Machine learning-based method and system for dynamically regulating and controlling installation precision of obliquely-spanned steel box arch bridge

Abstract

The invention relates to the technical field of construction control, and particularly discloses a method and a system for dynamically regulating and controlling the installation precision of an inclined-span steel box arch bridge based on machine learning. Wherein the real-time data comprises real-time stress of the arch rib, real-time deformation of the arch rib, real-time environment wind speed and real-time hoisting parameters; constructing a prediction model capable of analyzing the coupling effect of the wind load and the hoisting unbalance load based on a large amount of historical construction data; inputting real-time data into the prediction model to obtain a real-time coupling effect analysis result; outputting a hoisting sequence optimization instruction based on the decision engine, the coupling effect analysis result and a preset mechanical constraint condition of the obliquely-spanning steel box arch bridge; based on the deviation between the real-time deformation amount of the arch rib and a preset installation precision threshold value, a cable force grading adjustment scheme is generated; construction efficiency of the inclined-span steel box arch bridge is effectively improved, construction safety and installation precision are guaranteed, and stability and reliability of the bridge structure are guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of construction control, and in particular relates to a method and system for dynamic regulation and control of installation precision of a skew-span steel box arch bridge based on machine learning. BACKGROUND

[0002] Skew-span steel box arch bridges are widely used in modern transportation infrastructure construction due to their unique appearance and good mechanical properties. However, the installation process of skew-span steel box arch bridges is extremely challenging, especially the high requirement for installation precision, which not only affects the appearance of the bridge, but also directly relates to the structural stability and carrying capacity of the bridge, and is also related to the safety and durability of the bridge in the subsequent use process. In the construction process of skew-span steel box arch bridges, various complex factors interact with each other to affect the installation precision. Wind load, as an important natural factor, can change the stress state of the structure and cause vibration and displacement; and hoisting eccentric load may cause uneven stress of the structure due to the complexity of construction operation, and further affect the installation precision. In addition, as the construction progresses, the installation deviations in the early stages may accumulate in the subsequent construction, thereby increasing the difficulty of installation precision control. At present, applying machine learning technology to the installation precision regulation and control of skew-span steel box arch bridges has great potential. This not only helps to improve the quality of bridge construction and reduce the maintenance cost caused by installation precision problems, but also promotes the intelligent and fine construction technology of bridges, and has broad application prospects in the future bridge construction field.

[0003] However, the existing installation precision regulation and control technology of skew-span steel box arch bridges cannot effectively evaluate the comprehensive influence of complex factors and obtain real-time coupling effect analysis results, and cannot timely adjust the construction sequence to reduce the influence of adverse factors, thereby resulting in the inability to generate a scientific and reasonable cable force grading adjustment scheme, and making it difficult to dynamically and accurately regulate and control the installation precision, which ultimately affects the overall installation quality of skew-span steel box arch bridges.

[0004] Therefore, the present application proposes a method and system for dynamic regulation and control of installation precision of skew-span steel box arch bridges based on machine learning. SUMMARY

[0005] The application provides a machine learning-based installation precision dynamic regulation method and system for a skew-span steel box arch bridge, which can deeply mine potential relationships between various types of data in the construction process by means of the powerful data processing and analysis capabilities of machine learning, and realize accurate regulation of installation precision. A prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed using a large amount of historical construction data, and real-time coupling effect analysis results are obtained by inputting real-time construction data such as real-time stress of the arch rib, deformation, environmental wind speed and hoisting parameters into the model. The hoisting sequence optimization instruction is outputted by combining the decision engine and the preset mechanical constraint condition, and the cable force grading adjustment scheme is generated based on the deviation between the real-time deformation of the arch rib and the preset precision threshold, realizing the whole-process scientific management from data collection, effect prediction to construction instruction optimization and precision regulation, effectively improving the construction efficiency of the skew-span steel box arch bridge, ensuring the construction safety and installation precision, and ensuring the stability and reliability of the bridge structure.

[0006] The application provides a machine learning-based installation precision dynamic regulation method for a skew-span steel box arch bridge, comprising:

[0007] Real-time data in the construction process of the skew-span steel box arch bridge are collected, wherein the real-time data include real-time stress of the arch rib, real-time deformation of the arch rib, real-time environmental wind speed and real-time hoisting parameters;

[0008] A prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed based on a large amount of historical construction data;

[0009] Real-time data are inputted into the prediction model to obtain real-time coupling effect analysis results;

[0010] Based on the decision engine, the coupling effect analysis results and the preset mechanical constraint condition of the skew-span steel box arch bridge, a hoisting sequence optimization instruction is outputted;

[0011] Based on the deviation between the real-time deformation of the arch rib and the preset installation precision threshold, a cable force grading adjustment scheme is generated.

[0012] Preferably, the real-time data in the construction process of the skew-span steel box arch bridge are collected, including:

[0013] Real-time axial stress and real-time radial stress of the arch rib in the construction process of the skew-span steel box arch bridge are collected based on fiber bragg grating sensors arranged longitudinally and uniformly along the arch rib of the skew-span steel box arch bridge as the real-time stress of the arch rib;

[0014] Three-dimensional coordinates of the arch rib in the construction process of the skew-span steel box arch bridge are collected in real time based on laser trackers arranged on both sides of the arch rib piers, and transverse bending deformation and axial deviation are calculated based on the collected three-dimensional coordinates in real time as the real-time deformation of the arch rib;

[0015] Instantaneous wind speed and direction in the construction process of the skew-span steel box arch bridge are collected based on wind speed sensors arranged on the top of the arch rib as the real-time environmental wind speed;

[0016] Collecting real-time segment physical parameters, real-time hoisting posture parameters and real-time motion state parameters in the construction process of the inclined cross steel box arch bridge as real-time hoisting parameters.

[0017] Preferably, a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed based on a large amount of historical construction data, including:

[0018] The combination of wind load and hoisting eccentric load in a large amount of historical construction data is simulated to obtain preset group sample data;

[0019] All group sample data is divided into a training set and a validation set at a preset ratio, and a neural network model is trained based on the training set and the validation set, while minimizing the mean square error of the predicted deformation and the actual deformation through the Adam optimizer until the mean square error does not exceed a preset error threshold, thereby obtaining a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load.

[0020] Preferably, the construction process of the decision engine further includes:

[0021] Based on the actual deformation and the allowable deformation amount and the hoisting construction period delay rate, a reward function of a reinforcement learning algorithm is generated;

[0022] A decision engine is constructed based on the reward function of the reinforcement learning algorithm.

[0023] Preferably, based on the decision engine and the coupling effect analysis result and the preset mechanical constraint condition of the inclined cross steel box arch bridge, a hoisting sequence optimization instruction is output, including:

[0024] A hoisting sequence thread is generated based on each selectable hoisting sequence;

[0025] The reward value of each hoisting sequence thread is determined based on the reward function of the reinforcement learning algorithm in the decision engine, and a hoisting sequence optimization instruction is generated based on the hoisting sequence thread corresponding to the maximum reward value.

