Search

CN-122020789-A - Intelligent identification method for quality defects of assembled structure based on VR technology

CN122020789ACN 122020789 ACN122020789 ACN 122020789ACN-122020789-A

Abstract

The invention discloses an intelligent identification method for quality defects of an assembled structure based on a VR technology, and relates to the field of quality detection of constructional engineering. The method comprises the eight steps of VR scene initialization, model loading, defect feature library construction and the like, BIM, GIS and machine learning technologies are fused, a 1:1 virtual scene aligned with a real + -2 cm space is built, a detection path is optimized, multiple virtual detection tools are called, hidden and combined defects are accurately identified, multi-source data fusion generates an interactive report, VR rectification simulation and mechanics verification are combined to ensure rectification effects, and a system iteration continuously optimizes an identification and rectification scheme. The method solves the problems of missed detection, low efficiency, blindness in rectification and the like in the traditional detection, realizes the full-flow intelligent control of the quality defects of the assembled structure, greatly improves the detection precision and efficiency, and reduces the reworking risk.

Inventors

  • WANG PUDONG

Assignees

  • 王普东

Dates

Publication Date
20260512
Application Date
20260121

Claims (9)

  1. 1. The intelligent identification method for the quality defects of the assembled structure based on the VR technology is characterized by comprising the following steps: Step S1, VR scene initialization and model loading, namely building a three-dimensional virtual scene matched with an actual assembly building 1:1 by means of a Unity engine, integrating the geometric shape, material properties and mechanical parameters of components, optimizing a BIM model by means of an autonomously developed lightweight plug-in unit, introducing GIS geographic information data to realize the spatial alignment of the virtual scene and a real site, supporting multi-format BIM model file analysis, carrying out fine modeling on key nodes and setting collision detection grids, adopting a material baking technology to enhance immersion, presetting multiple types of detection scenes and supporting quick switching; S2, constructing and marking a defect feature library, namely, using key detection point coordinates in a BIM model database as space positioning references, developing a parameter-defect association algorithm to match common defect types, constructing a defect three-dimensional model by combining multi-source data through a reverse modeling technology, introducing a defect equivalent elastic modulus calculation model to represent the influence of defects on the mechanical properties of materials, developing a semantic label system to associate three-dimensional attributes of geometry, characteristics and specifications for the defect model, adopting a hierarchical defect database architecture, and supporting quick calling, personalized generation and dynamic updating of the defect model; Step S3, VR interactive detection path planning, namely developing a defect priority algorithm, determining detection priority by combining defect severity level and detection difficulty, binding a detection path with an assembly type construction flow, dynamically filtering detection points of an unworked area and embedding time dimension constraint, adopting an improved Dijkstra algorithm to plan the detection path, setting defect differentiation weight, developing intelligent voice guidance and visual guidance functions, and linking a path planning result with a defect library updating mechanism; S4, invoking a virtual detection tool by a detection personnel through VR handle motion induction, wherein the virtual detection tool comprises a virtual ultrasonic detector, a laser range finder, a piezoresistor resistance sensor, an impact vibration detection tool and a virtual rebound instrument, each tool calculates and feeds back a detection result in real time according to preset parameters and defect feature library data, and a part of the tools are matched with a touch feedback function; S5, intelligent defect identification and marking, namely, extracting data characteristic parameters acquired by a virtual detection tool by a system, comparing the data characteristic parameters with preset model parameters of a defect characteristic library, identifying defect types and grades in a mode of threshold judgment, relevance analysis and the like, marking defects in a specific visual form in a VR scene, associating a correction scheme, introducing an AR space marking function to realize synchronization of virtual defect marking and real space, and supporting new addition of novel defects and feature extraction; Step S6, multi-source data fusion and report generation, namely matching defect marking information with BIM model space coordinates to form a three-dimensional data chain of defect-component-project, fusing normative attribute and correction scheme data, developing a multi-mode data fusion engine to integrate multi-tool detection data, automatically generating an interactable three-dimensional PDF report and a light-weight format report, and associating relevant responsibility party information by a development responsibility tracing module to generate a report file containing a defect position two-dimensional code; Step S7, defect correction simulation and verification, namely screening an optimal correction scheme according to defect types and related conditions, constructing a correction simulation environment in a VR scene, developing a virtual mechanical property verification module, verifying the mechanical properties of the corrected component based on a finite element model and a deflection and elastoplastic stress calculation model, supporting construction flow pre-modeling and cross-specialty collaborative simulation, generating a virtual correction acceptance report, and encrypting and storing acceptance data through a block chain technology; S8, system iteration and learning are carried out, namely, correction simulation key data are collected, effectiveness of a correction scheme is evaluated, priority of the correction scheme is adjusted, a K-means algorithm is adopted to cluster complex correction cases to generate a defect template, a machine learning model is retrained by combining novel defect data through a transfer learning technology, standard clauses and defect classification standards are updated regularly, hardware interaction logic is optimized, and a development module expansion interface supports system function expansion.
