CN-122020018-A - Construction state data driving oriented intelligent positioning method and device for transverse struts of cable-stayed bridge

CN122020018ACN 122020018 ACN122020018 ACN 122020018ACN-122020018-A

Abstract

The application provides a construction state data-driven intelligent positioning method and device for a cable-stayed bridge cross brace, which comprise the steps of realizing phase locking acquisition of camera exposure and double-angle oblique illumination based on a linear speed encoder and a cutter head phase, executing speed compensation and geometric shearing on continuous frames to obtain a space-time registration sequence, extracting sub-pixel edge curves and generating a narrow-band interested region along a normal direction, calculating a flicker ratio and a relative phase difference between alternate illumination frames, generating a burr salient map by combining a normal gradient and a fine tooth periodic energy along the edge direction, inputting the salient map and the normal brightness profile into a convolutional neural network, outputting pixel-level burr probability, amplitude proxy quantity and periodic scoring, and carrying out self-adaptive threshold judgment by combining linear speed, tension and foil thickness, and giving an alarm in association with the cutter head phase. The detection method and the detection system realize high detection rate, low false alarm and millisecond delay under the high light reflection and the linear speed of more than or equal to 200m/min, and are suitable for multi-path on-line quality control and process linkage.

Inventors

SUN SHUDONG
YANG LIHUA
YANG WEIGANG
ZHANG YUANDONG
XIE ZHIJIE

Assignees

惠州交投公路建设有限公司

Dates

Publication Date: 20260512
Application Date: 20251222

Claims (10)

