Urban underground space structure health detection method and detection system

Abstract

The invention discloses an urban underground space structure health detection method and detection system, and the method comprises the steps: S1, collecting the multi-modal data of an underground space structure through the deployment of diversified sensors; s2, preprocessing the collected data, wherein the preprocessing comprises filtering denoising and feature extraction; s3, performing pattern recognition and health state evaluation on the processed data by using a deep learning model; s4, dynamically adjusting a threshold value of health assessment according to the environmental parameters, and predicting a future health trend; and S5, the evaluation result is displayed through a visual platform, a corresponding health examination report is generated, the health condition of the underground space structure can be monitored in real time by integrating multiple sensors and utilizing a data processing technology, potential problems can be found in time, an early warning threshold value can be dynamically adjusted according to real-time monitoring data through a dynamic threshold value adjusting mechanism, and the early warning efficiency is improved. The problems of false alarm and missing alarm caused by a fixed threshold value are avoided, and the accuracy and reliability of early warning are improved. The method is suitable for various underground space structures.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring of underground space structures, and in particular to a health detection method and a detection system for urban underground space structures. Background Art

[0002] With the acceleration of urbanization, underground space structures play an increasingly important role in modern urban infrastructure. However, underground space structures face complex geological conditions and long-term operational pressure, which are prone to structural health problems. Traditional health monitoring methods mainly rely on manual inspections, which have problems such as low efficiency, low accuracy and high cost. In recent years, with the development of sensor technology, signal processing technology and machine learning, automated and intelligent health monitoring methods have gradually become a research hotspot.

[0003] After searching, the invention patent with Chinese patent number CN118566346A discloses a safety monitoring method and system for highway tunnel structures, wherein the method includes: setting a vibration sensing module at a preset position in the tunnel, and using the vibration sensing module to collect vibration data generated in the tunnel; extracting features from the vibration data to obtain target feature parameters related to the health of the tunnel structure; based on the target feature parameters, using a target safety assessment model to assess the health status of the tunnel and obtain a health assessment result. Compared with the prior art, the invention patent with Chinese patent number CN118566346A integrates seismic sensors into the existing highway tunnel monitoring system, which can monitor the health status of the tunnel structure in real time, further improve the ability to detect and respond to events and anomalies, and reduce the workload of manual inspections, improve efficiency, and has important application value and broad market prospects.

[0004] However, in actual use, the above invention only uses basic methods (such as denoising and filtering) for data preprocessing, which fails to effectively improve data quality; feature extraction mainly uses FFT and wavelet transform, and fails to make full use of more advanced feature extraction methods; the security assessment model uses support vector machines and neural networks, but fails to achieve higher accuracy and generalization capabilities; the early warning mechanism does not specify the dynamic threshold adjustment, resulting in low accuracy and effectiveness of the early warning.

[0005] Therefore, a method and system for detecting the health of urban underground space structures are proposed. Summary of the invention

[0006] The purpose of the present invention is to solve the problems existing in the prior art that data preprocessing only uses basic methods (such as denoising and filtering) and fails to effectively improve data quality; feature extraction mainly uses FFT and wavelet transform and fails to make full use of more advanced feature extraction methods; the safety assessment model uses support vector machines and neural networks, but fails to achieve higher accuracy and generalization ability; the early warning mechanism does not explain in detail the dynamic threshold adjustment, resulting in low accuracy and effectiveness of the early warning, and proposes a method and system for detecting the health of urban underground space structures.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for detecting the health of an urban underground space structure comprises the following steps:

[0009] S1: Collect multimodal data of underground space structure by deploying diverse sensors;

[0010] S2: preprocessing the collected data, including filtering, denoising and feature extraction;

[0011] S3: Use deep learning models to perform pattern recognition and health status assessment on processed data;

[0012] S4: Dynamically adjust the threshold of health assessment according to environmental parameters and predict future health trends;

[0013] S5: The evaluation results are displayed through a visualization platform and a corresponding health check report is generated.