[0026] Preferably, based on the deviation of the arch rib real-time deformation amount from the preset installation precision threshold, a cable force grading adjustment scheme is generated, including:

[0027] The arch rib real-time deflection value is determined based on the arch rib real-time deformation amount;

[0028] The preset maximum preset deflection value is taken as the preset installation precision threshold;

[0029] When the deviation of the arch rib real-time deflection value from the preset installation precision threshold does not exceed a first deviation threshold, a first proportional initial tensioning is performed as the cable force grading adjustment scheme;

[0030] When the deviation of the real-time deflection value of the arch rib from the preset installation accuracy threshold exceeds the first deviation threshold and does not exceed the second deviation threshold, a first proportion to a second proportion staged tension is performed as a cable force staged adjustment scheme;

[0031] When the deviation of the real-time deflection value of the arch rib from the preset installation accuracy threshold exceeds the second deviation threshold, a maximum proportion tension and a recalculation of the cable force correction amount of all hangers calculated based on the matrix influence method are performed as a cable force staged adjustment scheme.

[0032] Preferably, the cable force correction amount of all hangers is calculated based on the matrix influence method, including:

[0033] The influence matrix is constructed with the hanger number of the skew-span steel box arch bridge as the row vector and the arch rib key control section as the column vector;

[0034] The basis deviation vector is generated based on the deviation of the real-time deflection value of all arch rib key control sections from the preset installation accuracy threshold;

[0035] The basis deviation vector is corrected based on the preset wind-induced correction coefficient and the temperature correction coefficient to obtain the corrected deviation vector;

[0036] The multi-dimensional cable force correction vector is solved by the regularized least squares method based on the influence matrix and the corrected deviation vector;

[0037] The multi-dimensional cable force correction vector is subjected to safety constraint filtering, balanced constraint filtering, and construction feasibility constraint filtering to obtain the cable force correction amount of all hangers.

[0038] Preferably, the multi-dimensional cable force correction vector is subjected to balanced constraint filtering, including:

[0039] A finite element model in the construction process of the skew-span steel box arch bridge is generated, and a force flow monitoring section is selected every preset distance along the longitudinal direction of the arch rib in the finite element model, a plurality of feature points are selected at each force flow monitoring section, the axial normal stress direction, the radial normal stress direction, and the shear stress direction of each feature point are extracted, and the force flow direction angle of each feature point is calculated based on the axial normal stress direction, the radial normal stress direction, and the shear stress direction of each feature point;

[0040] The continuous section with the same structural feature or stress feature in the finite element model is regarded as the to-be-analyzed section;

[0041] The force flow sharp turn section is screened out in all to-be-analyzed sections based on the force flow direction angle of all feature points in the finite element model;

[0042] The strength requirement of balanced constraint is determined based on the distribution position and sharp turn degree of all force flow sharp turn sections;

[0043] The multi-dimensional force correction vector is balanced and constrained based on the strength requirement of the balanced constraint.

[0044] Preferably, the force flow sharp turn section is screened out in all analysis sections based on the force flow direction angle of all feature points in the finite element model, including:

[0045] The difference value of the force flow direction angle of the feature points at the same position in each group of adjacent force flow monitoring sections and the ratio of the adjacent section spacing are taken as the corresponding direction angle change rate;

[0046] The standard deviation of the force flow direction angle of all feature points in the analysis section is defined as the force flow continuity index;

[0047] The ratio of the number of all direction angle abrupt change points to the number of all feature points in the analysis section is taken as the extreme point ratio;

[0048] The force flow sharp turn section is screened out in all analysis sections based on the direction angle change rate and the force flow continuity index and the extreme point ratio of all analysis sections.

[0049] The present application provides a kind of based on machine learning's oblique span steel box arch bridge installation precision dynamic regulation and control system, including:

[0050] Data acquisition module is used to collect real-time data in the construction process of oblique span steel box arch bridge, wherein real-time data includes arch rib real-time stress, arch rib real-time deformation, real-time environmental wind speed and real-time hoisting parameter;

[0051] Model construction module is used to construct prediction model capable of analyzing wind load and hoisting eccentric load coupling effect based on a large number of historical construction data;

[0052] Coupling analysis module is used to input real-time data into prediction model to obtain real-time coupling effect analysis result;

[0053] Sequential optimization module is used to output hoisting sequence optimization instruction based on decision engine, coupling effect analysis result and preset mechanical constraint condition of oblique span steel box arch bridge;

[0054] Scheme generation module is used to generate cable force grading adjustment scheme based on the deviation of arch rib real-time deformation and preset installation precision threshold value.

[0055] The beneficial effects generated by the present application relative to the prior art are: the potential relationship between various types of data in the construction process can be deeply mined by means of the powerful data processing and analysis capability of machine learning, and the installation precision can be accurately regulated. A prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed by using a large amount of historical construction data, and real-time coupling effect analysis results are obtained by inputting real-time construction data such as real-time stress, deformation, environmental wind speed and hoisting parameters of the arch rib into the model, and then the hoisting sequence optimization instruction is output by combining the decision engine and the preset mechanical constraint condition, and the cable force grading adjustment scheme is generated according to the real-time deformation of the arch rib and the preset precision threshold deviation, realizing the whole-process scientific management from data collection, effect prediction to construction instruction optimization and precision regulation, effectively improving the construction efficiency of the skew-span steel box arch bridge, ensuring the construction safety and installation precision, and ensuring the stability and reliability of the bridge structure.

[0056] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and achieved by the structure specifically pointed out in the present application document.

[0057] The technical solutions of the present application will be further described in detail below by means of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0058] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0059] Figure 1 A schematic diagram of the installation precision dynamic regulation method based on machine learning of the skew-span steel box arch bridge in the embodiment of the present application;

[0060] Figure 2 A real-time data acquisition flowchart in the construction process of the skew-span steel box arch bridge in the embodiment of the present application;

[0061] Figure 3 A prediction model construction method flowchart in the embodiment of the present application. DETAILED DESCRIPTION

[0062] The preferred embodiments of the present application will be described below in conjunction with the drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not limit the present application.

[0063] As shown in Figure 1 The present application provides an installation precision dynamic regulation method based on machine learning of a skew-span steel box arch bridge, comprising:

[0064] Collect real-time data in the construction process of the skew-span steel box arch bridge, wherein the real-time data include real-time stress of the arch rib, real-time deformation of the arch rib, real-time environmental wind speed and real-time hoisting parameters;

[0065] Based on a large amount of historical construction data, a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed;

[0066] The real-time data are input into the prediction model to obtain real-time coupling effect analysis results;

[0067] Based on the decision engine, the coupling effect analysis results and the preset mechanical constraint conditions of the skew-span steel box arch bridge, hoisting sequence optimization instructions are output;

[0068] Based on the deviation of the real-time deformation of the arch rib from the preset installation precision threshold, a cable force grading adjustment scheme is generated.

