Building quality inspection management system

Abstract

The invention discloses a building quality inspection management system, and the system comprises a distributed intelligent sensor network which comprises a multi-type sensor array disposed at a key node of a building structure, a mobile detection terminal, and an unmanned plane inspection unit; the multi-modal data fusion module is used for performing space-time alignment and feature extraction on the sensor data, the image data and the BIM model data; the digital twinborn visual platform comprises a quality evaluation engine, a defect traceability module and a three-dimensional visual interface; the defect detection rate in a dark light environment is obviously improved, full-coverage detection of floors can be realized in a short time, field sensor data and a BIM model are synchronized in real time through a 5G network, and when a quality problem is found, the system analyzes 12 types of influence factors such as materials, processes and environments through a Bayesian network, so that the quality problem is solved. The system can trace back to specific construction teams, equipment numbers and even meteorological data, and can reduce the requirements of personnel through intelligent detection, thereby reducing the working cost.

Description

Technical Field

[0001] The present invention relates to the technical field of building quality inspection, and in particular to a building quality inspection management system. Background Art

[0002] In the construction industry, quality inspection is crucial for ensuring structural safety and durability. Currently, the industry relies primarily on the following quality inspection methods: Inspectors rely on their experience to determine the quality of building components, such as concrete cracks and rebar spacing, and manually record the inspection results. This results in inconsistent judgments from different inspectors, resulting in high error rates and low efficiency. Furthermore, real-time monitoring is impossible, and paper records are easily lost, difficult to trace, and unable to form a structured database. Summary of the Invention

[0003] The technical problem solved by the present invention is to overcome the defects of the prior art such as reliance on paper records, discrete data, and non-traceability of processes, and to provide a building quality inspection and management system.

[0004] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: a building quality inspection and management system, comprising: a distributed intelligent sensor network comprising a multi-type sensor array, mobile detection terminals, and drone inspection units deployed at key nodes of a building structure;

[0005] Multimodal data fusion module, used for spatiotemporal alignment and feature extraction of sensor data, image data, and BIM model data;

[0006] Digital twin visualization platform, including a quality assessment engine, defect tracing module, and 3D visualization interface;

[0007] Edge computing nodes, deployed at construction sites and running lightweight quality analysis models;

[0008] A cloud server cluster that stores a database on building quality throughout its lifecycle and runs deep learning algorithms;

[0009] The user interaction terminal is equipped with AR quality inspection glasses and a mobile quality inspection APP.

[0010] Furthermore, the distributed intelligent sensor network meets the following deployment requirements:

[0011] (a) The sensor array includes at least eight types of sensors among stress sensors, displacement sensors, temperature and humidity sensors, and vibration sensors;

[0012] (b) The vertical sensor spacing is ≤50 cm, and the horizontal direction adopts a hexagonal tessellation topology;

[0013] (c) The drone inspection unit is equipped with an optical camera with a resolution of ≥40 million pixels and an infrared thermal imager with an accuracy of ±0.1°C.

[0014] Furthermore, the multimodal data fusion module performs the following processing flow:

[0015] (a) Establishing a spatiotemporal coordinate system mapping relationship to align sensor data with BIM model coordinates;

[0016] (b) Using an improved Kalman filter algorithm for noise suppression, the signal-to-noise ratio is improved by ≥15dB;

[0017] (c) Dynamically calculate the weight of each data source through the entropy weight method to generate a comprehensive quality score Q_score:

[0018] The weight factor wi is dynamically adjusted according to the reliability of historical data, and αcorr is the environmental correction coefficient.

[0019] Furthermore, the digital twin visualization platform includes:

[0020] (a) Quality assessment engine: This runs a convolutional neural network model based on transfer learning. The input layer receives multi-source data, and the output layer generates a probability distribution of quality defects.

[0021] (b) Defect tracing module: Constructs a Bayesian network to analyze the causes of defects and supports at least five levels of causal chain tracing;

[0022] (c) 3D visualization interface: displays quality heat maps using color gradients and supports real-time defect annotation in AR mode.

[0023] Furthermore, the edge computing node has:

[0024] (a) Localized data processing capabilities, running the compressed YOLOv5s model for real-time image analysis;

[0025] (b) Dynamic bandwidth allocation mechanism, automatically switching to LoRa communication protocol when network delay > 100ms;

[0026] (c) Security protection module, which uses the SM4 national encryption algorithm to encrypt the transmitted data.

