Embedded part deviation prevention method based on dynamic twinning and real-time deviation correction

Abstract

The invention discloses an embedded part deviation prevention method based on dynamic twinning and real-time deviation correction, and the method comprises the steps: building a BIM entity model with a GUID for each embedded part, and generating a two-dimensional code digital identity card through collision detection; an automatic leveling total station is used for millimeter-level lofting, NB-IoT dip angle-displacement sensors are arranged in the reinforcing steel bar-concrete stage, and coordinates are uploaded in real time; the cloud pushes yellow / red early warning and three-dimensional deviation correction vectors according to double criteria of a threshold value and a trend, a screw jack or a nut fine adjustment mechanism is used for real-time deviation correction, and pouring is carried out after repeated measurement is qualified; writing a completion coordinate into the IFC4 digital twinborn body for delivery; all deviation data are input into a database, an association rule mining algorithm is operated to generate a Top3 high-frequency mode CSV, and technical disclosure attributes of a next project are written in batches through a BIM plug-in. On-site verification shows that the lofting error is smaller than or equal to + / -1mm, the installation deviation is smaller than or equal to + / -1.8 mm, the rework rate approaches zero, and data in the whole process can be traced. The method is suitable for concrete structures such as housing buildings, bridges and subways.

Description

Technical Field

[0001] This invention relates to the field of building engineering measurement and quality control technology, and in particular to a method for correcting millimeter-level deviations of embedded parts before concrete pouring by using BIM solid models, NB-IoT sensors and cloud closed-loop control. Background Technology

[0002] The installation accuracy of embedded parts directly affects the quality of subsequent electromechanical, curtain wall, and decoration work. Traditional embedded part construction mainly relies on two-dimensional CAD drawings, requiring on-site personnel to manually interpret plans, elevations, sections, and detailed nodes. Due to the large number of embedded parts, their varying specifications, and complex spatial relationships, misunderstandings are prone to occur. After design changes, paper drawings are often outdated, leading to on-site construction based on old drawings and a high rework rate. Industry statistics show that embedded part deviations caused by errors in drawing interpretation alone account for more than 30% of the total rework.

[0003] In recent years, although individual technologies such as BIM clash detection, robotic layout, and IoT sensors have emerged, existing technologies have not yet integrated "BIM clash detection—robotic layout—real-time sensor monitoring—mechanical fine-tuning—knowledge base feedback" into a complete process that can be replicated on-site. Furthermore, they lack verifiable quantitative accuracy data (layout error, sensor battery life, fine-tuning resolution) and closed-loop management parameters (early warning threshold, regression time window, knowledge base update cycle). Design, construction, and acceptance parties still use different information carriers, making real-time data synchronization impossible. The accountability traceability period is as long as 7–14 days, with rework costs of 2,000–5,000 yuan per item, and rework costs for large projects often reaching millions of yuan.

[0004] Therefore, the industry urgently needs a closed-loop management method that prioritizes prevention, ensures process control, and enables real-time correction, in order to fundamentally eliminate the accumulation of deviations in embedded parts and the resulting high rework costs. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method for preventing deviation of embedded parts based on dynamic twinning and real-time correction, which aims to detect and correct the installation deviation of embedded parts in real time and at the millimeter level before concrete pouring, so as to eliminate rework afterward.

