle, current construction time status, and current lifting height.

[0037] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the logic for acquiring the construction parameters includes:

[0038] Obtain the current net span of the main arch ring, retrieve the first region corresponding to the current net span of the main arch ring, select any point within the first region, and obtain the straight-line distance between the crane boom and the nearest high-voltage power pole, the straight-line distance between the wire rope and the nearest high-voltage power pole, and the straight-line distance between the hoisted object and the nearest high-voltage power pole at that point. Record these as safety distances. If any safety distance is less than the standard safety distance, jump to the next point. If all safety distances are greater than or equal to the standard safety distance, set the point as the location of the crane's center of gravity. Set the straight-line distance between the location of the crane's center of gravity and the midpoint of the line segment formed between the endpoints of the net span of the main arch ring as the current crane distance data. Adjust the position of the crane relative to the net span of the main arch ring based on the current crane distance data.

[0039] Obtain the current lifting capacity, match the corresponding current slewing angle based on the first image, and adjust the slewing angle of the crane according to the current slewing angle;

[0040] Obtain the current pitch angle, match the corresponding current lifting height according to the upper boundary function, and adjust the lifting height of the crane according to the current lifting height;

[0041] Obtain the current wind force level and the current hourly average rainfall. Based on the calculation logic of the first average value, obtain the first average value corresponding to the current wind force level and the current hourly average rainfall. When the current first average value is equal to 0.8, set the current construction time status to "construction can be carried out today". When the current first average value is not equal to 0.8, set the current construction time status to "construction cannot be carried out today".

[0042] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the logic for analyzing the construction environment includes:

[0043] Acquire videos of the construction environment, construct a digital twin scene based on the videos, align the digital twin scene with the timeline of the on-site video stream, pre-mark K AR target positions in the digital twin scene, and have operators on-site use an AR headset to place real targets onto corresponding physical points. Measure the global coordinates of the real targets using a total station, and bridge the virtual coordinate system (VCS) and the on-site coordinate system (GCS) using the targets to obtain a coarse transformation matrix. Activate the depth camera in the AR headset to scan points on the crane boom, obtaining a point cloud, and then integrate the point cloud with the digital twin scene. ICP iteration is performed, allowing the crane to perform slewing and hoisting actions in sequence, and continuously comparing the error between the virtual boom and the real boom. The error is represented by the straight-line distance between the center of gravity of the virtual boom and the center of gravity of the real boom. The sixth value is set as the deviation threshold, and the error is compared with the sixth value. When the error is greater than the sixth value, ICP iteration is repeated until the error is less than or equal to the sixth value. Then, ICP iteration is stopped, and the registration step is completed. The coarse transformation matrix at this time is set as the registration matrix, and the registration matrix is ​​set as the analysis result.

[0044] Based on the analysis results, the crane parameters are registered to obtain the registration results. The crane parameters include crane distance data, slewing angle, pitch angle and lifting height.

[0045] As a preferred embodiment of the digital twin-based cantilever bridge construction risk monitoring system described in this invention, the construction logic of the digital twin scenario includes:

[0046] The construction environment video is extracted into continuous frames at a fixed time step. The motion recovery structure algorithm is run on each frame to automatically track key feature points and generate sparse point clouds and camera pose sequences. Using the sparse point cloud as a seed, a multi-view stereo algorithm is used for diffusion matching to obtain a dense point cloud with color information. Poisson reconstruction is performed on the dense point cloud to generate a renderable 3D mesh model. An instance segmentation network is used to label the "sky, ground, crane, and utility pole" in the mesh to form a semantic shell. For the segmented "crane" area, the coarse mesh is replaced with a CAD model. The local coordinate system of the reconstructed model is transformed to the design coordinate system through on-site control points. Each frame is assigned a timestamp from the original video to complete the construction of the digital twin scene.

[0047] The logic for generating the registration matrix includes:

[0048] The AR head-mounted display identifies the center pixel coordinates of the real target, and the total station automatically aims at the same target and outputs global coordinates, forming a correspondence between the pixel and the global coordinates. A coarse transformation matrix is ​​obtained using a three-point coplanar algorithm. The AR depth camera sweeps the crane boom, aggregates depth frames, generates a local point cloud of the boom, and automatically filters out outliers. Each vertex of the point cloud is matched with the nearest neighbor triangle in the digital twin scene to establish a correspondence between the observation point and the model face. The ICP algorithm is used to iteratively update the coarse transformation matrix. After each iteration, the crane performs a slewing motion before a lifting motion, and the error between the virtual boom center of gravity and the real boom center of gravity is calculated. If the error is greater than the sixth value, the ICP iteration is re-executed; otherwise, the current coarse transformation matrix is ​​locked and set as the registration matrix.