[0069] In this embodiment, the skew-span steel box arch bridge and its construction process: the skew-span steel box arch bridge is a structure applied in bridge construction, which has a unique shape and good mechanical properties. The construction process involves multiple links, including component hoisting, cable force adjustment, etc., and is affected by various factors during the process, such as environmental factors (wind load) and construction operation factors (hoisting eccentric load), which have very high requirements for installation precision.

[0070] In this embodiment, a large amount of historical construction data: refers to the data accumulated in the construction of numerous skew-span steel box arch bridges in the past, which covers various aspects of construction information, such as arch rib stress and deformation data at different construction stages, environmental wind speed and hoisting parameters at each stage, and actual deformation results under the combined working conditions of wind load and hoisting eccentric load, etc.

[0071] In this embodiment, the prediction model for analyzing the coupling effect of wind load and hoisting eccentric load: used to analyze the effect produced by the combined action of wind load and hoisting eccentric load, i.e. the influence of the combined action of wind load and hoisting eccentric load on the installation precision of the arch rib under high-wind environment on the plateau.

[0072] In this embodiment, the real-time coupling effect analysis results: after inputting the data such as real-time stress of the arch rib, real-time deformation of the arch rib, real-time environmental wind speed and real-time hoisting parameters collected in the construction process of the skew-span steel box arch bridge into the prediction model for analyzing the coupling effect of wind load and hoisting eccentric load, the analysis results about the effect (or the influence on the installation precision of the arch rib) under the combined action of current wind load and hoisting eccentric load output by the model.

[0073] In this embodiment, the decision engine: used to combine the real-time coupling effect analysis results and the preset mechanical constraint conditions to assist in making decisions about hoisting sequence.

[0074] In this embodiment, the preset mechanical constraint condition of the skew-span steel box arch bridge: the limit condition preset in the design and construction planning stage of the skew-span steel box arch bridge, based on the bridge structure mechanics and construction safety, such as the maximum stress that the structure can withstand, the deformation range, the carrying capacity of each part, etc. These conditions limit various operations in the construction process.

[0075] In this embodiment, the hoisting sequence optimization instruction: an instruction for guiding the hoisting sequence in the construction of the skew-span steel box arch bridge. The optimization instruction example is as follows:

[0076] For the skew-span steel box arch bridge, symmetric and alternating hoisting is adopted, that is, left arch foot segment→right arch foot segment→left 1 / 4 span segment→right 1 / 4 span segment;

[0077] Dynamic adjustment combined with real-time monitoring data (such as preferentially hoisting the segment on the opposite side to balance the stress when the stress of the arch rib on one side exceeds 250 MPa);

[0078] The goal is to control the cumulative deformation of the arch rib to be less than or equal to 28 mm, and to ensure the accuracy of the 20 mm expansion joint reserved at the closure opening.

[0079] As shown in Figure 2 In order to comprehensively and accurately obtain real-time data in the construction process of the skew-span steel box arch bridge, real-time data in the construction process of the skew-span steel box arch bridge is collected, including:

[0080] Based on the fiber Bragg grating sensors uniformly arranged along the arch rib of the skew-span steel box arch bridge, the real-time axial stress and real-time radial stress of the arch rib in the construction process of the skew-span steel box arch bridge are collected as the real-time stress of the arch rib;

[0081] Based on the laser tracker arranged on both sides of the pier of the arch rib, the three-dimensional coordinates of the arch rib in the construction process of the skew-span steel box arch bridge are collected in real time, and the transverse bending deformation and axial deviation based on the real-time collected three-dimensional coordinates are calculated as the real-time deformation of the arch rib;

[0082] Based on the wind speed sensor arranged at the top of the arch rib, the instantaneous wind speed and wind direction in the construction process of the skew-span steel box arch bridge are collected as the real-time environmental wind speed;

[0083] The real-time segment physical parameters, real-time hoisting posture parameters, and real-time motion state parameters in the construction process of the skew-span steel box arch bridge are collected as real-time hoisting parameters.

[0084] In this embodiment, the fiber grating sensors are uniformly arranged along the longitudinal direction of the arch rib of the skew-span steel box arch bridge. The axial and radial stresses of the arch rib are collected in real time during the construction process, such as the stress change value recorded by the sensor when the components are hoisted. Laser trackers are arranged on both sides of the pier of the arch rib, and the three-dimensional coordinates of the arch rib are collected in real time during the construction, such as the coordinate change recorded by the tracker when the segments are installed. Then, the transverse bending deformation and axial deviation are calculated based on the three-dimensional coordinates of the arch rib collected by the laser tracker in real time, such as the deformation and deviation calculated after the segment is installed to evaluate the construction. In addition, the top of the arch rib refers to the highest point of the arch rib of the skew-span steel box arch bridge, which can be used for installing equipment such as wind speed sensors.

[0085] This embodiment collects the physical parameters of the segment (such as the weight of a single segment, the cross-sectional size, and the center of gravity coordinates), the hoisting attitude parameters (such as the stage lifting angle: vertical inclination, lateral deviation angle; the force distribution of the lifting point, that is, the tension value of each sling), and the motion state parameters (such as the lifting speed, the horizontal movement speed, the relative displacement during high-altitude butt joint, etc.) during the construction of the skew-span steel box arch bridge.

[0086] As shown in Figure 3 To construct a prediction model that can analyze the coupling effect of wind load and hoisting eccentric load, a prediction model that can analyze the coupling effect of wind load and hoisting eccentric load is constructed based on a large amount of historical construction data, including:

[0087] The combination of wind load and hoisting eccentric load in a large amount of historical construction data is simulated to obtain a preset group of sample data;

[0088] All group sample data is divided into a training set and a validation set at a preset ratio. The neural network model is trained based on the training set and the validation set, and the mean square error of the predicted deformation and the actual deformation is minimized through the Adam optimizer until the mean square error does not exceed the preset error threshold, and a prediction model that can analyze the coupling effect of wind load and hoisting eccentric load is obtained.

[0089] This embodiment obtains a set of pre-set sample data by simulating the combination of wind load and hoisting eccentric load in a large amount of historical construction data, such as simulating the combination of different wind forces and hoisting eccentric loads to generate corresponding samples.

[0090] The preset ratio mentioned in this embodiment is a certain ratio value set in advance, which may be used for sample division and the like, such as dividing the sample into a training set and a test set at a preset ratio of 7:3.

[0091] In this embodiment, the training set and the validation set are used to train the neural network model, and the model parameters are continuously adjusted by means of the Adam optimizer to minimize the mean square error between the predicted deformation and the actual deformation. When the mean square error is less than or equal to the preset error threshold, a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is obtained, for example, the training model predicts the deformation of the bridge under a specific working condition, and the optimization is performed until the error meets the requirements.

[0092] The predicted deformation in this embodiment refers to the deformation calculated by the model, and the actual deformation refers to the deformation measured in the actual construction scene. For example, the model predicts that the bridge deformation is 5 cm, and the actual measurement deformation is 4.8 cm.