[0006] The technical solution adopted by this invention to solve the technical problem is as follows: A method for preventing deviation of embedded parts based on dynamic twinning and real-time correction includes the following steps: S1. Create a BIM entity model with a unique GUID code for each embedded part, and store the design coordinates and allowable deviation threshold of ±3 mm in the model properties. S2. Perform virtual collision detection on the BIM entity model and the reinforcement, pipelines and formwork to generate an optimized target coordinate dataset; S3. Store the target coordinate dataset into a QR code label, and paste the QR code label onto the surface of the corresponding entity embedded part to form a digital ID card that corresponds one-to-one with the GUID. S4. Import the target coordinate dataset into a total station with automatic leveling function. The total station projects the center point of the embedded part on site to complete the millimeter-level layout with a layout error of ≤±1mm. S5. During the rebar binding and concrete pouring stage, a low-power NB-IoT tilt-displacement sensor is temporarily magnetically fixed on the embedded part. The sensor has a coordinate accuracy of ±1mm, a sampling interval of 1s, and a battery life of ≥7d. It collects three-dimensional coordinates, horizontality and verticality in real time and uploads them to the cloud data center. S6. The cloud data center performs threshold judgment on the received coordinate values: when the absolute value of the real-time deviation is ≥2.4mm, a yellow warning is immediately pushed to the mobile phone of the responsible personnel; when the linear regression slope of the continuous sampling points within 5 minutes is >0.5mm / min and the absolute value of the deviation after 2 minutes of prediction is ≥3mm, a red warning is pushed and the specific correction amount is given. S7. Based on the three-dimensional correction vector pushed by the mobile app, the responsible personnel use the readily available screw jacks or nut fine-tuning mechanisms on site to lift or lower the embedded parts. After adjustment, the coordinates are remeasured using the same total station. The remeasurement results are transmitted back to the cloud via the mobile app. The transmitted data includes GUID, measured coordinates and remeasurement timestamp, forming a closed loop of early warning, adjustment and remeasurement. S8. After the concrete is poured and the formwork is removed, use the same total station to measure the actual as-built coordinates of the embedded parts, write the measured values ​​into the BIM entity model, and generate an as-built digital twin that is consistent with the actual location on site. S9. Record the measured deviation values ​​of all embedded parts into the deviation set in the database. The set fields should include at least GUID, deviation amount, and deviation reason. Run the association rule mining algorithm on the local computer with support ≥40% and confidence ≥90% to generate a Top 3 high-frequency deviation pattern list. Then, use the BIM plugin to write the list into the technical disclosure attribute fields of the BIM entity model of the next project in batches to achieve continuous improvement. The Top 3 list is saved in CSV format and delivered with the BIM model.

[0007] Furthermore, in step S3, the QR code label uses 3M VHB adhesive, which is alkali-resistant and temperature-resistant from -20 to 80°C. On-site scanning of the code can display the target coordinates and technical briefing video.

[0008] Furthermore, in step S5, the NB-IoT tilt-displacement sensor has a size of ≤30mm×20mm×8mm, a magnetic base with a suction force of ≥5kg, and the assembly and disassembly process does not damage the galvanized layer of the embedded parts.

[0009] Furthermore, in step S6, the linear regression calculation uses only the measured coordinate data from the field, the significance level of the regression coefficient is α=0.05, the prediction step size is fixed at 2min, and it does not rely on an external training model.

[0010] Furthermore, in step S7, the single adjustment accuracy of the screw jack or nut fine-tuning method is ≤0.5mm, and the absolute value of the deviation after adjustment is <2.4mm is considered as closed-loop qualified.

[0011] Furthermore, in step S8, the as-built digital twin is exported in IFC4 format, including GUID, measured coordinates, signature of the responsible person, and processing timestamp, for use as a reference for subsequent electromechanical installation.

[0012] Furthermore, in step S9, the local computer is connected to the construction site's local area network and can run the association rule mining algorithm without relying on an external artificial intelligence interface. The algorithm is only executed on the deviation cause field, and the running time is ≤30s.

[0013] Furthermore, the association rule mining algorithm adopts a TID-list bitmap compression structure, and completes the calculation of Top 3 patterns with support ≥40% and confidence ≥90% within 30 seconds in a single-threaded 4C8G environment.

[0014] Furthermore, the database employs column-oriented parallel aggregation, with the export time for 900,000 deviation records being ≤2 seconds, and automatically generates a CSV file for use by the association rule mining algorithm.

[0015] The working principle of this invention can be summarized as a six-step closed loop of "prevention-monitoring-early warning-fine-tuning-retesting-knowledge feedback". The entire process is data-driven and millimeter-level controlled. The database and association rule algorithm can be implemented using domestic or foreign software. The specific process is as follows: 1. Prevention: BIM Digital Identity Card (1) Create a BIM entity model with GUID for each embedded part, and write the design coordinate ±3mm allowable deviation in the attribute; (2) Collision detection between the model and the steel bars / pipelines / formwork, eliminate spatial conflicts in advance, and generate an optimized target coordinate dataset; (3) Write the dataset into a QR code label (3M VHB adhesive), paste it on the surface of the embedded part to form a "digital ID card", and scan the code to obtain the three-dimensional coordinates, handover video, and acceptance standards.

[0016] 2. Monitoring: Robotic layout + NB-IoT sensors (1) Import the target coordinates into the automatic leveling total station, project the center point on site, and the layout error is ≤ ±1mm; (2) During the steel bar binding-concrete pouring stage, the magnetic NB-IoT tilt-displacement sensor (30×20×8mm, battery life ≥7d, coordinate accuracy ±1mm, 1s sampling) uploads the three-dimensional coordinates, horizontality and verticality to the cloud data center in real time.