[0049] The beneficial effects of this invention are as follows: It generates crane construction spectrum diagrams using historical big data, transforming experience into quantifiable boundaries, eliminating human subjective judgment errors, comparing current operating parameters with the spectrum diagram in seconds, detecting over-limit trends in advance, and achieving early warning rather than post-event remediation. Through VR, AR targets, point clouds, and ICP, it ensures zero deviation between warning commands and actual posture. Wind force and rainfall are written into the registration matrix in real time, allowing the crane's working range to automatically expand and contract with the weather, reducing the false alarm rate of downtime. The four key parameters of distance, slewing, pitch, and hoisting are registered and locked at one time, eliminating the need for the operator to switch between multiple systems and improving work efficiency. Attached Figure Description

[0050] Figure 1 This is a basic flowchart of a digital twin-based cantilever bridge construction risk monitoring system provided in one embodiment of the present invention. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0052] Example, refer to Figure 1 As an embodiment of the present invention, a construction risk monitoring system for cantilever bridges based on digital twins is provided, including a construction module and an analysis module;

[0053] The construction module constructs a crane construction spectrum based on historical construction data, and obtains construction parameters based on current construction data and the crane construction spectrum.

[0054] The analysis module analyzes the construction environment and registers the crane parameters based on the analysis results to obtain the registration results.

[0055] This invention generates crane operation spectrum diagrams using historical big data, transforming experience into quantifiable boundaries and eliminating human subjective judgment errors. Current operating parameters are compared with the spectrum diagram in seconds to detect over-limit trends in advance, achieving early warning rather than post-event remediation. Through VR, AR targets, point clouds, and ICP, it ensures zero deviation between warning commands and actual position and posture. Wind force and rainfall are written into the registration matrix in real time, and the crane's allowable working area automatically expands and contracts with the weather, reducing the false alarm rate of downtime. The four key parameters of distance, slewing, pitch, and hoisting are registered and locked at one time, eliminating the need for the operator to switch between multiple systems and improving work efficiency.

[0056] The historical construction data includes the historical main arch ring clear span, historical crane model, historical environmental data, historical operation time and historical crane distance data. The historical construction process type corresponding to each historical construction data is the cantilever assembly method.

[0057] The historical crane distance data is represented as the distance between the midpoint of the line segment formed by the crane's center of gravity and the endpoint of the main arch's net span.

[0058] The historical crane model refers to the model of the crane used in the construction of the historical cantilever bridge;

[0059] The historical environmental data includes historical wind speed and historical average hourly rainfall.

[0060] In practice, all samples are limited to the "cantilever assembly method" to avoid mixing in different construction methods (jacking, rotation), ensuring consistency between historical data and the physical mechanisms of the current construction scenario, and improving prediction reliability. The "straight-line distance between the crane's center of gravity and the midpoint of the main arch's clear span endpoint" is used as a unified spatial scale to eliminate dimensional differences under different bridge spans and swing amplitudes. Historical data can be directly compared horizontally, achieving "one map for multiple uses." Simultaneous recording of "wind force + average hourly rainfall" and "operation time" quantifies the real impact of severe weather on construction continuity, providing time-varying data for subsequent dynamic thresholds and weather compensation modules. The penalty coefficient of the label reduces false alarms and missed alarms. Using "operation duration" as an implicit quality label, short-term trial lifting records are automatically downgraded, effectively filtering disturbing data in the historical database, making the unit confidence calculation more robust. The historical crane model is bound to the rated parameter database once, and only the model needs to be entered on site to retrieve the corresponding rated wind force, rated lifting capacity and other thresholds in seconds, eliminating repeated entry and improving deployment efficiency. Once the current construction is completed, the new "distance-environment-duration" record is immediately written back to the historical database, realizing online rolling updates of the spectrum. The more the system is used, the "smarter" it becomes, and the risk boundary continues to converge with engineering experience.

[0061] Obtain historical crane models, retrieve the crane database, input the historical crane models into the crane database, and match the rated lifting capacity, rated slewing angle, rated pitch angle, rated wind force level, and rated lifting height corresponding to the historical crane models;

[0062] The rated slewing angle is the maximum clockwise rotation angle or the maximum counterclockwise rotation angle of the uppermost structure of the enclosure. The rated pitch angle is the maximum pitch angle or the maximum tilt angle of the crane boom. The rated wind force level is the dividing wind force level between continued operation and cessation of operation when the crane is operating outdoors, and the rated wind force level is the maximum wind force level for continued operation. The rated lifting height is the maximum vertical distance that the hook can be raised.