[0093] The preset error threshold in this embodiment is a pre-set error limit value for judging whether the model training meets the requirements. For example, it is set to 0.5, and the model mean square error is not more than 0.5, which meets the requirements.

[0094] In order to generate a reward function based on the actual deformation, the allowable deformation amount and the hoisting construction period delay rate, and further construct a decision engine to realize more scientific construction decision, the construction process of the decision engine is proposed, which also includes:

[0095] Generating a reward function of a reinforcement learning algorithm based on the actual deformation, the allowable deformation amount and the hoisting construction period delay rate;

[0096] Constructing a decision engine based on the reward function of the reinforcement learning algorithm.

[0097] The actual deformation in this embodiment refers to the deformation value of the arch rib and other structural components actually occurring in the actual construction process of the skew-span steel box arch bridge, which reflects the actual shape change of the structure at the construction time. For example, after installing a key component, it is measured that the actual deformation of the arch rib is 3 mm.

[0098] The allowable deformation amount is pre-set according to the bridge design standard, safety specification and the like, and is the maximum deformation range that the structural component can withstand in the construction process. It is an important index to ensure the safety and function of the bridge structure. For example, the design specifies that the allowable deformation amount of a certain section of arch rib is 5 mm at a specific construction stage.

[0099] The hoisting construction period delay rate is used to measure the proportion of the actual construction period compared with the planned construction period in the hoisting construction process of the skew-span steel box arch bridge. For example, the planned hoisting construction period is 30 days, and the actual hoisting construction period is 35 days, so the hoisting construction period delay rate = (35-30) ÷ 30 x 100% ≈ 16.7%.

[0100] In this embodiment, the actual deformation, the allowable deformation amount and the hoisting construction period delay rate are used as variables to construct the reward function in the reinforcement learning algorithm:

[0101] Reward = 0.6 x (1 - actual deformation / allowable deformation) + 0.4 x (1 - delay rate of construction period).

[0102] This embodiment is based on the reward function of the reinforcement learning algorithm constructed above to build a decision engine. The decision engine will optimize and adjust various decisions during the construction process according to different reward values fed back by the reward function, such as adjusting the lifting sequence and construction process, to achieve better construction results. For example, according to the feedback of the reward function, the decision engine decides to increase temporary support in a lifting link to reduce deformation and avoid delay of construction period.

[0103] To optimize the lifting sequence, a lifting sequence optimization instruction is output based on the decision engine, the coupling effect analysis results, and the preset mechanical constraint conditions of the skew steel box arch bridge, including:

[0104] Generating a lifting sequence thread based on each selectable lifting sequence;

[0105] Determining the reward value of each lifting sequence thread based on the reward function of the reinforcement learning algorithm in the decision engine, and generating a lifting sequence optimization instruction based on the lifting sequence thread corresponding to the maximum reward value.

[0106] The selectable lifting sequence in this embodiment refers to the arrangement of multiple feasible sequences for the lifting of each component during the construction of the skew steel box arch bridge. For example, for three different arch rib segments A, B, and C, the selectable lifting sequence may include A-B-C, A-C-B, B-A-C, and so on.

[0107] In this embodiment, each selectable lifting sequence is converted into an independent lifting sequence thread. The thread can be understood as a flow unit that performs a specific lifting sequence task. Each thread simulates or performs lifting operations according to the corresponding lifting sequence. For example, for the A-B-C selectable lifting sequence, a dedicated thread is generated to perform the relevant simulation or operation according to this sequence.

[0108] In this embodiment, the reward function based on the reinforcement learning algorithm in the decision engine is used to evaluate each lifting sequence thread and determine the corresponding reward value of each thread. The reward value reflects the rationality and degree of excellence of the lifting sequence in actual construction. For example, for the A-B-C lifting sequence thread, if the actual deformation is small and does not exceed the allowable deformation, and the lifting delay rate is low, the reward function will give a higher reward value; otherwise, if the deformation is too large or the construction period is delayed seriously, the reward value will be lower.

[0109] The decision engine based on the reinforcement learning algorithm regards the lifting sequence as a "sequence of actions" and evaluates the advantages and disadvantages of different sequences through the reward function. The specific rules are:

[0110] Base sequence library: preset 3 types of initial sequences (symmetric alternation, from arch spring to arch apex, from arch apex to arch spring), such as "left arch spring segment -> right arch spring segment -> left 1 / 4 span segment -> right 1 / 4 span segment" (symmetric alternation);

[0111] Dynamic adjustment logic:

[0112] When the coupling effect analysis result is "low risk" (wind speed < 5 m / s, eccentric load < 3%): prefer to select "efficiency priority sequence" (continuous hoisting of adjacent segments on the same side, shorten the construction period);

[0113] When the analysis result is "medium risk" (wind speed 5-10 m / s, eccentric load 3%-5%): switch to "symmetric balance sequence" (left and right alternation hoisting, each side hoists 1 segment and then changes side), reduces asymmetric load;

[0114] When the analysis result is "high risk" (wind speed ≥ 10 m / s or eccentric load ≥ 5%): trigger "pause-dispersed sequence" (first hoist the hoisted segment to the temporary fixed position, then hoist the segment on the opposite side after 30 minutes, to avoid load superposition).

[0115] In order to formulate different proportion tensioning schemes as cable force grading adjustment schemes according to different deviations of arch rib real-time deflection value and preset installation accuracy threshold value, realize accurate cable force adjustment, based on the deviation of arch rib real-time deformation and preset installation accuracy threshold value, a cable force grading adjustment scheme is generated, including:

[0116] Determine the arch rib real-time deflection value based on the arch rib real-time deformation;

[0117] Take the preset maximum preset deflection value as the preset installation accuracy threshold value;

[0118] When the deviation of the arch rib real-time deflection value and the preset installation accuracy threshold value does not exceed the first deviation threshold value, the first proportion initial tensioning is executed as the cable force grading adjustment scheme;

[0119] When the deviation of the arch rib real-time deflection value and the preset installation accuracy threshold value exceeds the first deviation threshold value and does not exceed the second deviation threshold value, the first proportion to the second proportion grading tensioning is executed as the cable force grading adjustment scheme;

[0120] When the deviation of the arch rib real-time deflection value and the preset installation accuracy threshold value exceeds the second deviation threshold value, the maximum proportion tensioning and the cable force correction amount of all suspender calculated based on the matrix influence method are executed as the cable force grading adjustment scheme.

[0121] In this embodiment, the real-time deformation amount of the arch rib: In the construction process of the skew-span steel box arch bridge, the amount of change in the spatial position of the arch rib is obtained in real time, such as the lateral bending deformation amount and the axial deviation of the arch rib calculated by collecting data by the laser tracker.

[0122] In this embodiment, the real-time deflection value of the arch rib is determined based on the real-time deformation amount of the arch rib: the real-time three-dimensional coordinates of the key control section of the arch rib (such as the arch top, 1 / 4 span, and arch foot) are collected by the laser tracker, and the difference between the vertical coordinates and the design coordinates is calculated, that is, the real-time deflection value of the corresponding section. The difference of the arch top section is taken as the representative deflection value of the arch rib under the current state.