1. The intelligent positioning method for the cable-stayed bridge cross braces facing construction state data driving is characterized by comprising the following steps: s1, acquiring multisource construction state data of a cable-stayed bridge in a construction stage, simultaneously acquiring a field video image of a transverse strut area acquired by a camera, and generating a working condition coding vector based on construction procedure information; s2, preprocessing multisource construction state data to obtain construction state feature vectors, carrying out camera calibration, target area detection, feature point tracking and three-dimensional posture inversion on a field video image to obtain three-dimensional displacement and posture parameters of the end part and the cross section of the transverse support, forming the video image feature vectors, and carrying out time alignment with the construction state feature vectors to obtain a construction state feature vector fused with multiple modes; S3, inputting the fused multi-mode construction state feature vector into a digital twin model of a cable-stayed bridge at a construction stage, and calibrating model parameters to obtain a digital twin model matched with the current construction state; s4, generating a plurality of cross brace positioning candidate schemes based on the cross brace design position and construction allowable deviation in the geometric constraint range given by the digital twin model, and constructing a cross brace positioning characteristic parameter for each cross brace positioning candidate scheme; S5, inputting the fusion multi-mode construction state feature vector, the working condition coding vector and the transverse strut positioning feature parameter into a convolutional neural network prediction model to obtain a target transverse strut positioning final scheme and a corresponding structural response index; s6, generating a cross brace installation control instruction according to a final target cross brace positioning scheme, continuously acquiring video images in the cross brace installation process to estimate the end gesture of the cross brace, comparing the actual measurement gesture with a structural response index, and outputting an adjustment prompt when the deviation exceeds a preset threshold.
2. The intelligent positioning method for a cable-stayed bridge cross brace driven by construction state data according to claim 1, wherein the multi-source construction state data comprises bridge tower displacement, main girder displacement, stay cable force, strain, temporary support counter force and environmental parameters.
3. The intelligent positioning method for the cable-stayed bridge cross brace driven by construction state data according to claim 1, wherein the generating of the working condition coding vector based on the construction procedure information comprises the steps of collecting current construction procedure information, wherein the construction procedure information at least comprises one or more of a hoisting sequence number of a current girder segment, a tensioning completion state of a corresponding stay cable, a tower beam connection state and an enabling state of a temporary support, mapping the construction procedure information into a working condition category identification representing a construction stage type according to a preset construction stage division rule, and representing the working condition category identification as the working condition coding vector through multidimensional binary coding.
4. The intelligent positioning method for the cable-stayed bridge cross brace driven by construction state data according to claim 1, wherein the preprocessing of the multi-source construction state data to obtain construction state feature vectors comprises the steps of obtaining original monitoring data collected by a displacement sensor, a strain gauge, a cable force meter, a support counter-force meter and a thermometer which are arranged on a bridge tower, a main beam and a stay cable, restraining high-frequency noise by adopting low-pass filtering on the original monitoring data, carrying out coordinate conversion and dimension normalization processing on displacement, strain, cable force and support counter-force data of each monitoring point, and splicing the preprocessed displacement, strain, cable force, support counter-force and temperature feature vectors according to a preset section and component arrangement sequence to form the construction state feature vectors.
5. The intelligent positioning method for the transverse struts of the cable-stayed bridge driven by construction state data according to claim 1 is characterized in that the intelligent positioning method for the transverse struts of the cable-stayed bridge comprises the steps of calibrating internal parameters and external parameters of a camera arranged in a transverse strut area, establishing a space mapping relation between a camera pixel coordinate system and a full-bridge coordinate system of the cable-stayed bridge, then performing target area detection on an on-site video image, extracting image areas of transverse strut members, transverse strut end markers and preset transverse strut section structural features, selecting a plurality of feature points in the image areas, tracking pixel coordinates of the feature points in adjacent video frames based on an optical flow tracking algorithm, utilizing initial three-dimensional coordinates of camera calibration parameters and the feature points in a reference frame, combining a perspective-n-point solving algorithm, inverting to obtain three-dimensional coordinates of transverse strut end and preset section feature points in the full-bridge coordinate system, solving three-dimensional coordinate differences of the three-dimensional coordinates relative to the corresponding feature points in a reference construction state, obtaining three-dimensional displacement of the transverse strut end and section, calculating attitude angle parameters of the transverse strut end and the section according to a preset attitude angle vector of the fitting transverse strut axis and the local section normal vector, and splicing the three-dimensional angle parameters to obtain the video image displacement parameters.
6. The intelligent positioning method for the cable-stayed bridge cross brace driven by construction state data according to claim 1 is characterized in that the step S3 comprises the steps of extracting three-dimensional displacement of the top end and the middle section of a bridge tower, deflection and plane displacement of a main beam control section, actual measurement response of a stay cable force and a temporary support counter force from a fused multi-mode construction state characteristic vector, carrying out one-to-one correlation on the actual measurement response and corresponding calculation nodes and components in a digital twin model, selecting at least one of temperature load distribution, construction additional load, support rigidity and material elastic modulus as parameters to be calibrated, carrying out structural response forward analysis calculation in the digital twin model by taking the parameters to be calibrated as variables to obtain model calculation response, constructing a parameter inverse analysis model by taking a difference value between the model calculation response and the corresponding actual measurement response as an objective function, adopting an iterative least square Kalman filtering algorithm, carrying out repeated updating iteration on the parameters to be calibrated, and obtaining the digital twin model matched with the current construction state when the displacement deviation and the cable force deviation of a key monitoring position are smaller than a preset threshold value.