[0017] 3. Early Warning: Threshold + Trend Dual Criteria (1) Threshold criterion: |Real-time deviation|≥2.4mm (80% limit) → Yellow warning for mobile phones; (2) Trend criterion: 5-minute sliding window linear regression slope>0.5mm / min and |deviation|≥3mm after prediction 2 minutes → Red warning and output specific correction amount (X / Y / Z adjustment vector).

[0018] 4. Fine-tuning: Mechanical jacking + closed-loop retesting (1) The responsible personnel use the three-dimensional vector of the mobile app to lift / lower the device using an M24 screw jack or nut fine adjustment mechanism (0.25mm per turn, ≤0.5mm per time); (2) The same total station is immediately re-measured, and the re-measured data (GUID, coordinates, timestamp) is sent back to the cloud. If the |deviation| < 2.4mm, the closed loop is qualified, and the average closed loop time is 8min.

[0019] 5. Re-testing and Completion: Digital Twin Delivery After the formwork is removed, the total station measures the as-built coordinates, automatically updates the BIM entity model, and generates an as-built digital twin in IFC4 format, including GUID, measured coordinates, responsible person's signature, and timestamp, which serves as the benchmark for subsequent electromechanical installation.

[0020] 6. Knowledge Feedback: Continuous Improvement of Database + Association Rule Algorithm (1) All deviation data are entered into the database deviation collection (fields: GUID, deviation amount, deviation reason); (2) The local computer runs the association rule mining algorithm (support ≥40%, confidence ≥90%, ≤30s) to generate the Top 3 high frequency pattern CSV; (3) The BIM plugin automatically writes the CSV into the BIM technology disclosure attribute of the next project, realizing the design-construction linkage optimization, and the rework rate approaches zero.

[0021] This invention, through the above six-step closed-loop process, stably controls the installation deviation of embedded parts within ±1.8mm, with a rework rate close to 0%, and the entire process data is traceable and reusable.

[0022] The positive and beneficial effects of this invention are as follows: 1. Improved accuracy: Layout error ≤ ±1mm, installation deviation stable within ±1.8mm, meeting the ±3mm specification limit, and rework rate close to zero.

[0023] 2. Cost savings: The construction period of a single project is shortened by 3 to 5 days, avoiding additional losses caused by structural cutting and secondary pouring.

[0024] 3. Reduced risk: Correction is completed before concrete pouring, eliminating the damage to reinforcing bars, waterproof layer and prestress caused by subsequent water drilling, making the structure safer.

[0025] 4. Traceability of responsibility: GUID + QR code + timestamp enables full-chain data synchronization from design to construction to acceptance, shortening the responsibility confirmation period from 7-14 days to real time.

[0026] 5. Knowledge Reuse: Database + Association Rule Algorithm automatically generates Top 3 high-frequency deviation patterns, which are updated in real time during the next project design briefing, continuously reducing the defect rate.

[0027] 6. On-site replicability: Total station, NB-IoT sensor, and screw jack are all in stock. Closed loop can be completed in 30 seconds within the local area network, without the need for additional large equipment.

[0028] 7. Value-added delivery: Provide an IFC4 as-built digital twin, allowing subsequent MEP and curtain wall coordinates to be directly accessed, avoiding secondary measurements and improving overall construction efficiency. Detailed Implementation

[0029] The present invention will be further explained and described below with reference to specific embodiments: Example 1: A standard floor of a high-rise residential building, with 420 embedded parts, a design allowable deviation of ±3mm, a floor height of 3.0m, C30 concrete, and a raft foundation thickness of 0.6m.

[0030] 1. BIM Solid Model and Collision Detection (S1-S2) Software: Revit 2025 + Navisworks 2025 Operation: Create a family of embedded parts, add shared parameters GUID, DesignX / Y / Z, AllowDev=3 mm; perform collision checks with the rebar, pipeline, and formwork models, eliminate 16 conflicts in advance, and export the optimized coordinate CSV (fields: GUID, X, Y, Z).

[0031] 2. Digital ID Card (S3) QR code content: GUID|X|Y|Z|Explanation video URL Label: 30mm×30mm PET QR code, 3M VHB adhesive, alkali resistant -20℃~80℃, can be read in 1 second by scanning on site.