[0123] In this embodiment, the real-time deflection value of the arch rib: that is, the value reflecting the current vertical deformation degree of the arch rib calculated according to the real-time deformation amount of the arch rib.

[0124] In this embodiment, the preset maximum preset deflection value: the maximum deflection value that the arch rib can withstand is preset in the bridge design stage according to the use function, structural characteristics, and safety standards of the bridge.

[0125] In this embodiment, the first deviation threshold: such as 5mm.

[0126] In this embodiment, the first proportional initial tensioning is performed: when the deviation between the real-time deflection value of the arch rib and the preset maximum preset deflection value reaches the first deviation threshold, the suspender is initially tensioned, and the tensioning degree is performed according to the preset first proportion. This step aims to preliminarily control the deformation of the arch rib by adjusting the suspender cable force, for example, the first proportion is 30%, that is, the suspender is tensioned to 30% of the design cable force.

[0127] In this embodiment, the second deviation threshold: such as 15mm.

[0128] In this embodiment, the first proportion to the second proportion is executed: when the deviation between the real-time deflection value of the arch rib and the preset maximum preset deflection value reaches the second deviation threshold, the suspender is continuously tensioned on the basis of the first proportion initial tensioning, and the tensioning degree is gradually increased from the first proportion to the second proportion. The hierarchical way is adopted to more finely control the deformation of the arch rib and the distribution of the internal force of the structure. For example, gradually tensioning from 30% to 70%.

[0129] In this embodiment, the maximum proportion tensioning is performed: if the real-time deflection value of the arch rib is not effectively controlled after the initial tensioning and the hierarchical tensioning, and approaches or reaches the preset maximum preset deflection value, the suspender is tensioned at the maximum proportion to adjust the deformation of the arch rib as much as possible to return it to the safe range. The maximum proportion is usually 100% of the design cable force.

[0130] In this embodiment, the cable force correction amount of all hangers calculated based on the matrix influence method is used for re-tuning: the cable force correction amount of all hangers is calculated by using the matrix influence method, which considers the mutual influence between the hangers and the effect on the overall structure of the arch rib. Then, according to the calculated cable force correction amount, the hanger cable force is adjusted again to further optimize the stress state and deformation of the arch rib, and to ensure that the bridge structure meets the design requirements. For example, it is calculated by the matrix influence method that the cable force of a certain hanger needs to be increased by 5kN, and re-tuning operation is performed.

[0131] In order to calculate the cable force correction amount of all hangers, it is proposed to calculate the cable force correction amount of all hangers based on the matrix influence method, including:

[0132] The influence matrix is constructed by taking the hanger number of the skew-span steel box arch bridge as the row vector and the arch rib key control section as the column vector.

[0133] The basic deviation vector is generated based on the deviation of the real-time deflection value of all arch rib key control sections from the preset installation accuracy threshold.

[0134] The basic deviation vector is corrected based on the preset wind-induced correction coefficient and temperature correction coefficient to obtain the corrected deviation vector.

[0135] The multi-dimensional cable force correction vector is solved by the regularized least squares method based on the influence matrix and the corrected deviation vector.

[0136] The multi-dimensional cable force correction vector is filtered by safety constraint, balanced constraint and construction feasibility constraint to obtain the cable force correction amount of all hangers.

[0137] In this embodiment, the influence matrix is constructed by taking the hanger number of the skew-span steel box arch bridge (such as 12 pairs of hangers of the whole bridge, numbered G1-G12) as the row vector and the arch rib key control section (2 arch feet, 2 1 / 4 spans, 1 arch top, 1 3 / 4 span, a total of 6 sections) as the column vector: in the analysis of the skew-span steel box arch bridge, the number of each hanger is sequentially formed as a row, and the key control section of the arch rib for controlling the structure state is taken as a column to form a matrix. The elements of this matrix reflect the influence relationship of each hanger on each arch rib key control section. For example, the influence degree of hanger 1 on arch rib key control section 1 is 0.5, which is recorded as 0.5 at the corresponding position of the matrix, which is used for subsequent analysis of the effect of hanger cable force change on different sections of the arch rib.

[0138] The matrix element M(i,j) represents the vertical deflection change value (mm) of the jth control section when the ith hanger is tensioned by a unit force (100kN), which is determined by the following method:

[0139] ANSYS is used to establish a refined finite element model to simulate the structural response under different load cases (including wind load and hoisting eccentric load coupling effect), and the initial calculation of M(i, j) baseline value;

[0140] For the diagonal span side (asymmetrically stressed side) of the suspender (such as G1-G4), the additional effect of arch rib transverse offset is considered, and the M(i, j) value is multiplied by a correction coefficient of 1.2 (because the diagonal span structure is prone to asymmetric deflection).

[0141] In this embodiment, a basic deviation vector is generated based on the deviation of the real-time deflection values of all arch rib key control sections from the preset installation precision threshold: the real-time deflection values of each control section of the arch rib are compared with the preset installation precision threshold, the deviation values of each control section are calculated, and these deviation values are sequentially composed into a vector. For example, the control section 1 deviation is 2mm, the control section 2 deviation is-3mm, etc., to form the basic deviation vector [2,-3,…], which reflects the deviation of the current control section from the ideal installation state.

[0142] In this embodiment, the preset wind-induced correction coefficient is: when v=10m / s, the preset wind-induced correction coefficient=1.25;

[0143] A three-dimensional coupling factor of "wind speed-wind direction-section position" is introduced:

[0144] When the angle between the wind direction and the arch rib axis θ≤30°, the preset wind-induced correction coefficient is 1+0.03v (v is the wind speed, unit m / s) ;

[0145] When 30°<θ≤60°, the preset wind-induced correction coefficient is 1+0.02v×sinθ (considering the decomposition effect of lateral wind) ;

[0146] When θ>60°, the preset wind-induced correction coefficient is 1+0.015v×cosθ (the wind load is mainly in the axial direction, and the influence on deflection is weakened) ;

[0147] For the windward and leeward sections of the hexagonal arch rib (such as the left 1 / 4 span and the right 1 / 4 span), the preset wind-induced correction coefficients are multiplied by the asymmetric coefficients of 1.2 and 0.8 respectively (considering the uneven distribution of wind load).

[0148] In this embodiment, the temperature correction coefficient: because the temperature change will cause the thermal expansion and contraction of the bridge structure material, affecting the deformation of the structure, the temperature correction coefficient is set to correct the influence of temperature change on the deformation of the structure. For example, when the temperature changes by 1℃, the correction ratio of the structure deformation is set to 0.005, which is the temperature correction coefficient, and its value is determined by the structure material and the environmental temperature change characteristics.