[0063] In practice, simply inputting the "historical crane model" is sufficient for the system to automatically retrieve the complete set of rated indicators, eliminating the need for manual retrieval of manuals and PDFs. Deployment time is reduced from hours to seconds. Unique engineering definitions are provided for easily confused concepts such as "slewing," "pitch," and "wind force level." Data from different projects and manufacturers interact within the same coordinate system, completely eliminating the hidden dangers of "different meanings for the same name." All rated values ​​are derived from factory type tests and national standard inspection certificates. When comparing with real-time values ​​on site, the rated wind force level is directly compared with the real-time wind speed. The system can instantly provide a "continue / stop" binary decision without the need for experience coefficients, achieving zero-delay judgment of weather risks. When the crane is replaced or an auxiliary machine is added on site, the threshold is updated immediately upon entering the new model number. Historical spectrum and current warning logic do not need to be recoded, supporting rapid adaptation to "one machine, one policy." Each new model-rated record expands the database, allowing for zero-cost reuse of the same type of equipment in subsequent projects. The enterprise-level knowledge base continues to grow with the project schedule.

[0064] Extract any historical construction data point to construct historical sample data;

[0065] A five-dimensional grid is established with the first value being the distance step size, the second value being the lifting weight step size, the third value being the angle step size, and the fourth value being the wind force level step size.

[0066] Historical sample data of the same main arch ring net span are selected, and the 90% and 10% values ​​of the corresponding historical crane distance data are statistically analyzed and denoted as D90 and D10 respectively. A two-dimensional safety zone is generated with D10 as the inner boundary and D90 as the outer boundary, which is denoted as the first zone.

[0067] Within each five-dimensional grid corresponding to the net span of the main arch ring, the total number of historical sample data is counted and recorded as the first quantity. The historical working time corresponding to each historical sample data is sorted in descending order, and the historical working time with the largest value is selected to calculate the unit confidence.

[0068] Using the unit confidence level as the pixel value, a two-dimensional safety heat map of historical lifting weight and historical rotation angle is obtained, which is denoted as the first image;

[0069] Obtain the rated wind force level corresponding to each crane model in the clear span of the main arch ring. When the wind force level of the historical sample data is less than or equal to the corresponding rated wind force level, set the first safety factor to 0.8; otherwise, set the first safety factor to 0.4. Obtain the historical average hourly rainfall in the clear span of the main arch ring. Set the first rainfall amount as the rainfall threshold. Calculate the first difference between the historical average hourly rainfall amount and the first rainfall amount. Set the fifth value as the first difference threshold. Compare the first difference with the fifth value. When the first difference is less than or equal to the fifth value, set the second safety factor to 0.8; otherwise, set the second safety factor to 0.4. Calculate the first average value of the first safety factor and the second safety factor.

[0070] The net span of the main arch ring and the crane model in the historical sample data of the net span of the main arch ring are combined to obtain the first combination. The sample points of each corresponding pitch angle and lifting height in any first combination are extracted and convex hull operation is performed to obtain the boundary polygon. The upper boundary function of the boundary polygon is fitted by the least squares method.

[0071] The value of the first proportion of the boundary function is set as the upper limit of the allowable range, and the first proportion is distributed between 0 and 1.

[0072] In practice, continuous physical quantities are divided into statistically discretizable units using a "distance step - lifting weight step - angle step - wind force step" approach. Data from different bridge spans and different crane models can be directly superimposed within the same grid space, solving the pain point of "difficulty in aligning large datasets." A "first region" is generated with D10 (10th percentile) as the inner edge and D90 as the outer edge. If a crane falls outside the zone, a distance over-limit warning is immediately triggered, eliminating the need for manually setting an "experienced safety distance." The threshold adaptively evolves with historical samples, and the unit confidence level simultaneously penalizes both "few samples" and "operational" factors. The "short" design allows for varying shades of color in the heat map, making risks easily identifiable to drivers. A double 0.8 is applied when wind speed is less than or equal to the rated continuous value and rainfall is less than or equal to the threshold; exceeding either value drops the threshold to 0.4. The average value directly determines the "construction can proceed today" status, simplifying complex meteorological criteria into a single 0.8 / non-0.8 switch, resulting in zero interpretation costs on-site. A convex hull and least squares fit is used to obtain the "upper boundary function" for pitch-lift, which, multiplied by 0.9 (the first ratio), yields the allowable upper limit. An alarm is triggered if the lifting height is falsely exceeded, physically preventing the risk of "hook head hitting the boom" or breakage due to over-the-top contact.