[0032] 3. Robotic layout (S4) Equipment: Leica iCR80 total station, automatic leveling, angle accuracy 1″, distance accuracy ±(0.5mm + 1ppm); Process: Import the CSV file into the total station, project the crosshair center point on site, and the layout takes 1.2 hours. The layout error is ≤ ±1 mm (n=420, σ=0.34 mm).

[0033] 4. Sensor monitoring (S5) Sensor: STM32L4 + MPU9250 + BC95-B5, dimensions 30×20×8mm, magnetic base with 5kg suction force, 10-day battery life; Installation: Magnetic attachment to the back rib of the embedded part, completed in 30 seconds; sampling frequency 1Hz, 4G IoT card direct transmission to Alibaba Cloud IoT, end-to-end latency <3s.

[0034] 5. Dual-criteria early warning (S6) Threshold criterion: |Δ| ≥ 2.4 mm → Yellow warning Trend criterion: 5-minute sliding window linear regression, slope > 0.5 mm / min and |Δ| ≥ 3 mm after 2 minutes of prediction → red alert; Push notifications: DingTalk API, average push time 1.8s.

[0035] 6. Mechanical fine-tuning and retesting closed loop (S7) Tools: M24 screw jack, 0.25mm per turn, single adjustment accuracy ≤0.5mm; Closed loop: After adjustment, the same total station was used for re-measurement, and the re-measurement data (GUID, X, Y, Z, timestamp) was transmitted back via the App; the re-measurement pass criterion was |Δ|<2.4 mm, the average closed loop time was 8 minutes, and the pass rate was 100% (420 / 420).

[0036] 7. As-built digital twin (S8) Actual measurement: After the template was removed, the same total station was used for remeasurement. The maximum deviation was 1.8 mm, and the average was 0.7 mm. Delivery: IFC4 format, file size 1.2MB, including GUID, measured coordinates, electronic signature of the responsible person, and timestamp, which can be directly used for subsequent electromechanical installation.

[0037] 8. Continuous improvement of the knowledge base (S9) Database: MongoDB deviation collection, documents {GUID:string, devX:float, devY:float, devZ:float, cause:string, date:ISODate}; Algorithm: Running Apriori on local computer, support=0.4, confidence=0.9, running time 28s (single-threaded 4C8G). Results: The Top 3 pattern (loosening of template tie rods → floating of embedded parts, support rate 42%) automatically generates CSV, and the BIM plugin batch writes the technical disclosure attributes of the BIM entity model of the next project; the floating deviation of subsequent projects is reduced to 0.6mm (a decrease of 67%).

[0038] Example 2: The project overview, hardware configuration, early warning threshold, and fine-tuning mechanism in this example are exactly the same as in Example 1, except that: 1. Database: A domestic cloud-native distributed database (Tencent Cloud TDSQL-C MySQL columnar storage version, 1 million QPS per node, and national cryptographic transmission with security level protection 3.0+) is used. The writing time for 4.2 million deviation records is 1.9 seconds, which is 52% shorter than that of Example 1. 2. Algorithm: The domestic high-performance association rule algorithm Mare (JDK17 single-threaded, support 0.4, confidence 0.9, running time 6.2s, peak memory usage 0.7GB) is adopted, and the output Top 3 pattern is completely consistent with that in Example 1; 3. Compliance and Cost: Data does not leave the country, saving 15 working days of third-party compliance audits and reducing cloud resource fees by 45%.

[0039] The remaining steps, fields, CSV format, and BIM plugin writing methods are consistent with those in Example 1, and will not be repeated here.

[0040] As can be seen from Examples 1 and 2, both software solutions can complete the Top 3 pattern mining with a support rate of ≥40% and a confidence rate of ≥90% within ≤30 seconds, with a rework rate approaching zero. This fully demonstrates that the present invention does not depend on any specific software, and equivalent software replacements from both domestic and international sources can achieve the same result.

[0041] This invention is also applicable to concrete structures such as bridge piers and subway stations. Only the number of sensors and the model of the layout robot need to be adjusted according to the size of the component. The rest of the steps are the same, and the rework rate remains 0%.

[0042] The above are merely preferred embodiments of the present invention. Based on the technical solutions defined in the claims, those skilled in the art can make equivalent substitutions or adjustments to the BIM software version, sensor model, fine-tuning mechanism form, or algorithm parameters without departing from the spirit of the invention; any modifications, equivalent substitutions, or improvements made within the scope of the principles of the present invention should be considered to fall within the protection scope of the present invention.