[0149] Arranging 16 temperature sensors (1 every 5 m along the longitudinal direction of the arch rib, and 1 on each of the 6 faces in the circumferential direction), and calculating:

[0150] Longitudinal temperature gradient ΔTlongitudinal (temperature difference between arch foot and arch top): when ΔTlongitudinal> 12℃, ktemp longitudinal = 1+0.02×ΔTlongitudinal (affecting overall deflection);

[0151] Circumferential temperature gradient ΔTcircumferential (maximum temperature difference in the same section): when ΔTcircumferential> 8℃, ktemp circumferential = 1+0.01×ΔTcircumferential (affecting local bending, only correcting lateral deviation).

[0152] In this embodiment, the basic deviation vector is modified based on the preset wind-induced correction coefficient and the temperature correction coefficient to obtain a modified deviation vector: each element in the basic deviation vector is multiplied by the preset wind-induced correction coefficient and the temperature correction coefficient, respectively, to comprehensively consider the influence of wind load and temperature on the deviation, and to obtain the modified deviation vector. For example, the basic deviation vector element is 5, the preset wind-induced correction coefficient is 0.8, and the temperature correction coefficient is 0.9, then the modified element becomes 5×0.8×0.9=3.6, and all elements after modification form the modified deviation vector. The modified deviation vector Δδmod = basic deviation vector × preset wind-induced correction coefficient × ktemp longitudinal × ktemp circumferential, realizing fine modification of the spatial and temporal distribution of the temperature field.

[0153] In this embodiment, the multi-dimensional cable force correction vector is solved based on the influence matrix and the modified deviation vector by using the regularization least square method: a multi-dimensional vector is calculated by using the constructed influence matrix and the modified deviation vector by using the mathematical method of Tikhonov regularization least square method. Each dimension in this multi-dimensional vector corresponds to a correction value of different suspender cable forces, which can minimize the structural deformation deviation considering various influencing factors, so as to determine how to adjust each suspender cable force.

[0154] In this embodiment, the multi-dimensional cable force correction vector: the vector obtained by the above calculation, each element of which represents the correction amount of different suspender cable forces, and from multiple dimensions reflects the numerical value of each suspender cable force that should be changed to make the structure return to the ideal state. For example, the multi-dimensional cable force correction vector [10, -5, 8, …] indicates that the first suspender cable force needs to be increased by 10 kN, the second suspender cable force needs to be reduced by 5 kN, etc.

[0155] In this embodiment, the multi-dimensional cable force correction vector is subjected to safety constraint filtering, balanced constraint filtering and construction feasibility constraint filtering to obtain the cable force correction of all the hangers: for each element in the multi-dimensional cable force correction vector, the safety constraint condition (such as single hanger correction ≤ ±180kN (not more than 10% of the design cable force), additional limit of diagonal span side hanger ≤ ±150kN (control arch foot stress ≤250MPa)), balanced constraint condition (make the distribution of each hanger cable force more uniform and reasonable, such as the difference between adjacent hanger correction ≤25kN, avoid load mutation), construction feasibility constraint condition (the correction needs to adapt to the jack tensioning accuracy (0.4 level), and is rounded to the multiple of 5kN (such as the calculated value 132kN is adjusted to 130kN)) are sequentially screened and adjusted, and finally the cable force correction of all the hangers meeting various actual requirements is obtained. For example, the cable force correction of a certain hanger is found to be out of the safety range after safety constraint filtering and needs to be adjusted to be within the safety range; after balanced constraint filtering, it may be further fine-tuned to ensure that the stress of each hanger is more balanced; finally, after construction feasibility constraint filtering, the correction that cannot be realized by the existing construction process is excluded, and the final cable force correction is obtained.

[0156] In order to determine the balanced constraint strength requirement, the multi-dimensional cable force correction vector is subjected to balanced constraint filtering, which comprises:

[0157] A finite element model in the construction process of the diagonal span steel box arch bridge is generated, a force flow monitoring section is selected every preset distance along the longitudinal direction of the arch rib in the finite element model, a plurality of feature points are selected at each force flow monitoring section, the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point are extracted, and the force flow direction angle of each feature point is calculated based on the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point.

[0158] The continuous section with the same structural feature or stress feature in the finite element model is regarded as a to-be-analyzed section;

[0159] The force flow sharp turn section is screened out in all the to-be-analyzed sections based on the force flow direction angle of all the feature points in the finite element model;

[0160] The strength requirement of the balanced constraint is determined based on the distribution position and sharp turn degree of all the force flow sharp turn sections;

[0161] The multi-dimensional cable force correction vector is subjected to balanced constraint filtering based on the strength requirement of the balanced constraint.

[0162] In this embodiment, the finite element model of the construction process of the skew-span steel box arch bridge is generated: with the help of professional finite element analysis software, according to the design drawings, material properties, construction technology and other information of the skew-span steel box arch bridge, a digital model capable of simulating the mechanical behavior of the bridge at each stage of construction is constructed. This model discretizes the bridge structure into a large number of finite elements, and analyzes the stress and deformation of the whole bridge by calculating the interaction between the elements. For example, the arch rib, suspender, deck and other structural components are simulated by appropriate element types, and the corresponding material parameters and boundary conditions are assigned, so as to obtain a finite element model that can be used for analysis.

[0163] In this embodiment, the preset distance is, for example, 0.5 meters.

[0164] In this embodiment, the force flow monitoring section: in the finite element model of the skew-span steel box arch bridge, a specific cross section is selected, and these sections are used to monitor the transmission and distribution of force in the structure.

[0165] In this embodiment, multiple feature points are selected at each force flow monitoring section: on the force flow monitoring section that has been determined, multiple representative points are selected, and the mechanical parameters of these points can reflect the stress characteristics of the section. For example, on a circular force flow monitoring section, edge points, center points and points distributed at a certain angle are selected as feature points in order to comprehensively obtain the force information on the section.

[0166] In this embodiment, the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point are extracted: for each selected feature point, the axial (along the length direction of the component), radial (perpendicular to the length direction of the component) and shear stress directions are obtained through the calculation results of the finite element model. The axial normal stress direction of the feature point is along the left direction of the bridge longitudinal direction, the radial normal stress direction is perpendicular to the section and points to the inside, and the shear stress direction is at a certain angle with a coordinate axis.

[0167] In this embodiment, the force flow direction angle of each feature point is calculated based on the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point: using the relevant principles and calculation formulas of material mechanics, and combining the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point, an angle value is calculated, which represents the flow direction of force at the feature point, and is called the force flow direction angle. The force flow direction angle can intuitively show the transmission trend of force at the feature point. For example, the inverse tangent value of the ratio of twice the shear stress to the difference between the axial normal stress and the radial normal stress, and then dividing the result by 2, can obtain the force flow direction angle of the feature point.

[0168] In this embodiment, the continuous section with the same structural characteristics or stress characteristics in the finite element model is taken as the section to be analyzed: in the finite element model of the skew-span steel box arch bridge, the structure is divided, and those parts which are consistent and continuous in structural form (such as straight segment arch rib, same cross section shape of suspender, etc.) or stress characteristics (such as being subjected to axial tension, similar bending moment distribution, etc.) are analyzed as a whole, which is the section to be analyzed. For example, the arch rib part with constant cross section shape and mainly subjected to axial compression in a length is divided into a section to be analyzed.