[0014] The above technical solution further includes:

[0015] Preferably, S2 includes:

[0016] Use wavelet transform to perform multi-scale signal processing, including multi-resolution analysis of the collected vibration signals to extract useful information in different frequency bands;

[0017] Adopt Kalman filter for adaptive filtering, including dynamically adjusting filtering parameters according to signal characteristics to remove noise interference;

[0018] Use PCA dimensionality reduction technology to extract features, including extracting the most representative features from high-dimensional data to simplify subsequent processing steps;

[0019] LASSO regression was used for feature selection, including screening out the most representative features through regularization methods.

[0020] Preferably, S3 includes:

[0021] Use convolutional neural network (CNN) for image recognition, including feature extraction and classification of collected image data to identify cracks;

[0022] Use long short-term memory networks (LSTMs) for time series analysis, including processing the collected time series data to identify structural health change trends;

[0023] Use ensemble learning models to improve prediction accuracy, including combining multiple models for ensemble learning to improve prediction accuracy.

[0024] Preferably, S4 includes:

[0025] Dynamically adjusting thresholds for health assessments based on environmental parameters, including dynamically adjusting criteria for health assessments based on environmental parameters;

[0026] Use ARIMA models for time series forecasting, including using the autoregressive integrated moving average model ARIMA to predict the health of the structure over a period of time in the future;

[0027] Survival analysis techniques are used to predict the remaining life of structures, including estimating the remaining life of structures through survival analysis techniques, providing a basis for maintenance decisions.

[0028] A detection system for a method for detecting the health of an urban underground space structure, comprising:

[0029] A data acquisition module, used to collect multimodal data by deploying diverse sensors;

[0030] Data processing module, used for preprocessing the collected data, including filtering, denoising and feature extraction;

[0031] A deep learning module for pattern recognition and health status assessment of processed data using deep learning models;

[0032] An adaptive assessment module, which is used to dynamically adjust the threshold of health assessment according to environmental parameters and predict future health trends;

[0033] Visual management platform, used to display assessment results and generate corresponding health check reports.

[0034] Preferably, the data acquisition module includes:

[0035] Vibration sensors to monitor structural vibrations;

[0036] Temperature sensor, used to monitor ambient temperature changes;

[0037] Humidity sensor, used to monitor changes in ambient humidity;

[0038] Fiber Bragg grating (FBG) sensors are used to monitor stress changes inside structures;

[0039] LiDAR, for non-contact monitoring of structural surface deformation;

[0040] Infrared sensors to monitor the temperature distribution on the surface of the structure;

[0041] Groundwater quality monitoring sensors to assess corrosion risk.

[0042] Preferably, the data processing module includes:

[0043] Multi-scale signal processing unit, used to perform multi-resolution analysis on the collected vibration signals and extract useful information in different frequency bands;

[0044] Adaptive filtering unit, used to dynamically adjust filtering parameters according to signal characteristics and remove noise interference;

[0045] Feature extraction unit, used to extract the most representative features from high-dimensional data and simplify subsequent processing steps;

[0046] The feature selection unit is used to select the most representative features through regularization methods.

[0047] Preferably, the deep learning module includes:

[0048] An image recognition unit, used to extract features and classify the collected image data and identify cracks;

[0049] The time series analysis unit is used to process the collected time series data and identify the trend of structural health changes;

[0050] The integrated learning unit is used to combine multiple models for integrated learning to improve prediction accuracy.

[0051] Preferably, the adaptive evaluation module comprises:

[0052] A dynamic threshold adjustment unit, used to dynamically adjust the health assessment criteria according to environmental parameters;

[0053] The time series forecasting unit is used to predict the health status of the structure in the future using the autoregressive integrated moving average model ARIMA;

[0054] The remaining life prediction unit is used to estimate the remaining life of the structure through survival analysis technology to provide a basis for maintenance decisions.

[0055] Preferably, the visual management platform includes:

[0056] A three-dimensional visualization unit, which is used to present the structural layout and health status of underground spaces through 3D modeling technology;

[0057] Immersive management unit, used to provide immersive management experience through virtual reality VR or augmented reality AR technology;

[0058] Interactive interface unit, used to design user interfaces, supports drag and drop, zoom interaction, and supports remote access to the system;

[0059] The report generation unit is used to automatically generate regular health check reports and support exporting to PDF format.