Claims

1. A pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction, characterized in that, The method comprises the following steps: S1, a BIM entity model with a unique GUID code is established for each embedded part, and the design coordinates and the allowable deviation threshold ±3 mm are stored in the model attribute; S2, virtual collision detection is performed on the BIM entity model, steel bars, pipelines and templates to generate an optimized target coordinate data set; S3, the target coordinate data set is stored in a two-dimensional code label, and the two-dimensional code label is pasted on the surface of the corresponding entity embedded part to form a digital identity certificate corresponding to the GUID; S4, the target coordinate data set is imported into a total station with an automatic leveling function, and the total station projects the center point of the embedded part on site to complete the millimeter-level setting-out with an error of ≤±1 mm; S5, during the steel bar binding and concrete pouring stage, a low-power NB-IoT inclination-displacement sensor is temporarily magnetically attracted and fixed on the embedded part, the sensor has a coordinate accuracy of ±1 mm, a sampling interval of 1 s, and a continuous running time of ≥7 d, and real-time collection of three-dimensional coordinates, levelness and perpendicularity is uploaded to a cloud data hub; S6, the cloud data hub performs threshold judgment on the received coordinate values: when the absolute value of the real-time deviation is ≥2.4 mm, a yellow warning is immediately pushed to the mobile phone of the responsible person; when the linear regression slope of the continuous sampling points within 5 min is >0.5 mm / min and the absolute value of the deviation after 2 min is predicted to be ≥3 mm, a red warning is pushed and the specific correction amount is given; S7, the responsible person adjusts the embedded part by using the on-site commonly used screw jacks or nut fine adjustment mechanisms according to the three-dimensional correction vector pushed by the mobile phone App, and the same total station is used to re-measure the coordinates, the re-measured results are returned to the cloud through the mobile phone App, and the returned data includes the GUID, the measured coordinates and the re-measurement timestamp, forming a closed loop of warning, adjustment and re-measurement; S8, after the concrete pouring is completed and the formwork is removed, the actual completed coordinates of the embedded part are measured by using the same total station, the measured values are written into the BIM entity model, and a completed digital twin body consistent with the position of the on-site physical object is generated; S9, the measured deviation values of all embedded parts are recorded in the database deviation set, and the set field at least includes GUID, deviation amount and deviation reason; an associated rule mining algorithm is run in a local computer with a support degree ≥40% and a confidence degree ≥90%, a Top3 high-frequency deviation mode list is generated, and the list is written into the technical disclosure attribute field of the BIM entity model of the next project through a BIM plug-in to realize continuous improvement; the Top3 list is saved in CSV format and delivered together with the BIM model.

2. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1, characterized in that: In step S3, the two-dimensional code label uses 3M VHB adhesive, is alkali-resistant and temperature-resistant-20-80℃, and can display the target coordinates and technical disclosure video by scanning the code on site.

3. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1, characterized in that: In step S5, the NB-IoT inclination-displacement sensor has a size of ≤30mm×20mm×8mm, a magnetic attraction base with a suction force of ≥5kg, and does not damage the galvanized layer of the embedded part during disassembly and assembly.

4. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1, characterized in that: In step S6, the linear regression calculation only uses the on-site measured coordinate data, the regression coefficient significance level α=0.05, and the prediction step is fixed at 2 min, which does not depend on external training models.

5. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1, characterized in that: In step S7, the screw jack or nut fine adjustment mode single adjustment precision ≤0.5mm, adjustment is completed after the retest deviation absolute value <2.4mm is considered as closed loop qualified.

6. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1, characterized in that: In step S8, the completed digital twin is exported in IFC4 format, including GUID, measured coordinates, responsible person signature and processing timestamp, for subsequent mechanical and electrical installation reference.

7. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1, characterized in that: In step S9, the local computer is connected with the construction site LAN, and can run the association rule mining algorithm without relying on external artificial intelligence interface, and the algorithm is only executed on the deviation reason field, and the running time is ≤30s.

8. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1 or 7, characterized in that: The association rule mining algorithm adopts TID-list bitmap compression structure, and completes Top3 mode calculation with support ≥40% and confidence ≥90% in 30s in single-thread 4C8G environment.

9. The pre-embedded part deviation prevention method based on dynamic twin and real-time deviation correction according to claim 1 or 7, characterized in that: The database adopts column storage parallel aggregation, and 900,000 deviation records are exported in ≤2s, and CSV is automatically generated for association rule mining algorithm calling.