[0169] In this embodiment, the force flow sharp turning section: in the section to be analyzed, if the force flow direction angle changes sharply in a short distance, it is called the force flow sharp turning section. The force transmission path of such area changes suddenly, which may cause local stress concentration of the structure and potential threat to the safety of the structure. For example, at the connection between the arch rib and the pier, the force flow direction may change sharply from the direction along the arch rib to the direction of the pier, and this area is the force flow sharp turning section.

[0170] In this embodiment, the strength requirement of the balanced constraint is determined based on the distribution position and sharp turning degree of all force flow sharp turning sections: the position of all force flow sharp turning sections in the finite element model and the sharp degree of the change of the force flow direction angle are comprehensively considered to determine the strength required when the structure is balanced constrained. For example:

[0171] If the force flow sharp turning sections are concentrated in the arch foot steel-concrete joint section and the direction angle change rate reaches 10° / m, the adjacent suspender force difference constraint strength needs to be increased to ≤3% to suppress the mutation;

[0172] When the 1 / 4 span area appears mild sharp turning (change rate 6° / m), the constraint strength can be appropriately relaxed to force difference ≤5%;

[0173] If there is no sharp turning section in the arch top circular curve section, only the conventional constraint strength (force difference ≤8%) is required.

[0174] If the force flow sharp turning sections are concentrated in the arch foot steel-concrete joint section and the direction angle change rate reaches 10° / m, the adjacent suspender force difference constraint strength needs to be increased to ≤3% to suppress the mutation;

[0175] When the 1 / 4 span area appears mild sharp turning (change rate 6° / m), the constraint strength can be appropriately relaxed to force difference ≤5%;

[0176] If there is no sharp turning section in the arch top circular curve section, only the conventional constraint strength (force difference ≤8%) is required.

[0177] In this embodiment, the multi-dimensional cable force correction vector is filtered based on the balanced constraint strength requirement: according to the previously determined balanced constraint strength requirement, each element in the multi-dimensional cable force correction vector is screened and adjusted. The corrected cable force distribution ensures that the structure is more uniform in the force flow rapid turning section and its surroundings, avoiding unreasonable cable force leading to further deterioration of the force flow distribution. For example, if the cable force correction of a certain suspender will exacerbate the stress concentration of the force flow rapid turning section, and the balanced constraint strength requirement is high, the cable force correction needs to be adjusted to meet the requirement of uniform structural stress.

[0178] In order to screen the force flow rapid turning section in the to-be-analyzed section, the force flow direction angle of all feature points in the finite element model is used to screen the force flow rapid turning section in all to-be-analyzed sections, including:

[0179] The ratio of the difference value of the force flow direction angle of the feature points at the same position in each group of adjacent force flow monitoring sections to the distance between adjacent sections is taken as the corresponding direction angle change rate;

[0180] The standard deviation of the force flow direction angle of all feature points in the to-be-analyzed section is defined as the force flow continuity index;

[0181] The ratio of the number of all direction angle abrupt change points to the number of all feature points in the to-be-analyzed section is taken as the extreme point ratio;

[0182] Based on the direction angle change rate, the force flow continuity index, and the extreme point ratio in each to-be-analyzed section, the force flow rapid turning section is screened in all to-be-analyzed sections.

[0183] In this embodiment, the direction angle abrupt change point: in the analysis of the finite element model of the diagonal-span steel box arch bridge, the direction angle abrupt change point refers to the point at which the force flow direction angle changes suddenly and significantly. On a continuous structure section, the force flow direction angle usually changes relatively smoothly, but at some points, the direction angle will change sharply due to changes in structure form, sudden changes in stress conditions, etc. For example, at the connection position of the arch rib and the special structural member, the force flow direction angle may change from a certain direction to another direction with a large difference in an instant. This position is the direction angle abrupt change point. It has an important indicating effect on judging the abnormality of internal force transmission and the local stress concentration area of the structure.

[0184] In this embodiment, based on the direction angle change rate, the force flow continuity index, and the extreme point ratio in each to-be-analyzed section, the force flow rapid turning section is screened in all to-be-analyzed sections:

[0185] A higher change rate means that the force flow direction changes greatly in a short distance. The lower the force flow continuity index, the more unsmooth the force flow transmission in the section, and there may be a force mutation or concentration. A higher extreme point ratio means that the force flow direction angle changes frequently and sharply in the section.

[0186] Screening force flow sharp turn section: considering the above three indexes comprehensively, those sections with high direction angle change rate, low force flow continuity index and high extreme point proportion are screened out from all the sections to be analyzed, and these sections are force flow sharp turn sections. For example, if the direction angle change rate of a section to be analyzed is higher than the set threshold, the force flow continuity index is lower than the specific standard, and the extreme point proportion exceeds a certain proportion, it is determined as a force flow sharp turn section.

[0187] The application provides an embodiment of a machine learning-based installation precision dynamic regulation and control system for a skew-span steel box arch bridge, comprising:

[0188] A data acquisition module is configured to acquire real-time data during construction of the skew-span steel box arch bridge, wherein the real-time data includes arch rib real-time stress, arch rib real-time deformation, real-time environmental wind speed and real-time hoisting parameters.

[0189] A model construction module is configured to construct a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load based on a large amount of historical construction data.

[0190] A coupling analysis module is configured to input the real-time data into the prediction model to obtain real-time coupling effect analysis results.

[0191] A sequential optimization module is configured to output hoisting sequence optimization instructions based on a decision engine, coupling effect analysis results and preset mechanical constraint conditions of the skew-span steel box arch bridge.

[0192] A scheme generation module is configured to generate a cable force grading adjustment scheme based on the deviation of the arch rib real-time deformation from a preset installation precision threshold.

[0193] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A method for dynamic regulation and control of installation precision of a skew-span steel box arch bridge based on machine learning, characterized in that, The method comprises the following steps: Collecting real-time data in the construction process of the skew-span steel box arch bridge, wherein the real-time data comprises real-time stress of the arch rib, real-time deformation of the arch rib, real-time environmental wind speed and real-time hoisting parameters; Based on a large amount of historical construction data, a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed; The real-time data is input into the prediction model to obtain real-time coupling effect analysis results; Based on the decision engine, the coupling effect analysis results and the preset mechanical constraint conditions of the skew-span steel box arch bridge, hoisting sequence optimization instructions are output; Based on the deviation of the real-time deformation of the arch rib from the preset installation accuracy threshold, a cable force grading adjustment scheme is generated.