[0060] The present invention has the following beneficial effects:

[0061] 1. In the present invention, by integrating multiple sensors and using data processing technology, the health status of underground space structures can be monitored in real time and potential problems can be discovered in time. Data preprocessing technology can significantly improve the reliability and accuracy of data. By reducing the workload of manual inspection, the efficiency and automation level of monitoring work are improved, and the labor cost is reduced. By using feature extraction methods, the accuracy and robustness of monitoring results are improved, and the nonlinear characteristics of signals can be better captured, thereby improving the accuracy of feature extraction.

[0062] 2. In the present invention, a safety assessment model is adopted to improve the generalization ability of the model, so that it can maintain high accuracy and stability in different underground space structures. By combining multiple weak learners, overfitting can be effectively reduced and the stability and generalization ability of the model can be improved. The dynamic threshold adjustment mechanism improves the accuracy and effectiveness of the early warning, and can promptly remind maintenance personnel to pay attention to potential structural health problems. The dynamic threshold adjustment mechanism can dynamically adjust the early warning threshold according to real-time monitoring data, avoiding the false alarm and missed alarm problems caused by fixed thresholds, and improving the accuracy and reliability of the early warning. It is applicable to a variety of underground space structures and has a wide range of applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A flow chart of a method for detecting the health of an urban underground space structure proposed by the present invention;

[0064] Figure 2 This is a system architecture diagram of a detection system for a method for detecting the health of an urban underground space structure in the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0066] like Figure 1 As shown, a method for detecting the health of an urban underground space structure proposed by the present invention comprises the following steps:

[0067] S1: Collect multimodal data of underground space structure by deploying diverse sensors;

[0068] S2: preprocessing the collected data, including filtering, denoising and feature extraction;

[0069] S3: Use deep learning models to perform pattern recognition and health status assessment on processed data;

[0070] S4: Dynamically adjust the threshold of health assessment according to environmental parameters and predict future health trends;

[0071] S5: The evaluation results are displayed through a visualization platform and a corresponding health check report is generated.

[0072] A detection system for a method for detecting the health of an urban underground space structure, comprising:

[0073] A data acquisition module, used to collect multimodal data by deploying diverse sensors;

[0074] Data processing module, used for preprocessing the collected data, including filtering, denoising and feature extraction;

[0075] A deep learning module for pattern recognition and health status assessment of processed data using deep learning models;

[0076] An adaptive assessment module, which is used to dynamically adjust the threshold of health assessment according to environmental parameters and predict future health trends;

[0077] Visual management platform, used to display assessment results and generate corresponding health check reports.

[0078] In one embodiment, the data acquisition module is implemented

[0079] Diversified sensor deployment

[0080] Vibration Sensor

[0081] Type: Adopts high-sensitivity accelerometer sensor.

[0082] Location: Install vibration sensors at key locations of the structure (such as beams, columns, walls, etc.).

[0083] Purpose: To monitor the vibration of structures and identify potential structural defects.

[0084] Implementation: Choose a MEMS (micro-electromechanical system) accelerometer sensor, such as the ADXL345 or BMA220, which have high sensitivity and low power consumption.

[0085] Connection method: Connect to the data acquisition unit via SPI or I2C interface.

[0086] Temperature Sensor

[0087] Type: Use digital temperature sensor.

[0088] Location: Install temperature sensors at various locations on the structure.

[0089] Purpose: To monitor ambient temperature changes and assess the impact of the environment on the structure.

[0090] Implementation: Choose a digital temperature sensor such as DS18B20 or TMP36, which can directly output temperature data.

[0091] Connection method: Connect to the data acquisition unit via single-line or multi-line communication protocol.

[0092] Humidity Sensor

[0093] Type: Adopt high precision humidity sensor.

[0094] Location: Install moisture sensors in key areas of underground spaces.

[0095] Application: To monitor changes in ambient humidity and assess corrosion risks.

[0096] Implementation: Choose a high-precision humidity sensor like DHT11 or DHT22, which can measure both temperature and humidity.

[0097] Connection method: connected to the data acquisition unit via a single-line serial communication interface.

[0098] Fiber Bragg Grating (FBG) Sensors

[0099] Type: Uses Fiber Bragg Grating (FBG) sensor.

[0100] Location: Embed the FBG sensor inside the structure.

[0101] Purpose: To monitor stress changes within the structure and assess its internal health status.

