Precision control and deviation correction method for large-span space truss roof sliding construction track

Abstract

The invention discloses a large-span space truss roof sliding construction rail precision control and deviation correction method, and belongs to the technical field of steel structure construction. According to the method, a three-stage closed-loop control system is adopted, wherein in first-stage reference control, high-precision positioning of the pre-buried track is driven by + / -0.5 mm through a BIM model; in the second-stage dynamic monitoring, a multi-source sensor network comprises a laser tracker, a tilt angle sensor and a strain gauge array, track displacement, attitude and stress data are collected in real time, and millimeter-level monitoring is achieved in combination with a Kalman filtering algorithm; a hydraulic servo system is adopted for three-stage active deviation correction, the deviation trend is predicted based on an LSTM neural network, a hydraulic execution mechanism is driven through a fuzzy PID control algorithm, the response time is shorter than 8 seconds, and the deviation correction precision is + / -0.1 mm. According to the method, the construction precision is improved to + / -1.5 mm, the deviation rectifying efficiency is improved by 90%, and the method is suitable for large-span projects such as stadiums and airport terminals.

Description

Technical Field

[0001] The present invention relates to a large-span steel structure construction technology, and is particularly aimed at the track precision control and dynamic deviation correction problems in the sliding construction of a space truss roof with a span exceeding 150m. Background Art

[0002] Traditional sliding construction relies on manual measurement and mechanical limiters, which suffer from low accuracy of ±5mm, slow response times (often exceeding 30 minutes), and uncontrollable cumulative errors. Comparative literature uses a single laser rangefinder for monitoring, which cannot detect track torsion. Other literature uses hydraulic jacking for correction, but this lacks predictive control. This present invention utilizes multi-source sensing and intelligent algorithms to achieve submillimeter-level real-time correction. Summary of the Invention

[0003] In order to solve the problem of track accuracy and dynamic correction in the sliding construction of large-span structures:

[0004] (1) High-precision pre-embedding: Generate CNC machining codes for embedded parts based on the BIM model to ensure that the initial installation error is ≤±0.5mm; perform high-precision pre-embedding work accurately.

[0005] (2) Intelligent monitoring:

[0006] Laser tracker networking enables total station-level positioning; strain gauge arrays detect local buckling risks in the track; positioning and detection are performed simultaneously to ensure the normal operation of the intelligent monitoring system.

[0007] (3) Predictive correction:

[0008] LSTM network learns historical deviation patterns; participates in the first task of predictive deviation correction;

[0009] Fuzzy PID control of hydraulic cylinder output:

[0010] (1) Construction axis deviation ≤ ±1.5mm, the original traditional method ±5mm;

[0011] (2) Correction response time ≤ 8 seconds, the original traditional 30 minutes;

[0012] (3) Reduce manual intervention by 75%.

[0013] The technical solution adopted in the present invention is as follows:

[0014] A track precision control and deviation correction system for sliding construction of a large-span space truss roof is characterized by including the following contents:

[0015] In the early stage of sliding construction, a pre-embedded positioning subsystem is established. A numerical control positioning device (2) based on the BIM model is used to control the installation accuracy of the track embedded parts to ≤±0.5mm. Track embedded parts usually consist of the following parts: fixing parts, positioning parts and connecting parts. The fixing parts are used to fix the connection between the track base and the embedded parts, and are fixed by bolts, welding, etc. The positioning parts are used to ensure the accuracy of the position and relative position of the track. The connecting parts are used to connect the embedded parts and the track base to ensure their rigid connection and stability.

[0016] During the sliding construction, a multi-source monitoring subsystem (3) is established. The multi-source monitoring subsystem includes a laser tracker (4) with an accuracy of ≤0.1mm. The multi-source control subsystem also includes a MEMS tilt sensor with a resolution of ≤0.01°. The multi-source control subsystem is finally composed of a monitoring network composed of the laser tracker (4), the MEMS tilt sensor, and the distributed optical fiber strain sensor. The node spacing is ≤10m. The whole system can accurately control the track accuracy (5).

[0017] After the multi-source monitoring subsystem (3) completes its work, a hydraulic correction subsystem (6) is implemented: it includes a bidirectional hydraulic servo cylinder (7) with an output of 200kN and a stroke of 200mm, a servo proportional valve (8) with a frequency response of ≥50Hz, and a high-precision displacement sensor (9) with an accuracy of 0.1mm; in conjunction with the sliding construction correction control system, it includes distributed edge computing nodes, a wireless Mesh communication network, and a visual monitoring platform, supporting multi-node collaborative decision-making and manual intervention operations;

[0018] An intelligent control platform is used for the entire process of sliding construction: an LSTM deviation prediction model and a fuzzy PID control algorithm are integrated to realize closed-loop control from data acquisition to deviation correction execution (1); the topological structure of the LSTM neural network includes an input layer, including a 12-dimensional feature vector, 3 hidden layers, each with 128 nodes, and an output layer, including the X / Y / Z axis deviation prediction value for the next 5 seconds; the input layer performs intelligent control work on the working platform, and finally the hidden layer cooperates with the output layer to perform the final work.