[0073] The expression for historical sample data is:

[0074] R = {L, M, D, W, ...} , , H, V, T};

[0075] Where R represents historical sample data, L represents historical net span of the main arch ring, M represents historical crane model, D represents historical crane distance data, and W represents rated wind force level. This represents the historical average rainfall. For the rated rotation angle, H is the rated pitch angle, V is the rated lifting height, and T is the historical operating time.

[0076] The expression for calculating the unit confidence level is as follows:

[0077] ;

[0078] Where C is the unit confidence level, N is the first quantity, and T max The longest historical working time is represented by the largest value, where e is a natural number.

[0079] The expression for the upper boundary function is:

[0080] ;

[0081] Where a, b, and c are constants.

[0082] In practical implementation, R = {L, M, D, W, ...} , , The five dimensions of "bridge span, space, environment, capability, and time" are packaged into a self-consistent vector. Subsequent mesh generation, confidence calculation, and convex hull fitting all reuse the same set of fields, avoiding the accuracy loss caused by "intermediate conversion". Once fitted, a, b, and c are fixed constants. The embedded controller only needs 3 floating-point operations to obtain H_max, which is the maximum value of the upper boundary function. The computational load is <0.1ms, which is suitable for direct calling by PLC / industrial control computer.

[0083] The logic for constructing a crane construction spectrum based on historical construction data includes:

[0084] Acquire the distance data of the first region and the current crane. If the current crane distance data is not distributed in the corresponding first region, send a distance over-limit warning. Do not send a distance over-limit warning until the current crane distance data is distributed in the corresponding first region.

[0085] The system acquires the first image, the current lifting weight, and the current slewing angle. Based on the first image and the current lifting weight, it obtains the corresponding theoretical slewing angle. The system compares the current slewing angle with the theoretical slewing angle. If the current slewing angle is different from the theoretical slewing angle, a slewing angle over-limit warning is sent. The system continues to send a slewing angle over-limit warning until the current slewing angle is the same as the theoretical slewing angle.

[0086] Obtain the first average value. If the first average value is not 0.8, send a weather over-limit warning. Continue to send a weather over-limit warning until the weather improves and the first average value is 0.8.

[0087] Obtain the upper boundary function, current pitch angle, and current lift height. Input the current pitch angle into the upper boundary function to obtain the current function result. Calculate the first product of the current function result and the first ratio, and record it as the allowable upper limit. Compare the current lift height with the allowable upper limit. If the current lift height is greater than the allowable upper limit, send a lift height over-limit message. Continue until the lift height is less than or equal to the allowable upper limit, then do not send a lift height over-limit message.

[0088] In practice, the four-level locking system (distance, turning angle, weather, altitude) employs a combination of "hard boundaries and soft prompts" at each step: exceeding limits triggers an immediate audible and visual alarm; failure to do so results in automatic speed limiting or vehicle locking, preventing the driver from forcibly continuing and achieving "physical-level" risk prevention. The first zone, theoretical turning angle, 0.8 weather coefficient, and upper boundary function are all derived from historical big data, allowing for real-time comparison and instantaneous feedback to reduce warning delays. All boundaries are updated online with historical samples, resolving the frequent readjustment issues caused by bridge type and equipment replacement in traditional "experience tables." Once deployed, the system becomes increasingly accurate with use. The four-level logic is interconnected, allowing passage only when "position + attitude + environment + space" are simultaneously compliant. Single-point sensor drift will not trigger a complete shutdown, reducing the false alarm rate.

[0089] Based on the current construction data and the crane construction diagram, the construction parameters are obtained. The current construction data includes the current main arch clear span, current crane model, current pitch angle, current lifting capacity, current wind force level, and current hourly average rainfall. The construction parameters include the current crane distance data, current slewing angle, current construction time status, and current lifting height.

[0090] The logic for obtaining the construction parameters includes:

[0091] Obtain the current net span of the main arch ring, retrieve the first region corresponding to the current net span of the main arch ring, select any point within the first region, and obtain the straight-line distance between the crane boom and the nearest high-voltage power pole, the straight-line distance between the wire rope and the nearest high-voltage power pole, and the straight-line distance between the hoisted object and the nearest high-voltage power pole at that point. Record these as safety distances. If any safety distance is less than the standard safety distance, jump to the next point. If all safety distances are greater than or equal to the standard safety distance, set the point as the location of the crane's center of gravity. Set the straight-line distance between the location of the crane's center of gravity and the midpoint of the line segment formed between the endpoints of the net span of the main arch ring as the current crane distance data. Adjust the position of the crane relative to the net span of the main arch ring based on the current crane distance data.