  2. 2. The intelligent recognition method for the quality defects of the assembled structure based on the VR technology according to claim 1 is characterized in that in the step S1, the lightweight plug-in unit judges key components through analysis of mechanical parameter weights of components, the weight of a bearing component is more than or equal to 0.8, the weight of the bearing component is judged to be the key component, triangle grid simplification processing is adopted for non-key components, the number of grids is reduced by 60%, the loading speed of a model is improved by more than 40%, the component position and orientation errors of a virtual scene and a real field are controlled within +/-2 cm, the collision detection grid specification is 10cm multiplied by 10cm, and voice prompt is automatically triggered when the distance between a detection person and a key detection point is less than 50 cm.
  3. 3. The intelligent recognition method for the quality defects of the assembled structure based on the VR technology according to claim 1, wherein in the step S2, the defect equivalent elastic modulus calculation model formula is: Wherein: equivalent elastic modulus (MPa) for the defect region; the elastic modulus (MPa) of the material in the defect-free area is 3.0X10 4 MPa for concrete and 2.06X10 5 MPa for steel; for the defect type influence coefficient, the crack is 0.85, the hole is 0.92, and the honeycomb is 0.78; the length, width, depth (m) of the defect, respectively; taking the volume of the range of 0.5m multiplied by 0.5m around the center of the defect for the unit volume (m 3) of the component where the defect is located; The defect volume influence index is 1.2-1.5, the tension member is 1.5, and the compression member is 1.2 according to the stress level adjustment of the member; Taking a value of 0.3-0.5 as a material strength correction coefficient; Measuring the compressive strength (MPa) of the concrete cube for the component; Is the standard value (MPa) of the compressive strength of the concrete cube. The bottom layer of the hierarchical defect database architecture is a basic defect unit, the middle layer is a combined defect type, and the top layer is a project-level defect distribution model.
  4. 4. The intelligent identification method for the quality defects of the assembled structure based on the VR technology according to claim 1 is characterized in that in the step S3, a defect priority score calculation formula is that a priority score = severity level x 0.7+ detection difficulty x 0.3, a score of not less than 3.5 is judged to be a high priority defect, in the improved Dijkstra algorithm, the crack defect weight is 1.5, the hole defect weight is 2.0 and the rib defect weight is 2.5, and the vision guiding function is realized by superposing red arrows with the length of 10cm and the diameter of 1cm in a VR view.
  5. 5. The intelligent identification method for the quality defects of the assembled structure based on the VR technology is characterized in that in the step S4, the moving speed of a virtual probe of a virtual ultrasonic detector is 0.5m/S, the longitudinal wave speed 3500m/S and the transverse wave speed 2000m/S in concrete are referred, the design parameters of components of the laser range finder are preset, 3 times of measurement values are automatically recorded and averaged, red highlighting prompt is carried out when deviation exceeds +/-8 mm, the excitation frequency of the piezoresistor anti-sensor is divided into five steps which are adjustable, namely 180kHz, 200kHz, 220kHz, 240kHz and 260kHz, the touch feedback glove matched with the impact vibration detection tool simulates vibration handfeel when the defect that the compactness is lower than 70% is detected, the virtual rebound detector presets intensity data of a maintenance test block under the same conditions, and a virtual rebound value is displayed according to an intensity attenuation model.
  6. 6. The intelligent recognition method for the quality defects of the assembled structure based on the VR technology according to claim 1 is characterized in that in the step S5, when the waveform distortion rate detected by the virtual ultrasonic detector is more than or equal to 30%, the virtual ultrasonic detector is judged to be a slight gap defect and marked by a yellow semitransparent mask with the thickness of 5cm, when the perpendicularity deviation detected by the laser range finder is exceeded, the virtual ultrasonic detector is marked by a red three-dimensional arrow with the length of 10cm and the diameter of 1cm, when the peak width ratio detected by the impact vibration detection tool is exceeded by a normal range, the virtual ultrasonic detector is judged to be a defect of insufficient bonding of a splicing interface, an orange grid is used for covering a defect area, the grid density is dynamically changed along with the severity of the defect, and the delay of the AR space marking function is controlled within 0.5 seconds.
  7. 7. The intelligent identification method for quality defects of an assembled structure based on the VR technology according to claim 1 is characterized in that in step S6, fusion weights of the multi-mode data fusion engine to ultrasonic waveform data, laser ranging data and rebound value data are respectively 0.4, 0.3 and 0.3, a time axis step of the three-dimensional PDF report is 1 day, a defect detail information card comprises 10 contents, and a responsibility matrix table comprises fields of component numbers, defect types, production groups, installation units, detection personnel, responsibility proportions and the like.
  8. 8. The intelligent recognition method for the quality defects of the assembled structure based on the VR technology according to claim 1, wherein in the step S7, the screening standard of the optimal modification scheme is 80% of modification cost less than or equal to budget, modification period less than or equal to 7 days, qualification rate after modification is more than or equal to 98%, the finite element model is constructed by ANSYS software, mesh division size is 2cm, and the deflection calculation model formula is as follows: Wherein: Is the total deflection (m) of the component; is the deflection generated by the self-weight, For the deflection created by the live load, For the deflection caused by the temperature change, The residual effect is modified for defects to produce deflection.
  9. 9. Elastoplastic stress calculation model (based on Mises yield criterion): Wherein: Equivalent stress (MPa); three principal stresses (MPa), respectively; the stress is allowed for the material. The deflection allowable value is less than or equal to L/250, the allowable stress value is less than or equal to the allowable stress of the material.