[0019] Furthermore, the track precision control and deviation correction system for sliding construction of a large-span space truss roof is characterized by including the following contents:

[0020] Embedded track (16): The three-dimensional coordinates of the embedded parts are generated based on the BIM model, and the track connectors are processed using CNC machine tools. The track embedded parts usually consist of the following parts: fixing parts, positioning parts and connecting parts. The fixing parts are used to fix the connection between the track base and the embedded parts, and are fixed by bolts, welding, etc. The positioning parts are used to ensure the accuracy of the position and relative position of the track. The connecting parts are used to connect the embedded parts and the track base to ensure their rigid connection and stability.

[0021] Real-time monitoring (17): Track displacement, inclination, and strain data are collected at a frequency of ≥100 Hz through a multi-source sensor network. The hierarchical structure of multi-source sensor fusion can be divided into three levels according to the degree of abstraction of information processing: data layer fusion, feature layer fusion, and decision layer fusion.

[0022] Data-level fusion, also known as pixel-level fusion, directly fuses sensor observation data. This method requires homogeneous sensors and is computationally intensive, but produces the most accurate results. Feature-level fusion, an intermediate level, extracts features from each sensor's observation data and then fuses them into a single feature vector for processing. This requires relatively low computational effort and communication bandwidth, but accuracy is reduced. Decision-level fusion, a high-level fusion process, processes the condensed results of sensor data. While this method minimizes computational effort and communication bandwidth requirements, it produces relatively inaccurate results.

[0023] Deviation prediction (18): The monitoring data is input into the LSTM neural network to predict the track deviation trend within the next 5 seconds;

[0024] Correction decision (19): Based on the prediction results, the output of the hydraulic system is calculated through a multi-objective optimization algorithm;

[0025] Tiered Execution (20): Triggering three levels of response based on deviation thresholds:

[0026] Level 1, where the deviation is 2-5mm: sound and light alarm + automatic fine-tuning;

[0027] Level 2, where the deviation is 5-10mm: construction speed is reduced + continuous deviation correction;

[0028] Level 3, where the deviation is greater than 10mm: emergency braking + manual intervention;

[0029] Beneficial effects of the present invention

[0030] (1) For the first time, a numerical control positioning device based on the BIM model was proposed to improve the track accuracy, ensuring that the track positioning error during the entire construction process is controlled within ±2mm, which is better than the industry standard of ±5mm.

[0031] (2) Compared with manual deviation correction, the present invention adopts a multi-source monitoring subsystem to optimize efficiency, among which automatic deviation correction reduces the frequency of manual intervention by 90% and increases the daily slip progress by 40%.

[0032] (3) An integrated LSTM deviation prediction model and fuzzy PID control algorithm are proposed to reduce risks, ensure early prediction of deviation trends, and avoid sudden structural instability accidents.

[0033] (4) The present invention realizes sub-millimeter level real-time deviation correction through multi-source sensing and intelligent algorithms.

[0034] Ensure overall cost savings: Reduce rework losses and save 3%-5% of steel usage in typical projects.

[0035] (5) The present invention also has technical extensibility, and the core algorithm used can be migrated to fields such as bridge jacking and heavy equipment installation. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 : Track control flow chart for sliding construction of large-span space truss roof of the present invention;

[0037] Figure 2 : The overall construction drawing of the sliding construction track structure of the large-span space truss roof of the present invention;

[0038] Figure 3 : The present invention provides a numerical control positioning device for sliding construction tracks of large-span space truss roofs;

[0039] Figure 4 : The multi-source monitoring subsystem architecture diagram of the sliding construction track system of the large-span space truss roof of the present invention;

[0040] Figure 5 : Schematic diagram of the track system tracker device for the sliding construction of a large-span space truss roof according to the present invention;

[0041] Figure 6 : Schematic diagram of the hydraulic deviation correction subsystem device for the sliding construction track of the large-span space truss roof of the present invention;

[0042] In the figure, 1 is a three-level flow chart; 3 is a CNC positioning device, 2-1 is a fixing part; 2-2 is a positioning part; 2-3 is a connecting part; 2-4 is a square link; 2-5 is a CNC central axis; 3 is a structural diagram; 4 is a laser tracker; 5 is a sliding track; 6 is a hydraulic correction device; 6-1 is an upper adjustment mechanism; 6-2 is an intermediate adjustment mechanism; 6-3 is a lower adjustment mechanism; 6-4 is a sensing mechanism; 6-5 is a mounting bracket; DETAILED DESCRIPTION