7. The intelligent positioning method for the cable-stayed bridge cross brace driven by construction state data according to claim 1, wherein the step S4 comprises the steps of obtaining design coordinates and design attitude parameters of two ends of the cross brace under a design full-bridge coordinate system, wherein the design attitude parameters comprise a cross brace axis direction and design angles around a bridge longitudinal direction, a cross brace transverse direction and a vertical axis, determining a geometric constraint range of two end nodes of the cross brace in the vertical direction, the forward direction and the cross bridge direction and a minimum clearance requirement between the two end nodes and adjacent components based on a calibrated digital twin model of the cable-stayed bridge construction stage, and an allowable additional displacement and an internal force range of the cross brace after the installation of the cross brace, carrying out combined disturbance on the vertical elevation, the forward direction position, the cross bridge direction deviation and the attitude angles around a cross brace design position in the geometric constraint range according to preset step length to generate a plurality of cross brace positioning candidate schemes meeting geometric constraint, calculating three-dimensional coordinates, cross brace axis vectors and corresponding attitude angles of the two end nodes under the full-bridge coordinate system, and additionally matching with the candidate schemes, adjusting the cross brace positioning parameter positioning candidate parameters according to preset diagonal cable positioning parameter positioning candidate parameters.
8. The construction state data driven oriented intelligent positioning method for a cable-stayed bridge cross brace according to claim 1, wherein the convolutional neural network prediction model comprises: the construction state characteristic convolution sub-network is used for carrying out two-dimensional convolution coding on the construction state characteristic vector and the working condition coding vector which are fused in a multi-mode manner to obtain a first characteristic diagram of the current construction state and the construction stage; the cross brace positioning characteristic convolution sub-network is used for carrying out convolution coding on cross brace positioning characteristic parameters of each cross brace positioning candidate scheme to obtain a second characteristic diagram representing the spatial position and the gesture characteristics of different cross brace positioning schemes; The digital twin response embedding module is used for carrying out structural analysis and calculation on each transverse strut positioning candidate scheme to obtain a corresponding basic structure response result, and splicing the basic structure response result with the first characteristic diagram and the second characteristic diagram in the channel dimension to form a third characteristic diagram fused with physical information; The residual correction and grading output module is used for carrying out multi-scale convolution and residual connection operation on the third feature map, learning the deviation relation between the basic structure response result and the target structure response index, and outputting the structure response index of each cross brace positioning candidate scheme and the comprehensive grading value for indicating the target cross brace positioning final scheme, wherein the comprehensive grading value is used for automatically selecting the target cross brace positioning final scheme from a plurality of cross brace positioning candidate schemes.
9. The intelligent positioning method for the cable-stayed bridge cross-brace driven by construction state data according to claim 1, wherein the comparing the measured gesture with the structural response index in the process of installing the cross-brace comprises: in the process of installing the transverse brace, real-time acquisition of actual measurement attitude parameters of the end part of the transverse brace under a full-bridge coordinate system based on processing of field video images, wherein the actual measurement attitude parameters at least comprise three-dimensional coordinates of characteristic points of the end part of the transverse brace and actual measurement corners around longitudinal, transverse and vertical axes of a bridge; Extracting target attitude parameters corresponding to a final target transverse strut positioning scheme from the structural response index, wherein the target attitude parameters at least comprise target three-dimensional coordinates and target corners; Respectively calculating a difference vector of the actually measured three-dimensional coordinate and the target three-dimensional coordinate and a difference value of the actually measured corner and the target corner under a unified full-bridge coordinate system to obtain a displacement deviation amount and an angle deviation amount; And comparing the amplitude values of the displacement deviation amount and the angle deviation amount according to a preset displacement deviation threshold value and an angle deviation threshold value, judging that the gesture deviation exceeds the limit when any deviation amount exceeds a corresponding threshold value, and triggering and outputting an adjustment prompt or an alarm signal, wherein the adjustment prompt comprises one or more of reducing the hoisting speed, finely adjusting the end position of the cross brace and/or suspending the hoisting operation.
10. Construction state data driving-oriented intelligent positioning device for a cable-stayed bridge cross brace, which is characterized by comprising: the data acquisition module is used for acquiring multi-source construction state data of the cable-stayed bridge at the construction stage, acquiring a field video image of a transverse strut area acquired by a camera, and generating a working condition coding vector based on construction procedure information; The characteristic vector generation module is used for preprocessing the multisource construction state data to obtain construction state characteristic vectors, carrying out camera calibration, target area detection, characteristic point tracking and three-dimensional posture inversion on the field video image to obtain three-dimensional displacement and posture parameters of the end part and the cross section of the transverse support, forming the video image characteristic vectors, and carrying out time alignment with the construction state characteristic vectors to obtain the fused multi-mode construction state characteristic vectors; The digital twin model updating module is used for inputting the fused multi-mode construction state feature vector into a digital twin model of a cable-stayed bridge at a construction stage, calibrating model parameters and obtaining a digital twin model matched with the current construction state; The candidate scheme generating module is used for generating a plurality of cross brace positioning candidate schemes based on the cross brace design position and construction allowable deviation in the geometric constraint range given by the digital twin model, and constructing a cross brace positioning characteristic parameter for each cross brace positioning candidate scheme; The target cross brace positioning final scheme generating module inputs the fusion multi-mode construction state feature vector, the working condition coding vector and the cross brace positioning feature parameter into a convolutional neural network prediction model to obtain a target cross brace positioning final scheme and a corresponding structural response index; And the installation control module is used for generating a cross brace installation control instruction according to a final target cross brace positioning scheme, continuously acquiring video images in the cross brace installation process to estimate the end gesture of the cross brace, comparing the actual measurement gesture with a structural response index, and outputting an adjustment prompt when the deviation exceeds a preset threshold value.