[0027] Furthermore, the cloud server cluster is configured with:

[0028] (a) BIM quality database, which stores component attributes and inspection records in accordance with ISO 19650 standards;

[0029] (b) LSTM strength prediction model, which inputs concrete curing environment parameters and outputs the predicted value of 28-day compressive strength;

[0030] (c) Knowledge graph engine, which builds a database of associations between construction specifications, material standards, and historical cases.

[0031] Furthermore, the user interaction terminal implements the following functions:

[0032] (a) AR quality inspection glasses: superimposed display of component design parameters and deviation values from measured data;

[0033] (b) Mobile quality inspection app: supports voice input of defect descriptions and automatically generates rectification notices;

[0034] (c) Electronic signature module: uses blockchain technology to store tamper-proof acceptance records.

[0035] A construction quality inspection and management method, characterized by comprising the following steps:

[0036] S1. Mark key quality inspection points in the BIM model and generate a sensor deployment plan;

[0037] S2. Real-time acquisition of multi-source detection data through a hybrid communication network;

[0038] S3. Perform data preprocessing and preliminary quality determination at the edge node;

[0039] S4. Multimodal data fusion and in-depth analysis in the cloud;

[0040] S5. Generate a visual quality report and push it to relevant terminals;

[0041] S6. Update the quality status data in the building digital twin model.

[0042] Furthermore, the edge node processing in step S3 specifically includes:

[0043] (a) Performing improved non-local means denoising on image data;

[0044] (b) Using the lightweight MobileNetV3 model for preliminary crack identification;

[0045] (c) When it is detected that the key indicators exceed the threshold, an early warning signal is triggered immediately.

[0046] Furthermore, the depth analysis in step S4 includes the following algorithms:

[0047] (a) Rebar spacing detection algorithm: The aerial image is segmented using the U-Net model and the standard deviation of the point cloud spacing is calculated;

[0048] (b) Concrete vibration quality assessment algorithm: Analyze the spectrum characteristics of the vibration sensor and identify the under-vibration area;

[0049] (c) Construct safety warning algorithm: predict potential risk level based on finite element simulation results.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. Traditional manual inspections have a high misjudgment rate for defects such as cracks and hollows. However, this system utilizes a distributed sensor array (such as infrared thermal imagers and laser rangefinders) combined with an improved Kalman filter algorithm to achieve spatiotemporal alignment of multimodal data and noise suppression, reducing the overall detection error rate. Furthermore, the improved YOLOv5s algorithm automatically identifies very small cracks, variations in the thickness of rebar cover, and other defects that are difficult to detect manually. Especially during nighttime or high-altitude operations, the system significantly improves defect detection rates in low-light environments by fusing infrared thermal imaging data with visible light images.

[0052] 2. This system can achieve full coverage detection of floors in a short time through drone cluster inspection and edge computing real-time analysis.

[0053] 3. Real-time synchronization of on-site sensor data and BIM models through the 5G network to build a digital twin of the entire "design-construction-operation and maintenance" chain.

[0054] 4. When quality problems are discovered, the system uses Bayesian networks to analyze 12 influencing factors, including materials, processes, and environment, which can be traced back to specific construction teams, equipment numbers, and even meteorological data.

[0055] 5. This system can reduce personnel requirements and lower work costs through intelligent detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of the system of the present invention. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] Example 1: Please refer to Figure 1 ,The present invention provides a technical solution: a building quality inspection and management system, comprising: a distributed intelligent sensor network, including a multi-type sensor array, a mobile detection terminal and a drone inspection unit deployed at key nodes of the building structure;

[0059] Multimodal data fusion module, used for spatiotemporal alignment and feature extraction of sensor data, image data, and BIM model data;

[0060] A digital twin visualization platform, including a quality assessment engine, a defect tracing module 103b and a three-dimensional visualization interface;

[0061] Edge computing nodes, deployed at construction sites and running lightweight quality analysis models;

[0062] A cloud server cluster that stores a database on building quality throughout its lifecycle and runs deep learning algorithms;

[0063] The user interaction terminal is equipped with AR quality inspection glasses and a mobile quality inspection APP.

[0064] 2. A building quality inspection management system according to claim 1, characterized in that:

[0065] The distributed intelligent sensor network meets the following deployment requirements:

[0066] (a) The sensor array includes at least eight types of sensors among stress sensors, displacement sensors, temperature and humidity sensors, and vibration sensors;

[0067] (b) The vertical sensor spacing is ≤50 cm, and the horizontal direction adopts a hexagonal tessellation topology;

[0068] (c) The drone inspection unit 101c is equipped with an optical camera with a resolution of ≥40 million pixels and an infrared thermal imager with an accuracy of ±0.1°C.