[0043] In order to better explain the present invention and facilitate understanding, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings:

[0044] Refer to the attached Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 , Figure 5 , Figure 6 As a further improvement of the present invention,

[0045] The three-level control process 1 improves the track accuracy control and deviation correction during the sliding construction of large-span spatial truss roofs. Pre-embedding is driven by the BIM model. Tekla Structures is used to establish a track-foundation joint model. A multi-source monitoring subsystem 3 is added for monitoring, including a laser tracker 4. A hydraulic deviation correction subsystem 6 is then used to address key control points. The track centerline coordinate error is ±0.3mm, and the embedded part elevation error is ±0.2mm. CNC machining is performed 2 times. Track connectors are machined using a five-axis CNC machine tool, the DMG MORI NHX6300. Bolt hole machining accuracy is H7, with a tolerance of +0.018 / -0mm.

[0046] Refer to the attached Figure 3 As a further improvement of the present invention, the numerical control device controls the allowable deviation of the axis position to be ±0.5mm, and the measured maximum value is controlled to be 0.43mm.

[0047] Refer to the attached Figure 3 As a further improvement of the present invention, the numerical control device controls the allowable deviation of the embedded part elevation to ±0.5mm, and the measured maximum value is controlled to 0.37mm.

[0048] The numerical control device controls the allowable deviation of the horizontality to ±0.5 / 1000, and the measured maximum value is controlled at 0.3 / 1000.

[0049] Refer to the attached Figure 4 As a further improvement of the present invention, device 4 uses Leica AT960, with an arrangement spacing of one group every 15m, a total of 38 groups, a measurement frequency set to 120Hz, an accuracy index of ±0.015mm / m, and a JAE MW-9010V inclination sensor with an installation density of one monitoring point every 5m. The range is ±15° with a resolution of 0.001°.

[0050] Refer to the attached Figure 5 As a further improvement of the present invention, device 4 performs strain monitoring using a FOS-N-1550-50 fiber Bragg grating sensor. Figure 4 The layout plan is to arrange a monitoring section every 10m along the length of the track, with 4 measuring points arranged in each section, including 2 on the upper / lower flange.

[0051] Refer to the attached Figure 6 As a further improvement of the present invention, the hydraulic cylinder model selected for the device 6 for the correction process is HOB-200 / 200, with an output of 200 kN and a stroke of 200 mm; the servo valve model is MOOG D661-4651, with a frequency response of 65 Hz and a flow rate of 40 L / min; the displacement sensor model is MTS RHM0400MD, with a resolution of 0.001 mm and a repeatability of ±0.002 mm.

[0052] The device 6 adopts correction logic and formulates fuzzy PID parameter self-tuning rules. The input variables are: deviation e, unit mm, deviation change rate ec, unit mm / s, output variable: hydraulic cylinder output F, unit kN, fuzzy set division: NB, negative large, NS, negative small, ZO, zero, PS, positive small, PB, positive large, where the control period is 50ms, kept synchronized with sensor sampling.

Claims

1. A track precision control and deviation correction system for sliding construction of a large-span space truss roof, characterized by: include: A three-level closed-loop control system is established for the entire process of sliding construction: the first-level reference control drives the high-precision positioning of the embedded track (16) with a precision of ±0.5mm through the BIM model; the second-level dynamic monitoring uses a multi-source sensor network, including laser trackers, tilt sensors, and strain gauge arrays, to collect track displacement, posture, and stress data in real time, and combines the Kalman filter algorithm to achieve millimeter-level real-time monitoring (17); the third-level active deviation correction uses a hydraulic servo system, which predicts deviation trends based on the LSTM neural network (18), drives the hydraulic actuator through the fuzzy PID control algorithm, and finally generates a deviation correction decision (19). The response time is less than 8 seconds, and the deviation correction accuracy is ±0.1mm. The three-level closed-loop control system performs hierarchical execution (20). While each of them works independently, it also forms a continuous closed working form. Establish a pre-embedded positioning subsystem in the early stage of sliding construction, set up a numerical control positioning device based on the BIM model (2), and control the installation accuracy of the track pre-embedded parts to ≤±0.5mm; During the sliding construction, a multi-source monitoring subsystem (3) is established. The multi-source monitoring subsystem includes a laser tracker (4) with an accuracy of ≤0.1mm. The multi-source control subsystem also includes a MEMS tilt sensor with a resolution of ≤0.01°. The multi-source control subsystem is finally composed of a monitoring network composed of the laser tracker (4), the MEMS tilt sensor, and the distributed optical fiber strain sensor. The node spacing is ≤10m. The whole system can accurately control the track accuracy (5). After the multi-source monitoring subsystem (3) is completed, a hydraulic deviation correction subsystem (6) is implemented: comprising a bidirectional hydraulic servo cylinder (7) with an output of 200 kN and a stroke of 200 mm, a servo proportional valve (8) with a frequency response of ≥50 Hz, and a high-precision displacement sensor (9) with a resolution of 0.01°; An intelligent control platform is used for the entire process of sliding construction: integrating the LSTM deviation prediction model and the fuzzy PID control algorithm to achieve closed-loop control from data collection to deviation correction execution (1).

2. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 1 is characterized by: The control parameters of the hydraulic deviation correction subsystem (6) satisfy: Proportional gain Kp = 1.2 + 0.2 (1.2 - 0.2); Integration time Ti = 20s; Differential time Td = 5s.

3. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 1: the three-level closed-loop control system comprises the following steps: (1) Embedded track (16): Generate the three-dimensional coordinates of the embedded parts based on the BIM model, and use CNC machine tools to process the track connectors; Track embedded parts usually consist of the following parts: fixing parts (2-1), positioning parts (2-2) and connecting parts (2-3); fixing parts are used to fix the connection between the track base and the embedded parts, and are fixed by bolts, welding, etc.; positioning parts are used to ensure the position and relative position of the track; connecting parts are used to connect the embedded parts and the track base to ensure their rigid connection and stability; (2) Real-time monitoring (17): Track displacement, inclination, and strain data are collected at a frequency of ≥100 Hz through a multi-source sensor network. The hierarchical structure of multi-source sensor fusion can be divided into three levels according to the degree of abstraction of information processing: data layer fusion, feature layer fusion, and decision layer fusion. Data-level fusion, also known as pixel-level fusion, directly fuses sensor observation data. This method requires homogeneous sensors and is computationally intensive, but produces the most accurate results. Feature-level fusion, an intermediate level, extracts features from each sensor's observation data and then fuses them into a single feature vector for processing. This requires relatively low computational effort and communication bandwidth, but accuracy is reduced. Decision-level fusion, a high-level fusion process, processes the condensed results of sensor data. While this method minimizes computational effort and communication bandwidth requirements, it produces relatively inaccurate results. (3) Deviation prediction (18): The monitoring data is input into the LSTM neural network (26) to predict the orbit deviation trend within the next 5 seconds; (4) Correction decision (19): Based on the prediction results, the output of the hydraulic system is calculated using a multi-objective optimization algorithm; (5) Tiered Execution (20): Triggering three-level responses based on deviation thresholds: Level 1, where the deviation is 2-5mm: sound and light alarm + automatic fine-tuning; Level 2, where the deviation is 5-10mm: construction speed is reduced + continuous deviation correction; Level 3, where the deviation is greater than 10mm: emergency braking + manual intervention.

4. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 3 is characterized by: The topological structure of the LSTM neural network (26) includes an input layer, three hidden layers and an output layer. The input layer monitors data input. The three hidden layers improve the complexity and fitting ability of the network and enhance the stability of data output. Finally, the output layer obtains deviation prediction and obtains the future track deviation trend. Each layer of the hidden layer has 128 nodes and the output layer, including the X / Y / Z axis deviation prediction value for the next 5 seconds. The input layer performs intelligent control work on the working platform, and finally the hidden layer cooperates with the output layer to perform the final work.

5. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 3 is characterized by: The embedded track (16) generates track installation error pre-compensation parameters by fusing BIM model data (2) with on-site laser scanning point clouds, realizes three-dimensional dynamic calibration (21) in the track laying stage, and completes the sliding track layout.

6. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 5 is characterized by: The three-dimensional dynamic calibration (21) during the track laying phase is arranged along the sliding track, including a fiber grating sensor (10) including an outer resin (11), a cladding (12), an optical fiber core (13), a fiber core (14), a MEMS inertial unit and a multimodal sensing network (22) of a laser rangefinder, with a sampling frequency of not less than 50 Hz.

7. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 6 is characterized by: The multimodal perception network (22) performs real-time fusion processing on the perception data, constructs a track deformation state matrix (23), and predicts the deviation trend within the next 5 seconds, generating a prediction deviation (15).

8. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 7 is characterized by: When the predicted deviation (15) of the track deformation state matrix (23) exceeds a threshold, the three-level deviation correction actions of speed difference adjustment, lateral thrust and truss posture adjustment are triggered in sequence to form a closed-loop control (1).

9. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 8 is characterized by: The closed-loop control (1) includes distributed edge computing nodes, a wireless Mesh communication network and a visual monitoring platform (24), and supports multi-node collaborative decision-making and manual intervention operations.

10. The track precision control and deviation correction system for sliding construction of a large-span space truss roof according to claim 9 is characterized by: The wireless Mesh communication network and the visual monitoring platform (24) are integrated with a digital twin module, which can compare actual construction data with the simulation model in real time and dynamically optimize the correction parameters.

Images