[0102] Implementation: Use FBG sensors and read the changes in FBG reflection wavelength through a spectrum analyzer to infer the stress situation.

[0103] Connection method: connected to the data acquisition unit via optical fiber communication.

[0104] LiDAR

[0105] Type: Uses high-resolution lidar.

[0106] Location: Mount the lidar on the exterior of the structure.

[0107] Application: Non-contact monitoring to detect deformation of the surface of the structure.

[0108] Implementation: Choose a LiDAR (Light Detection and Ranging) sensor, such as the Velodyne VLP-16 or HDL-32E.

[0109] Connection method: Connect to the data acquisition unit via the Ethernet interface.

[0110] Infrared sensor

[0111] Type: Uses infrared thermal imaging sensor.

[0112] Location: Mount the infrared sensor on the exterior of the structure.

[0113] Purpose: To monitor the temperature distribution on the surface of a structure and assess the health of the structure.

[0114] Implementation: Choose a FLIR series thermal imaging camera, such as the FLIR T640 or A655.

[0115] Connection method: Connect to the data acquisition unit via USB or Ethernet interface.

[0116] Groundwater quality monitoring sensors

[0117] Type: Uses electrochemical sensor.

[0118] Location: Install water quality monitoring sensors in groundwater pipes or underground water tanks.

[0119] Purpose: To assess the corrosion risk of groundwater to structures.

[0120] Implementation: Choose electrochemical pH sensors and conductivity sensors, such as the series sensors from HACH.

[0121] Connection method: Connect to the data acquisition unit via RS-485 interface.

[0122] Intelligent Perception Network

[0123] Self-organizing network

[0124] Implementation: Sensor nodes form a self-organizing network through wireless communication protocols (such as Zigbee and LoRa).

[0125] Function: Sensor nodes can automatically discover neighboring nodes and form a network topology.

[0126] Detailed implementation: Use CC2530 chip as the core of Zigbee module and establish network connection through Zigbee protocol stack.

[0127] Configuration: Each node is configured as a router or terminal device. The router is responsible for forwarding data packets, and the terminal device is responsible for collecting data.

[0128] Self-diagnosis function

[0129] Implementation: Each sensor node has a built-in self-diagnosis module and performs self-checks regularly.

[0130] Function: When a fault is detected, the node can automatically adjust or send out an alarm signal.

[0131] Detailed implementation: Integrate self-diagnostic software in the sensor node and perform hardware self-check procedures regularly.

[0132] Configuration: When the self-test fails, the node sends an alarm message to the central server via the wireless network.

[0133] Energy Management

[0134] Implementation: Sensor nodes are equipped with solar panels or vibration energy harvesting devices.

[0135] Function: Automatically charge in sufficient light or frequent vibration to extend the life of the sensor.

[0136] Detailed implementation: Use a small solar panel (such as 5W or 10W) ​​to power the sensor and equip it with a lithium-ion battery to store electrical energy.

[0137] Configuration: Use an energy management system (such as the MAX17043) for power management to ensure that the sensor continues to work.

[0138] In one embodiment, the implementation of the data processing module

[0139] Data preprocessing

[0140] Multiscale signal processing

[0141] Implementation: Use wavelet transform to perform multi-resolution analysis on the collected vibration signal.

[0142] Function: Extract useful information from different frequency bands and improve signal resolution.

[0143] Detailed implementation:

[0144] Selection of wavelet basis function: Use Daubechies wavelet basis function (such as DB4 or Haar). These wavelet basis functions have good localization characteristics and are suitable for processing transient signals.

[0145] Multi-level decomposition: Through multi-level wavelet decomposition, the signal is decomposed into multiple sub-band signals, each sub-band contains information in a specific frequency range.

[0146] Specific steps:

[0147] Apply wavelet transform and select appropriate wavelet basis function.

[0148] Set the number of decomposition layers N, usually 3-5 layers are selected to cover a wide frequency range.

[0149] Separate the useful information of each frequency band and remove the noise components.

[0150] Tools / Libraries: Use MATLAB's Wavelet Toolbox or Python's PyWavelets library to implement wavelet transforms.

[0151] Configuration: Select a suitable wavelet basis function, such as db4, and set the number of decomposition levels N = 3.