[0069] The multimodal data fusion module performs the following processing flow:

[0070] (a) Establishing a spatiotemporal coordinate system mapping relationship to align sensor data with BIM model coordinates;

[0071] (b) Using an improved Kalman filter algorithm for noise suppression, the signal-to-noise ratio is improved by ≥15dB;

[0072] (c) Dynamically calculate the weight of each data source through the entropy weight method to generate a comprehensive quality score Q_score:

[0073] The weight factor wi is dynamically adjusted according to the reliability of historical data, and αcorr is the environmental correction coefficient.

[0074] The digital twin visualization platform includes:

[0075] (a) Quality assessment engine: This runs a convolutional neural network model based on transfer learning. The input layer receives multi-source data, and the output layer generates a probability distribution of quality defects.

[0076] (b) Defect tracing module: Constructs a Bayesian network to analyze the causes of defects and supports at least five levels of causal chain tracing;

[0077] (c) 3D visualization interface: displays quality heat maps using color gradients and supports real-time defect annotation in AR mode.

[0078] The edge computing node has:

[0079] (a) Localized data processing capabilities, running the compressed YOLOv5s model for real-time image analysis;

[0080] (b) Dynamic bandwidth allocation mechanism, automatically switching to LoRa communication protocol when network delay > 100ms;

[0081] (c) Security protection module, which uses the SM4 national encryption algorithm to encrypt the transmitted data.

[0082] The cloud server cluster is configured with:

[0083] (a) BIM quality database, which stores component attributes and inspection records in accordance with ISO 19650 standards;

[0084] (b) LSTM strength prediction model, which inputs concrete curing environment parameters and outputs the predicted value of 28-day compressive strength;

[0085] (c) Knowledge graph engine, which builds a database of associations between construction specifications, material standards, and historical cases.

[0086] The user interaction terminal implements the following functions:

[0087] (a) AR quality inspection glasses: superimposed display of component design parameters and deviation values from measured data;

[0088] (b) Mobile quality inspection app: supports voice input of defect descriptions and automatically generates rectification notices;

[0089] (c) Electronic signature module: uses blockchain technology to store tamper-proof acceptance records.

[0090] A construction quality inspection and management method comprises the following steps:

[0091] S1. Mark key quality inspection points in the BIM model and generate a sensor deployment plan;

[0092] S2. Real-time acquisition of multi-source detection data through a hybrid communication network;

[0093] S3. Perform data preprocessing and preliminary quality determination at the edge node;

[0094] S4. Multimodal data fusion and in-depth analysis in the cloud;

[0095] S5. Generate a visual quality report and push it to relevant terminals;

[0096] S6. Update the quality status data in the building digital twin model.

[0097] The edge node processing in step S3 specifically includes:

[0098] (a) Performing improved non-local means denoising on image data;

[0099] (b) Using the lightweight MobileNetV3 model for preliminary crack identification;

[0100] (c) When it is detected that the key indicators exceed the threshold, an early warning signal is triggered immediately.

[0101] The depth analysis in step S4 includes the following algorithms:

[0102] (a) Rebar spacing detection algorithm: The aerial image is segmented using the U-Net model and the standard deviation of the point cloud spacing is calculated;

[0103] (b) Concrete vibration quality assessment algorithm: Analyze the spectrum characteristics of the vibration sensor and identify the under-vibration area;

[0104] (c) Construct safety warning algorithm: predict potential risk level based on finite element simulation results.

[0105] Example 2:

[0106] Step 1: Drone Intelligent Aerial Photography

[0107] Equipment selected: DJI Matrice 300RTK drone, equipped with Zenmuse P1 full-frame camera (45MP) and L1 LiDAR;

[0108] Flight parameters:

[0109] Flight altitude: 8m (vertical accuracy ±1cm)

[0110] Overlap rate: heading 80% / sideways 70%

[0111] Shooting angle: straight + tilt 45° dual mode

[0112] Data collection:

[0113] 1. Visible light image resolution: 8192×5460

[0114] 2. Laser point cloud density: 500 points / ㎡

[0115] Step 2: Image Preprocessing

[0116] 1. Distortion correction: Use Brown-Conrady model to eliminate lens distortion x′=x(1+k1r 2 +k2r 4 )+2p1xy+p2(r 2 +2x 2 )

[0117] 2. Lighting balance: Use the Retinex algorithm to enhance shadow areas

[0118] 3. Feature matching: SIFT algorithm realizes multi-view image registration

[0119] Step 3: Rebar division and positioning

[0120] Improved U-Net model architecture:

[0121] Encoder: ResNet50 backbone network

[0122] Decoder: Introducing attention gating mechanism

[0123] Output layer: pixel-level classification (rebar / background)

[0124] Training data:

[0125] Dataset: 10,800 annotated images (including different lighting and occlusion scenes)

[0126] Enhancement strategy: random rotation (±15°), Gaussian noise (σ=0.01)

[0127] Performance indicators:

[0128] 1. mIoU: 98.7%

[0129] 2. Inference speed: 0.15s / frame (NVIDIA Jetson AGX Xavier)

[0130] Step 4: 3D point cloud reconstruction

[0131] 1. Dense matching: using SGM algorithm to generate disparity map

[0132] 2. Point cloud generation: Calculate 3D coordinates through forward intersection

[0133]

[0134] (where: f = focal length, B = baseline distance, d = parallax)

[0135] 3. Point cloud optimization:

[0136] Outlier removal: statistical filtering (neighborhood 50 points, standard deviation threshold 1.5)

[0137] Smoothing: Moving Least Squares (MLS) surface fitting

[0138] Step 5: Spacing calculation and deviation analysis

[0139] 1. Rebar axis extraction: Slice the point cloud (thickness 2mm) and fit the straight line equation using the RANSAC algorithm;

[0140] 2. Spacing calculation:

[0141] Main reinforcement spacing: minimum Euclidean distance between adjacent axes

[0142] Distribution uniformity: the ratio of standard deviation σ to mean μ (CV = σ / μ)

[0143] 3. Standard comparison:

[0144] Design value: D (e.g. 200mm)

[0145] Allowable deviation: ±10mm (GB50204-2015)

[0146] Automatically mark extreme points (red warning)

[0147] Step 6: Output

[0148] 1. Visual Report:

[0149] 3D point cloud model annotation deviation position generates PDF format detection table (including each steel bar coordinates and spacing value) 2. Data synchronization:

[0150] Update reinforcement properties in the BIM model

[0151] Out-of-tolerance areas automatically trigger rectification work orders.

[0152] Effect verification

[0153] A comparative test was conducted in a high-rise building project:

[0154]

[0155] Example 3:

[0156] Step 1: Smart sensor network deployment

[0157] Device configuration:

[0158] Vibration sensor array: piezoelectric accelerometer (range ±50g, frequency response 5-1000Hz), spacing ≤1.5m

[0159] Temperature monitoring unit:

[0160] Pre-buried optical fiber temperature measurement (accuracy ±0.5°C, 3 per cubic meter)

[0161] Infrared thermal imager (FLIR A8580, frame rate 30Hz)

[0162] Environmental monitoring station: real-time collection of wind speed, humidity, and sunshine intensity

[0163] Step 2: Data collection during pouring process

[0164] Vibration quality monitoring:

[0165] The vibration sensor records the vibrator operation characteristics at a sampling rate of 1000Hz

[0166] Feature extraction: Calculate the effective vibration time ratio (>80% is qualified)

[0167] Temperature field construction:

[0168] Fiber optic temperature measurement data is updated every 30 seconds

[0169] Infrared thermal imager scans surface temperature distribution (trigger warning when temperature difference > 20°C)

[0170] Step 3: Real-time processing at the edge

[0171] Warning rules:

[0172] Local temperature gradient > 3°C / min → trigger cooling measures

[0173] Vibration coverage < 85% → Automatically mark the vibration area

[0174] Step 4: Deep analysis in the cloud

[0175] 1. Strength prediction model:

[0176] Input parameters: temperature history, vibration energy, environmental conditions

[0177] LSTM network structure: 3 layers, 128 units, Dropout = 0.2

[0178] Output: 28-day compressive strength prediction value (error ≤ 3MPa);

[0179] Crack risk warning: Calculate equivalent tensile stress based on temperature stress theory,

[0180] σ T =E·α·△T / (1-v), an alarm is issued when σ_T>0.7f_t (tensile strength).

[0181] Step 5: Dynamic Construction Adjustment

[0182] Automatic control strategy:

[0183] Insufficient vibration area → Navigation vibrator to accurately supplement vibration

[0184] High temperature area → Start the spray cooling system (flow rate ≥ 5L / min·㎡)

[0185] Data traceability: The blockchain records the pouring parameters and quality inspection results of each truck of concrete.