[0092] Obtain the current lifting capacity, match the corresponding current slewing angle based on the first image, and adjust the slewing angle of the crane according to the current slewing angle;

[0093] Obtain the current pitch angle, match the corresponding current lifting height according to the upper boundary function, and adjust the lifting height of the crane according to the current lifting height;

[0094] Obtain the current wind force level and the current hourly average rainfall. Based on the calculation logic of the first average value, obtain the first average value corresponding to the current wind force level and the current hourly average rainfall. When the current first average value is equal to 0.8, set the current construction time status to "construction can be carried out today". When the current first average value is not equal to 0.8, set the current construction time status to "construction cannot be carried out today".

[0095] In practice, the straight-line distances between the boom, wire rope, and hoisted load and the high-voltage pole are calculated separately. If any value is less than the standard safety distance, the point is skipped, achieving "one machine, three objects" protection against electric shock. This eliminates the blind spots of traditional methods that only measure the machine's position. The "final qualified point" → "midpoint of the main arch ring's net span endpoint line segment" → "straight-line distance" are encapsulated as the current crane distance data, output entirely by the algorithm. No manual measuring or total station secondary measurement is required. Using the first image (historical lifting weight - slewing angle heatmap) for reverse lookup, the theoretical slewing angle is immediately given upon input of the current lifting weight. The on-site driver can then rotate according to the diagram, eliminating overswing or collision with the template caused by "empirical estimation." Risk mitigation: The upper boundary function H_max outputs the current maximum allowable point with a single click, and the PLC directly limits the position; compared to mechanical ramming blocks, it decelerates 2m in advance, eliminating impact and maintenance, extending the life of the wire rope; wind force + rainfall → first average value, equal to 0.8 allows passage, ≠ 0.8 immediately displays "Construction not allowed today"; drivers do not need to understand the meaning of Beaufort level or mm / h, mm / h represents the unit of rainfall, i.e., millimeters per hour; stop instructions have zero interpretation cost, execution is completed upon supervisor's signature; all four parameters are driven by a dual source of "historical big data + real-time sensing", ensuring that construction always operates within the "historically proven safe" envelope.

[0096] The logic for analyzing the construction environment includes:

[0097] Acquire videos of the construction environment, construct a digital twin scene based on the videos, align the digital twin scene with the timeline of the on-site video stream, pre-mark K AR target positions in the digital twin scene, and have operators on-site use an AR headset to place real targets onto corresponding physical points. Measure the global coordinates of the real targets using a total station, and bridge the virtual coordinate system (VCS) and the on-site coordinate system (GCS) using the targets to obtain a coarse transformation matrix. Activate the depth camera in the AR headset to scan points on the crane boom, obtaining a point cloud, and then integrate the point cloud with the digital twin scene. ICP iteration is performed, allowing the crane to perform slewing and hoisting actions in sequence, and continuously comparing the error between the virtual boom and the real boom. The error is represented by the straight-line distance between the center of gravity of the virtual boom and the center of gravity of the real boom. The sixth value is set as the deviation threshold, and the error is compared with the sixth value. When the error is greater than the sixth value, ICP iteration is repeated until the error is less than or equal to the sixth value. Then, ICP iteration is stopped, and the registration step is completed. The coarse transformation matrix at this time is set as the registration matrix, and the registration matrix is ​​set as the analysis result.

[0098] Based on the analysis results, the crane parameters are registered to obtain the registration results. The crane parameters include crane distance data, slewing angle, pitch angle and lifting height.

[0099] In practice, existing surveillance / aerial video footage is used to directly generate digital twin scenes without the need for additional laser scanners and without affecting continuous on-site operations. The target is simultaneously read by the AR headset (pixel-level) and the total station (millimeter-level), completing both the coarse transformation from VCS to GCS and from GCS to VCS in one go. The initial error is controlled within 1cm, providing high-quality initial values ​​for subsequent ICP. The depth camera generates 30,000 points / frame as the arm moves, matching with the virtual CAD surface ICP. The virtual-to-real center of gravity is locked when the distance is less than 6mm. The driver can directly see the "red-green" superposition through the head-mounted display, and can intuitively correct the error without understanding three-dimensional coordinates. The crane is forced to rotate before lifting, and the center of gravity error is continuously compared within 10 seconds throughout the process. If it exceeds the threshold (6mm), it is immediately recalculated to ensure that the registration matrix is ​​effective throughout the operation and to avoid virtual model drift caused by load deformation or wind load. Once the final matrix is ​​locked, the four parameters of distance, rotation angle, pitch angle and lifting height are all automatically converted to the global coordinate system. Subsequent warnings, limits and spectrum comparisons can be directly called without secondary conversion.