2. The machine learning-based installation precision dynamic regulation method for a cable-stayed steel box arch bridge according to claim 1, characterized in that, Collecting real-time data in the construction process of the skew-span steel box arch bridge, comprising: Based on the fiber Bragg grating sensors uniformly arranged along the longitudinal direction of the arch rib of the skew-span steel box arch bridge, the real-time axial stress and the real-time radial stress of the arch rib in the construction process of the skew-span steel box arch bridge are collected as the real-time stress of the arch rib; Based on the laser tracker arranged on the two sides of the arch rib, the three-dimensional coordinates of the arch rib in the construction process of the skew-span steel box arch bridge are collected in real time, and the transverse bending deformation and the axial deviation are calculated based on the real-time collected three-dimensional coordinates as the real-time deformation of the arch rib; Based on the wind speed sensor arranged at the top of the arch rib, the instantaneous wind speed and the wind direction in the construction process of the skew-span steel box arch bridge are collected as the real-time environmental wind speed; The real-time segment physical parameters, the real-time hoisting posture parameters and the real-time motion state parameters in the construction process of the skew-span steel box arch bridge are collected as the real-time hoisting parameters.

3. The machine learning-based installation precision dynamic regulation method for a cable-stayed steel box arch bridge according to claim 1, characterized in that, Based on a large amount of historical construction data, a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load is constructed, comprising: The combination of wind load and hoisting eccentric load in a large amount of historical construction data is simulated to obtain preset group sample data; All group sample data is divided into a training set and a validation set at a preset ratio, and a neural network model is trained based on the training set and the validation set, while the mean square error of the predicted deformation and the actual deformation is minimized through the Adam optimizer until the mean square error does not exceed the preset error threshold, thereby obtaining the prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load.

4. The machine learning-based installation precision dynamic regulation and control method for a cable-stayed steel box arch bridge according to claim 1, characterized in that, The construction process of the decision engine further comprises: Based on the actual deformation and the allowable deformation and the hoisting duration delay rate, a reward function of the reinforcement learning algorithm is generated; Based on the reward function of the reinforcement learning algorithm, the decision engine is constructed.

5. The machine learning-based installation precision dynamic regulation and control method for a cable-stayed steel box arch bridge according to claim 1, characterized in that, Based on the decision engine, the coupling effect analysis results and the preset mechanical constraint conditions of the skew-span steel box arch bridge, hoisting sequence optimization instructions are output, comprising: Based on each selectable hoisting sequence, a hoisting sequence thread is generated; Based on the reward function of the reinforcement learning algorithm in the decision engine, the reward value of each hoisting sequence thread is determined, and the hoisting sequence optimization instructions are generated based on the hoisting sequence thread corresponding to the maximum reward value.

6. The machine learning-based installation precision dynamic regulation and control method for a cable-stayed steel box arch bridge according to claim 1, characterized in that, Based on the deviation of the real-time deformation of the arch rib from the preset installation accuracy threshold, a cable force grading adjustment scheme is generated, comprising: Based on the real-time deformation of the arch rib, the real-time deflection value of the arch rib is determined; The preset maximum preset deflection value is taken as the preset installation accuracy threshold; When the deviation of the real-time deflection value of the arch rib from the preset installation accuracy threshold does not exceed the first deviation threshold, the first proportion initial tensioning is performed as the cable force grading adjustment scheme; When the deviation of the real-time deflection value of the arch rib from the preset installation accuracy threshold exceeds the first deviation threshold and does not exceed the second deviation threshold, a first proportion to a second proportion staged tension is performed as a cable force staged adjustment scheme; When the deviation of the real-time deflection value of the arch rib from the preset installation accuracy threshold exceeds the second deviation threshold, a maximum proportion tension and a re-tuning of the cable force correction amount of all hangers calculated based on the matrix influence method are performed as the cable force staged adjustment scheme.

7. The machine learning-based installation precision dynamic regulation and control method for a cable-stayed steel box arch bridge according to claim 6, characterized in that, The cable force correction amount of all hangers is calculated based on the matrix influence method, including: Taking the hanger number of the skew-span steel box arch bridge as a row vector and the arch rib key control section as a column vector, an influence matrix is constructed; A basic deviation vector is generated based on the deviation of the real-time deflection value of all arch rib key control sections from the preset installation accuracy threshold; The basic deviation vector is corrected based on the preset wind-induced correction coefficient and the temperature correction coefficient to obtain a corrected deviation vector; A multi-dimensional cable force correction vector is solved by a regularized least squares method based on the influence matrix and the corrected deviation vector; The multi-dimensional cable force correction vector is filtered by safety constraints, balanced constraints and construction feasibility constraints to obtain the cable force correction amount of all hangers.

8. The machine learning-based installation precision dynamic regulation method for a cable-stayed steel box arch bridge according to claim 7, characterized in that, Balanced constraint filtering of the multi-dimensional cable force correction vector includes: A finite element model in the construction process of the skew-span steel box arch bridge is generated, and a force flow monitoring section is selected every preset distance along the longitudinal direction of the arch rib in the finite element model, and a plurality of feature points are selected at each force flow monitoring section, and the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point are extracted, and the force flow direction angle of each feature point is calculated based on the axial normal stress direction, the radial normal stress direction and the shear stress direction of each feature point; The continuous section with the same structural characteristics or stress characteristics in the finite element model is regarded as the section to be analyzed; Based on the force flow direction angle of all feature points in the finite element model, the force flow sharp turning sections are screened out in all sections to be analyzed; The strength requirement of the balanced constraint is determined based on the distribution position and the sharp turning degree of all force flow sharp turning sections; The multi-dimensional cable force correction vector is filtered based on the strength requirement of the balanced constraint.

9. The machine learning-based installation precision dynamic regulation method for a cable-stayed steel box arch bridge according to claim 8, characterized in that, Based on the force flow direction angle of all feature points in the finite element model, the force flow sharp turning sections are screened out in all sections to be analyzed, including: The ratio of the difference value of the force flow direction angle of the feature points at the same position in each group of adjacent force flow monitoring sections to the adjacent section spacing is regarded as the corresponding direction angle change rate; The standard deviation of the force flow direction angle of all feature points in the section to be analyzed is defined as the force flow continuity index; The ratio of the number of all direction angle abrupt change points to the number of all feature points in the section to be analyzed is regarded as the extreme point ratio; Based on all the direction angle change rates and the force flow continuity index and the extreme point ratio in each section to be analyzed, the force flow sharp turning sections are screened out in all sections to be analyzed.

10. A machine learning-based installation precision dynamic regulation system for a skew-span steel box arch bridge, characterized in that, Including: The data acquisition module is used to collect real-time data in the construction process of the skew-span steel box arch bridge, wherein the real-time data includes arch rib real-time stress, arch rib real-time deformation, real-time environmental wind speed and real-time hoisting parameters; The model construction module is used to construct a prediction model capable of analyzing the coupling effect of wind load and hoisting eccentric load based on a large amount of historical construction data; a coupling analysis module, configured to input real-time data into the prediction model to obtain a real-time coupling effect analysis result; a sequence optimization module, configured to output a hoisting sequence optimization instruction based on the decision engine, the coupling effect analysis result, and preset mechanical constraint conditions of the cable-stayed steel box arch bridge; a scheme generation module, configured to generate a cable force staged adjustment scheme based on a deviation of the real-time deformation of the arch rib from a preset installation precision threshold.

Images