Description

Construction state data driving oriented intelligent positioning method and device for transverse struts of cable-stayed bridge Technical Field The invention relates to the technical field of intelligent positioning, in particular to an intelligent positioning method and device for a cable-stayed bridge cross brace driven by construction state data. Background Cable-stayed bridges are one of the main forms of large-span bridges, and temporary or permanent transverse bracing members are usually required to be arranged at the construction stage so as to improve the overall rigidity and the spatial stability of the structure in the construction period and limit the relative deformation of bridge towers, main beams and stay cables. The position, posture and cooperative relation with peripheral components of the cross brace directly influence the linear control, internal force distribution and the safety and durability of the final bridge formation state in the construction stage. In the prior engineering practice, the arrangement and the positioning of the transverse struts of the cable-stayed bridge are mainly arranged according to theoretical lines and structures given by design, lofting is carried out by using traditional measuring means such as total stations, level gauges and the like, and then installation and adjustment are carried out by combining experience of constructors. Along with the evolution of the construction working condition, the bridge tower and the main beam generate remarkable time-varying deformation under the multi-factor actions of temperature, wind load, cable force adjustment, temporary bracket deformation and the like, the actual space position often deviates from the design theoretical value, but the transverse strut positioning is still based on a static design value and a small amount of measuring point correction, and the real stress and the geometric state of the structure are difficult to reflect in time. In order to improve construction control precision, monitoring equipment such as a displacement meter, a strain gauge, a cable force meter, a GNSS (Global navigation satellite System) and the like are arranged in part of engineering, so that a structural health monitoring system in a construction period is formed and is used for collecting data such as bridge tower top displacement, main beam control section deflection, cable force change, support counter force, environmental temperature and the like. However, such monitoring systems usually mainly use point sensors, the number of measuring points is limited, the arrangement positions are fixed, the cross brace installation area and peripheral components are difficult to be covered on the whole, the problem of 'local density and overall sparseness' exists in the monitoring information, and fine information of local alignment and posture of the cross brace cannot be provided. On the other hand, a large number of video monitoring cameras are commonly installed on construction sites for safety monitoring and construction process recording. In the prior art, video images stay at the level of manual visual inspection or simple video playback, even if a small amount of attempts are made to utilize image processing to carry out displacement or deformation identification, single-point off-line post-processing analysis is often carried out, a unified coordinate system and a unified time scale are not formed with multi-source sensor data and a structural analysis model, and a systematic solution for accurate positioning and real-time checking of a transverse strut is not constructed. In recent years, digital twin technology is gradually introduced into the field of bridges, and the real stress and geometric state of a construction stage structure can be reflected to a certain extent by establishing a high-precision finite element model corresponding to actual engineering and carrying out inversion calibration on model parameters in combination with monitoring data. However, the existing digital twin application focuses on the analysis and evaluation of integral linearity and internal force, and the application of the installation position optimization and real-time construction control of local components facing to the cross braces is still less. In addition, model updating and calculation often depend on traditional finite element forward analysis, and the method is large in calculation amount, limited in response speed, and difficult to support multi-scheme rapid evaluation and on-site real-time decision. With the development of data driving methods such as deep learning, part of researches start to try to predict bridge displacement, deflection or cable force by using a neural network so as to assist in structural state evaluation. However, most of the existing methods are based on sensor data of a single source, rich visual information such as video images is not fully utilized, and unified modeling is not performed by combining to