[0100] The construction logic of the digital twin scenario includes:

[0101] The construction environment video is extracted into continuous frames at a fixed time step. The motion recovery structure algorithm is run on each frame to automatically track key feature points and generate sparse point clouds and camera pose sequences. Using the sparse point cloud as a seed, a multi-view stereo algorithm is used for diffusion matching to obtain a dense point cloud with color information. Poisson reconstruction is performed on the dense point cloud to generate a renderable 3D mesh model. An instance segmentation network is used to label the "sky, ground, crane, and utility pole" in the mesh to form a semantic shell. For the segmented "crane" area, the coarse mesh is replaced with a CAD model. The local coordinate system of the reconstructed model is transformed to the design coordinate system through on-site control points. Each frame is assigned a timestamp from the original video to complete the construction of the digital twin scene.

[0102] The logic for generating the registration matrix includes:

[0103] The AR head-mounted display identifies the center pixel coordinates of the real target, and the total station automatically aims at the same target and outputs global coordinates, forming a correspondence between the pixel and the global coordinates. A coarse transformation matrix is ​​obtained using a three-point coplanar algorithm. The AR depth camera sweeps the crane boom, aggregates depth frames, generates a local point cloud of the boom, and automatically filters out outliers. Each vertex of the point cloud is matched with the nearest neighbor triangle in the digital twin scene to establish a correspondence between the observation point and the model face. The ICP algorithm is used to iteratively update the coarse transformation matrix. After each iteration, the crane performs a slewing motion before a lifting motion, and the error between the virtual boom center of gravity and the real boom center of gravity is calculated. If the error is greater than the sixth value, the ICP iteration is re-executed; otherwise, the current coarse transformation matrix is ​​locked and set as the registration matrix.

[0104] In practice, instance segmentation first forms a "semantic shell," and then replaces the "crane" area with a high-precision CAD model. This preserves the texture of the surrounding environment while ensuring that the error of key mechanical dimensions is less than 1cm, thus resolving the contradiction between "realistic scene" and "precise machinery."

[0105] This invention generates crane operation spectrum diagrams using historical big data, transforming experience into quantifiable boundaries and eliminating human subjective judgment errors. Current operating parameters are compared with the spectrum diagram in seconds to detect over-limit trends in advance, achieving early warning rather than post-event remediation. Through VR, AR targets, point clouds, and ICP, it ensures zero deviation between warning commands and actual position and posture. Wind force and rainfall are written into the registration matrix in real time, and the crane's allowable working area automatically expands and contracts with the weather, reducing the false alarm rate of downtime. The four key parameters of distance, slewing, pitch, and hoisting are registered and locked at one time, eliminating the need for the operator to switch between multiple systems and improving work efficiency.

[0106] Those skilled in the art will understand that embodiments of the present invention can be provided as logic, system, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A construction risk monitoring system for cantilever bridges based on digital twins, characterized in that, Includes a build module and an analysis module; The construction module constructs a crane construction spectrum based on historical construction data, and obtains construction parameters based on current construction data and the crane construction spectrum. The historical construction data includes the historical main arch ring clear span, historical crane model, historical environmental data, historical operation time and historical crane distance data. The historical construction process type corresponding to each historical construction data is the cantilever assembly method. The historical crane distance data is represented as the distance between the midpoint of the line segment formed by the crane's center of gravity and the endpoint of the main arch's net span. The historical crane model refers to the model of the crane used in the construction of the historical cantilever bridge; The historical environmental data includes historical wind speed levels and historical average hourly rainfall. The construction module extracts any historical construction data and constructs historical sample data. The expression for the historical sample data is: R={L,M,D,W, , , ,H,V,T}; Where R represents historical sample data, L represents historical net span of the main arch ring, M represents historical crane model, D represents historical crane distance data, and W represents rated wind force level. This represents the historical average rainfall. For the rated rotation angle, H is the rated pitch angle, V is the rated lifting height, and T is the historical operating time. The logic for constructing the historical sample data also includes establishing a five-dimensional grid with the first value being the distance step size, the second value being the lifting weight step size, the third value being the angle step size, and the fourth value being the wind force level step size. Historical sample data of the same main arch ring net span are selected, and the 90% and 10% values ​​of the corresponding historical crane distance data are statistically analyzed and denoted as D90 and D10 respectively. A two-dimensional safety zone is generated with D10 as the inner boundary and D90 as the outer boundary, which is denoted as the first zone. Within each five-dimensional grid corresponding to the net span of the main arch ring, the total number of historical sample data is counted and recorded as the first quantity. The historical working time corresponding to each historical sample data is sorted in descending order, and the historical working time with the largest value is selected to calculate the unit confidence. Using the unit confidence level as the pixel value, a two-dimensional safety heat map of historical lifting weight and historical rotation angle is obtained, which is denoted as the first image; Obtain the rated wind force level corresponding to each crane model in the clear span of the main arch ring. When the wind force level of the historical sample data is less than or equal to the corresponding rated wind force level, set the first safety factor to 0.8; otherwise, set the first safety factor to 0.