[0152] formula:

[0153]

[0154] Among them, C j,k is the wavelet coefficient, j is the scale, k is the translation, ψ j,k (t) is the wavelet basis function.

[0155] Adaptive filtering

[0156] Implementation: Use Kalman Filter to filter the signal.

[0157] Function: Dynamically adjust filter parameters according to signal characteristics to remove noise interference.

[0158] Detailed implementation:

[0159] Extended Kalman Filter (EKF): Since real signals are usually nonlinear, an Extended Kalman Filter (EKF) is used to process nonlinear signals.

[0160] Specific steps:

[0161] Set initial state and the covariance matrix P 0|0 .

[0162] The state vector is iteratively calculated using the prediction step and the update step and the covariance matrix P k|k .

[0163] The observation matrix H and state transfer matrix F are adjusted according to the signal characteristics.

[0164] Tools / Libraries: Use MATLAB's Control System Toolbox or the FilterPy library in Python to implement the Kalman filter.

[0165] Configuration: Adjust the observation matrix H and state transfer matrix F according to signal characteristics.

[0166] formula:

[0167]

[0168] P k|k-1 =FP k-1|k-1 F T +Q

[0169] K k =P k|k-1 H T (HP k|k-1 H T +R) -1

[0170]

[0171] P k|k =(IK k H)P k|k-1

[0172] Feature extraction

[0173] Implementation: Use principal component analysis (PCA) to reduce the dimensionality of high-dimensional data.

[0174] Function: Extract the most representative features from high-dimensional data and simplify subsequent processing steps.

[0175] Detailed implementation:

[0176] PCA algorithm: The PCA algorithm is used to extract principal components and reduce data dimensions.

[0177] Specific steps:

[0178] Compute the covariance matrix C of the data.

[0179] Calculate the eigenvalues ​​λ of the covariance matrix i and the eigenvector v i .

[0180] Select the eigenvectors corresponding to the first k largest eigenvalues ​​as the new feature space.

[0181] Project the original data X into the new feature space.

[0182] Tools / Libraries: Use Statistics and Machine Learning Toolbox in MATLAB or Scikit-Learn library in Python to implement PCA.

[0183] Configuration: Select an appropriate number of principal components k, usually selecting principal components with a cumulative contribution rate of more than 80%.

[0184] formula:

[0185]

[0186] Cv i =λ i v i

[0187] Z=XV k

[0188] Where N is the number of samples, V k is the matrix consisting of the first k eigenvectors.

[0189] Feature Selection

[0190] Implementation: Use LASSO regression (Least Absolute Shrinkage and Selection Operator) for feature selection.

[0191] Function: Filter out the most representative features through regularization methods.

[0192] Detailed implementation:

[0193] LASSO regression model: Use LASSO regression model to select features.

[0194] Specific steps:

[0195] Construct a LASSO regression model and set the regularization parameter λ.

[0196] The optimal lambda value was selected using cross validation.

[0197] The LASSO regression model was used for feature selection.

[0198] Tools / Libraries: Use Statistics and Machine Learning Toolbox in MATLAB or Scikit-Learn library in Python to implement LASSO regression.

[0199] Configuration: Choose an appropriate regularization parameter λ, usually by cross-validation to determine the optimal λ value.

[0200] formula:

[0201]

[0202] In one embodiment, the implementation of the deep learning model

[0203] Pattern Recognition

[0204] Image Recognition

[0205] Implementation: Use Convolutional Neural Network (CNN) to extract features and classify the collected image data.

[0206] Function: Identify visual defects such as cracks.

[0207] Detailed implementation:

[0208] CNN architecture: Use a CNN model with VGG16 or ResNet50 architecture.

[0209] Specific steps:

[0210] Use a pre-trained VGG16 or ResNet50 model as a base.

[0211] Adjust the output categories of the last fully connected layer to suit the specific classification task.

[0212] Perform transfer learning to fine-tune the model using domain-specific data.

[0213] Tools / Libraries: Use MATLAB’s Deep Learning Toolbox or Keras / TensorFlow libraries in Python to implement CNNs.

[0214] Configuration: Use the ImageNet pre-trained model for transfer learning and adjust the output category of the last fully connected layer.