[0186] The effect was verified by comparative testing in a super high-rise core tube casting project in Hangzhou:

[0187]

[0188] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A building quality inspection management system, characterized in that: include: Distributed intelligent sensor networks, including multi-type sensor arrays, mobile detection terminals, and drone inspection units deployed at key nodes of building structures; Multimodal data fusion module, used for spatiotemporal alignment and feature extraction of sensor data, image data, and BIM model data; Digital twin visualization platform, including a quality assessment engine, defect tracing module, and 3D visualization interface; Edge computing nodes, deployed at construction sites and running lightweight quality analysis models; A cloud server cluster that stores a database on building quality throughout its lifecycle and runs deep learning algorithms; The user interaction terminal is equipped with AR quality inspection glasses and a mobile quality inspection APP.

2. A construction quality inspection management system according to claim 1, characterized in that: The distributed intelligent sensor network meets the following deployment requirements: (a) The sensor array includes at least eight types of sensors among stress sensors, displacement sensors, temperature and humidity sensors, and vibration sensors; (b) The vertical sensor spacing is ≤50 cm, and the horizontal direction adopts a hexagonal tessellation topology; (c) The drone inspection unit is equipped with an optical camera with a resolution of ≥40 million pixels and an infrared thermal imager with an accuracy of ±0.1°C.

3. A construction quality inspection management system according to claim 1, characterized in that: The multimodal data fusion module performs the following processing flow: (a) Establishing a spatiotemporal coordinate system mapping relationship to align sensor data with BIM model coordinates; (b) Using an improved Kalman filter algorithm for noise suppression, the signal-to-noise ratio is improved by ≥15dB; (c) Dynamically calculate the weight of each data source through the entropy weight method to generate a comprehensive quality score Q_score: The weight factor wi is dynamically adjusted according to the reliability of historical data, and αcorr is the environmental correction coefficient.

4. A construction quality inspection management system according to claim 1, characterized in that: The digital twin visualization platform includes: (a) Quality assessment engine: This runs a convolutional neural network model based on transfer learning. The input layer receives multi-source data, and the output layer generates a probability distribution of quality defects. (b) Defect tracing module: Constructs a Bayesian network to analyze the causes of defects and supports at least five levels of causal chain tracing; (c) 3D visualization interface: displays quality heat maps using color gradients and supports real-time defect annotation in AR mode.

5. A construction quality inspection management system according to claim 1, characterized in that: The edge computing node has: (a) Localized data processing capabilities, running the compressed YOLOv5s model for real-time image analysis; (b) Dynamic bandwidth allocation mechanism, automatically switching to LoRa communication protocol when network delay > 100ms; (c) Security protection module, which uses the SM4 national encryption algorithm to encrypt the transmitted data.

6. A construction quality inspection management system according to claim 1, characterized in that: The cloud server cluster is configured with: (a) BIM quality database, which stores component attributes and inspection records in accordance with ISO 19650 standards; (b) LSTM strength prediction model, which inputs concrete curing environment parameters and outputs the predicted value of 28-day compressive strength; (c) Knowledge graph engine, which builds a database of associations between construction specifications, material standards, and historical cases.

7. A construction quality inspection management system according to claim 1, characterized in that: The user interaction terminal implements the following functions: (a) AR quality inspection glasses: superimposed display of component design parameters and deviation values from measured data; (b) Mobile quality inspection app: supports voice input of defect descriptions and automatically generates rectification notices; (c) Electronic signature module: uses blockchain technology to store tamper-proof acceptance records.

8. A construction quality inspection and management method based on the system of any one of claims 1 to 7, characterized in that The following steps are involved: S1. Mark key quality inspection points in the BIM model and generate a sensor deployment plan; S2. Real-time acquisition of multi-source detection data through a hybrid communication network; S3. Perform data preprocessing and preliminary quality determination at the edge node; S4. Multimodal data fusion and in-depth analysis in the cloud; S5. Generate a visual quality report and push it to relevant terminals; S6. Update the quality status data in the building digital twin model.

9. The method according to claim 8, characterized in that: The edge node processing in step S3 specifically includes: (a) Performing improved non-local means denoising on image data; (b) Using the lightweight MobileNetV3 model for preliminary crack identification; (c) When it is detected that the key indicators exceed the threshold, an early warning signal is triggered immediately.

10. The method according to claim 8, characterized in that: The depth analysis in step S4 includes the following algorithms: (a) Rebar spacing detection algorithm: The aerial image is segmented using the U-Net model and the standard deviation of the point cloud spacing is calculated; (b) Concrete vibration quality assessment algorithm: Analyze the spectrum characteristics of the vibration sensor and identify the under-vibration area; (c) Construct safety warning algorithm: predict potential risk level based on finite element simulation results.

Images