Description

Intelligent identification method for quality defects of assembled structure based on VR technology Technical Field The invention relates to the technical field of quality detection of constructional engineering, in particular to an intelligent identification method for quality defects of an assembled structure based on a VR technology, which is suitable for quality defect detection, correction and management in the whole period of production, installation and operation and maintenance of the assembled building components, and can be used for realizing accurate identification and whole-flow management and control of multi-dimensional quality problems such as key nodes, material performance, geometric parameters and the like of the assembled structure. Background Along with the continuous improvement of the industrialization degree of the assembled building, the structural quality of the assembled building is directly related to the overall safety and service life of the building, but the assembled structural members are various in types and complex in splicing nodes, and the traditional quality detection method still has various limitations. At present, the quality detection of an assembled structure depends on manual on-site investigation, and is combined with single tools such as ultrasonic waves and rebound instruments to detect, so that the labor intensity is high, the detection efficiency is low, the detection omission and the erroneous judgment are easily caused by human experience differences, and the hidden defects such as the sleeve grouting is not full, the splicing seam is leaked, and the like are difficult to accurately identify. Meanwhile, the conventional BIM model has the problems of loading and clamping in visual application and insufficient accuracy of spatial alignment with a real site, can not provide an immersive and accurate detection environment for detection personnel, has a lack of scientific planning on a detection path, often has the conditions of repeated detection or missing of key detection points, and has the problems of scattered detection data and difficult effective fusion of multi-source data, so that quality report generation is delayed and responsibility tracing is difficult. In addition, the defect correction scheme lacks visual simulation and mechanical property verification, the correction effect is difficult to prejudge, the reworking waste is easily caused, the existing detection system lacks autonomous learning and iteration capability, the novel defect cannot be adapted to the updated industry specification, and the requirement of high-quality development of the fabricated building is difficult to meet. Disclosure of Invention The invention aims to provide an intelligent recognition method for quality defects of an assembled structure based on VR technology, so as to solve the problems of the background technology. In order to achieve the purpose, the invention provides the following technical scheme that the intelligent identification method for the quality defects of the assembled structure based on the VR technology is characterized by comprising the following steps: Step S1, VR scene initialization and model loading, namely building a three-dimensional virtual scene matched with an actual assembly building 1:1 by means of a Unity engine, integrating the geometric shape, material properties and mechanical parameters of components, optimizing a BIM model by means of an autonomously developed lightweight plug-in unit, introducing GIS geographic information data to realize the spatial alignment of the virtual scene and a real site, supporting multi-format BIM model file analysis, carrying out fine modeling on key nodes and setting collision detection grids, adopting a material baking technology to enhance immersion, presetting multiple types of detection scenes and supporting quick switching; S2, constructing and marking a defect feature library, namely, using key detection point coordinates in a BIM model database as space positioning references, developing a parameter-defect association algorithm to match common defect types, constructing a defect three-dimensional model by combining multi-source data through a reverse modeling technology, introducing a defect equivalent elastic modulus calculation model to represent the influence of defects on the mechanical properties of materials, developing a semantic label system to associate three-dimensional attributes of geometry, characteristics and specifications for the defect model, adopting a hierarchical defect database architecture, and supporting quick calling, personalized generation and dynamic updating of the defect model; Step S3, VR interactive detection path planning, namely developing a defect priority algorithm, determining detection priority by combining defect severity level and detection difficulty, binding a detection path with an assembly type construction flow, dynamically filtering detection points