[0215] formula:

[0216] y=softmax(W f z+b f )

[0217] Among them, y is the output category, W f and b f are the weights and biases of the fully connected layer, and z is the input feature.

[0218] Time Series Analysis

[0219] Implementation: Use the Long Short-Term Memory (LSTM) network to process the collected time series data (such as vibration signals).

[0220] Function: Identify structural health change trends.

[0221] Detailed implementation:

[0222] LSTM model: Use the LSTM model to process time series data.

[0223] Specific steps:

[0224] Build an LSTM model and set the input layer, hidden layer, and output layer.

[0225] Set the number of hidden layer units n h , learning rate α and other hyperparameters.

[0226] Use time series data for training and validation.

[0227] Tools / Libraries: Use MATLAB’s Deep Learning Toolbox or the Keras / TensorFlow libraries in Python to implement LSTM.

[0228] Configuration: Set the number of hidden layer units n h =128, learning rate α = 0.001, and multiple training and validation are performed.

[0229] formula:

[0230] f t =σ(W f [h t-1 ,x t ]+b f )

[0231] i t =σ(W i [h t-1 ,x t ]+b i )

[0232] C t =f t ·C t-1 +i t tanh(W C [ht-1 ,x t ]+b C )

[0233] o t =σ(W o [h t-1 ,x t ]+b o )

[0234] h t =o t tanh(C t )

[0235] Ensemble Learning

[0236] Implementation: Use ensemble learning models (such as random forest, gradient boosted tree, etc.) for ensemble learning.

[0237] Function: Improve forecast accuracy.

[0238] Detailed implementation:

[0239] Ensemble learning model: Use Random Forest or Gradient Boosting Tree model.

[0240] Specific steps:

[0241] Build a random forest or gradient boosted tree model.

[0242] Adjust the number of trees n t , maximum depth d and other hyperparameters.

[0243] Use the data for training and validation.

[0244] Tools / Libraries: Use Statistics and Machine Learning Toolbox in MATLAB or Scikit-Learn in Python to implement ensemble learning models.

[0245] Configuration: Adjust the number of trees n t =100, maximum depth d = 10, and multiple training and validation are performed.

[0246] formula:

[0247]

[0248] Where y is the output, γ t is the weight of the tree, f t is the prediction function of the tth tree, θ m are model parameters.

[0249] In one embodiment, the implementation of the adaptive assessment module

[0250] Dynamic threshold adjustment

[0251] Implementation: Dynamically adjust health assessment criteria based on environmental parameters (such as temperature, humidity, etc.).

[0252] Function: Improve the accuracy and reliability of assessments.

[0253] Detailed implementation:

[0254] Linear regression model: Use a linear regression model to adjust the threshold based on environmental parameters.

[0255] Specific steps:

[0256] Collect historical data, including environmental parameters and health status.

[0257] Linear regression models were used to fit the relationships between environmental parameters and health status.

[0258] The threshold is adjusted dynamically based on the fitted relationship.

[0259] Tools / Libraries: Use Statistics and Machine Learning Toolbox in MATLAB or Scikit-Learn library in Python to implement linear regression models.

[0260] Configuration: Collect at least one year of historical data and use 80% of the data for training and 20% for validation.

[0261] formula:

[0262] y=β0+β1x1+β2x2+…+β p x p +∈

[0263] Among them, y is the health status, x i is the environmental parameter, β i is the regression coefficient and ∈ is the error term.

[0264] Time Series Forecasting

[0265] Implementation: Use the AutoRegressive Integrated Moving Average (ARIMA) model for time series forecasting.

[0266] Function: Predict the health of the structure over a period of time in the future.

[0267] Detailed implementation:

[0268] ARIMA Model: Use ARIMA model for time series forecasting.

[0269] Specific steps:

[0270] Select appropriate p, d, and q parameters, where p represents the order of the autoregressive term, d represents the number of differences, and q represents the order of the moving average term.

[0271] Use time series data for model training and validation.

[0272] Tools / Libraries: Use MATLAB’s Econometrics Toolbox or the statsmodels library in Python to implement the ARIMA model.

[0273] Configuration: Choose p=2, d=1, q=1, and perform multiple training and validation.