4. Obtain the historical average hourly rainfall in the clear span of the main arch ring. Set the first rainfall amount as the rainfall threshold. Calculate the first difference between the historical average hourly rainfall amount and the first rainfall amount. Set the fifth value as the first difference threshold. Compare the first difference with the fifth value. When the first difference is less than or equal to the fifth value, set the second safety factor to 0.8; otherwise, set the second safety factor to 0.

4. Calculate the first average value of the first safety factor and the second safety factor. The net span of the main arch ring and the crane model in the historical sample data of the net span of the main arch ring are combined to obtain the first combination. The sample points of each corresponding pitch angle and lifting height in any first combination are extracted and convex hull operation is performed to obtain the boundary polygon. The upper boundary function of the boundary polygon is fitted by the least squares method. The value of the first proportion of the boundary function is set as the upper limit of the allowable range, and the first proportion is distributed between 0 and 1; The analysis module analyzes the construction environment and registers the crane parameters based on the analysis results to obtain the registration results.

2. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 1, characterized in that: The construction module obtains historical crane models, retrieves the crane database, inputs historical crane models into the crane database, and matches the rated lifting capacity, rated slewing angle, rated pitch angle, rated wind force level, and rated lifting height corresponding to the historical crane models. The rated slewing angle is the maximum clockwise rotation angle or the maximum counterclockwise rotation angle of the uppermost structure of the enclosure. The rated pitch angle is the maximum pitch angle or the maximum tilt angle of the crane boom. The rated wind force level is the dividing wind force level between continued operation and cessation of operation when the crane is operating outdoors, and the rated wind force level is the maximum wind force level for continued operation. The rated lifting height is the maximum vertical distance that the hook can be raised.

3. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 1, characterized in that: The construction module obtains construction parameters based on the current construction data and the crane construction spectrum. The current construction data includes the current main arch clear span, current crane model, current pitch angle, current lifting capacity, current wind force level, and current hourly average rainfall. The construction parameters include the current crane distance data, current slewing angle, current construction time status, and current lifting height.

4. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 1, characterized in that: The construction module calculates the unit confidence and sets the upper boundary function; The expression for calculating the unit confidence level is as follows: ； Where C is the unit confidence level, N is the first quantity, and T max The longest historical working time is represented by the largest value, where e is a natural number. The expression for the upper boundary function is: ； Where a, b, and c are constants.

5. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 1, characterized in that: The logic of the construction module to build the crane construction spectrum based on historical construction data includes: Acquire the distance data of the first region and the current crane. If the current crane distance data is not distributed in the corresponding first region, send a distance over-limit warning. Do not send a distance over-limit warning until the current crane distance data is distributed in the corresponding first region. The system acquires the first image, the current lifting weight, and the current slewing angle. Based on the first image and the current lifting weight, it obtains the corresponding theoretical slewing angle. The system compares the current slewing angle with the theoretical slewing angle. If the current slewing angle is different from the theoretical slewing angle, a slewing angle over-limit warning is sent. The system continues to send a slewing angle over-limit warning until the current slewing angle is the same as the theoretical slewing angle. Obtain the first average value. If the first average value is not 0.8, send a weather over-limit warning. Continue to send a weather over-limit warning until the weather improves and the first average value is 0.

8. Obtain the upper boundary function, current pitch angle, and current lift height. Input the current pitch angle into the upper boundary function to obtain the current function result. Calculate the first product of the current function result and the first ratio, and record it as the allowable upper limit. Compare the current lift height with the allowable upper limit. If the current lift height is greater than the allowable upper limit, send a lift height over-limit message. Continue until the lift height is less than or equal to the allowable upper limit, then do not send a lift height over-limit message.

6. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 3, characterized in that: The construction module obtains construction parameters, and the logic for obtaining the construction parameters includes: Obtain the current net span of the main arch ring, retrieve the first region corresponding to the current net span of the main arch ring, select any point within the first region, and obtain the straight-line distance between the crane boom and the nearest high-voltage power pole, the straight-line distance between the wire rope and the nearest high-voltage power pole, and the straight-line distance between the hoisted object and the nearest high-voltage power pole at that point. Record these as safety distances. If any safety distance is less than the standard safety distance, jump to the next point. If all safety distances are greater than or equal to the standard safety distance, set the point as the location of the crane's center of gravity. Set the straight-line distance between the location of the crane's center of gravity and the midpoint of the line segment formed between the endpoints of the net span of the main arch ring as the current crane distance data. Adjust the position of the crane relative to the net span of the main arch ring based on the current crane distance data. Obtain the current lifting capacity, match the corresponding current slewing angle based on the first image, and adjust the slewing angle of the crane according to the current slewing angle; Obtain the current pitch angle, match the corresponding current lifting height according to the upper boundary function, and adjust the lifting height of the crane according to the current lifting height; Obtain the current wind force level and the current hourly average rainfall. Based on the calculation logic of the first average value, obtain the first average value corresponding to the current wind force level and the current hourly average rainfall. When the current first average value is equal to 0.8, set the current construction time status to "construction can be carried out today". When the current first average value is not equal to 0.8, set the current construction time status to "construction cannot be carried out today".

7. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 1, characterized in that: The analysis module analyzes the construction environment, and the logic for analyzing the construction environment includes: Acquire videos of the construction environment, construct a digital twin scene based on the videos, align the digital twin scene with the timeline of the on-site video stream, pre-mark K AR target positions in the digital twin scene, and have operators on-site use an AR headset to place real targets onto corresponding physical points. Measure the global coordinates of the real targets using a total station, and bridge the virtual coordinate system (VCS) and the on-site coordinate system (GCS) using the targets to obtain a coarse transformation matrix. Activate the depth camera in the AR headset to scan points on the crane boom, obtaining a point cloud, and then integrate the point cloud with the digital twin scene. ICP iteration is performed, allowing the crane to perform slewing and hoisting actions in sequence, and continuously comparing the error between the virtual boom and the real boom. The error is represented by the straight-line distance between the center of gravity of the virtual boom and the center of gravity of the real boom. The sixth value is set as the deviation threshold, and the error is compared with the sixth value. When the error is greater than the sixth value, ICP iteration is repeated until the error is less than or equal to the sixth value. Then, ICP iteration is stopped, and the registration step is completed. The coarse transformation matrix at this time is set as the registration matrix, and the registration matrix is ​​set as the analysis result. Based on the analysis results, the crane parameters are registered to obtain the registration results. The crane parameters include crane distance data, slewing angle, pitch angle and lifting height.

8. The cantilever bridge construction risk monitoring system based on digital twin as described in claim 7, characterized in that: The analysis module constructs a digital twin scenario, and the construction logic of the digital twin scenario includes: The construction environment video is extracted into continuous frames at a fixed time step. The motion recovery structure algorithm is run on each frame to automatically track key feature points and generate sparse point clouds and camera pose sequences. Using the sparse point cloud as a seed, a multi-view stereo algorithm is used for diffusion matching to obtain a dense point cloud with color information. Poisson reconstruction is performed on the dense point cloud to generate a renderable 3D mesh model. An instance segmentation network is used to label "sky, ground, crane, and utility pole" in the mesh to form a semantic shell. For the segmented "crane" area, the coarse mesh is replaced with a CAD model. The local coordinate system of the reconstructed model is transformed to the design coordinate system through on-site control points. Each frame is assigned a timestamp from the original video to complete the construction of the digital twin scene. The logic for generating the registration matrix includes: The AR head-mounted display identifies the center pixel coordinates of the real target, and the total station automatically aims at the same target and outputs global coordinates, forming a correspondence between the pixel and the global coordinates. A coarse transformation matrix is ​​obtained using a three-point coplanar algorithm. The AR depth camera sweeps the crane boom, aggregates depth frames, generates a local point cloud of the boom, and automatically filters out outliers. Each vertex of the point cloud is matched with the nearest neighbor triangle in the digital twin scene to establish a correspondence between the observation point and the model face. The ICP algorithm is used to iteratively update the coarse transformation matrix. After each iteration, the crane performs a slewing motion before a lifting motion, and the error between the virtual boom center of gravity and the real boom center of gravity is calculated. If the error is greater than the sixth value, the ICP iteration is re-executed; otherwise, the current coarse transformation matrix is ​​locked and set as the registration matrix.

Images