[0274] formula:

[0275] Δ d Y t =φ1Δ d Y t-1 +…+φ p Δ d Y t-p +∈ t +θ1∈ t-1 +…+θ q ∈ t-q

[0276] Among them, Δ d Y t is the differenced sequence, φ i is the autoregressive coefficient, θ i is the moving average coefficient, ∈ t It's white noise.

[0277] Remaining life prediction

[0278] Implementation: Estimate the remaining life of the structure using survival analysis techniques such as the Cox proportional hazards model.

[0279] Function: Provide a basis for maintenance decisions.

[0280] Detailed implementation:

[0281] Cox proportional hazards model: Survival analysis was performed using the Cox proportional hazards model.

[0282] Specific steps:

[0283] Select appropriate risk factors, such as vibration intensity, temperature, humidity, etc.

[0284] Use historical data for model training and validation.

[0285] Tools / Libraries: Use Statistics and Machine Learning Toolbox in MATLAB or the lifelines library in Python to implement the Cox proportional hazards model.

[0286] Configuration: Select appropriate risk factors and use at least one year of historical data for training and validation.

[0287] formula:

[0288] h(t|X)=h0(t)exp(β1X1+β2X2+…+β p X p )

[0289] Among them, h(t|X) is the conditional risk function, h0(t) is the baseline risk function, and X i is the risk factor, β i is the regression coefficient.

[0290] In one embodiment, the implementation of the visualization management platform

[0291] 3D Visualization

[0292] Implementation: Use 3D modeling technology (such as Three.js, Blender, etc.) to show the structural layout and health status of the underground space.

[0293] Function: Visually display the health status of the structure.

[0294] Detailed implementation: Use the Three.js library for Web-based 3D modeling.

[0295] Configuration: Import structural CAD drawings, generate 3D models, and add interactive features.

[0296] Immersive Management

[0297] Implementation: Use virtual reality (VR) or augmented reality (AR) technology to provide an immersive management experience.

[0298] Function: Enhance user experience and improve management efficiency.

[0299] Detailed implementation: Use Unity3D or Unreal Engine to develop VR / AR applications.

[0300] Configuration: Integrate VR head-mounted display devices such as HTC Vive or Oculus Rift to provide an immersive experience.

[0301] Interactive interface

[0302] Implementation: Design a user-friendly interface that supports interactive operations such as dragging, zooming, and remote access to the system.

[0303] Function: convenient for user operation and supports remote monitoring.

[0304] Detailed implementation: Develop web applications using front-end frameworks such as React or Vue.js.

[0305] Configuration: Design a simple and clear operation interface, supporting access from multiple browsers and mobile devices.

[0306] Report Generation

[0307] Implementation: Automatically generate regular health check reports and support exporting to PDF and other formats.

[0308] Function: Convenient for archiving and sharing.

[0309] Detailed implementation: Generate PDF reports using the reportlab library in Python.

[0310] Configuration: Define report templates, automatically fill in data, support export and printing.

[0311] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting the health of urban underground space structures, characterized in that: The following steps are involved: S1: Collect multimodal data of underground space structure by deploying diverse sensors; S2: preprocessing the collected data, including filtering, denoising and feature extraction; S3: Use deep learning models to perform pattern recognition and health status assessment on processed data; S4: Dynamically adjust the threshold of health assessment according to environmental parameters and predict future health trends; S5: The evaluation results are displayed through a visualization platform and a corresponding health check report is generated.

2. A method for detecting the health of an urban underground space structure according to claim 1, characterized in that: The S2 includes: Use wavelet transform to perform multi-scale signal processing, including multi-resolution analysis of the collected vibration signals to extract useful information in different frequency bands; Adopt Kalman filter for adaptive filtering, including dynamically adjusting filtering parameters according to signal characteristics to remove noise interference; Use PCA dimensionality reduction technology to extract features, including extracting the most representative features from high-dimensional data to simplify subsequent processing steps; LASSO regression was used for feature selection, including screening out the most representative features through regularization methods.

3. The urban underground space structure health detection method according to claim 1, characterized in that: The S3 includes: Use convolutional neural network (CNN) for image recognition, including feature extraction and classification of collected image data to identify cracks; Use long short-term memory networks (LSTMs) for time series analysis, including processing the collected time series data to identify structural health change trends; Use ensemble learning models to improve prediction accuracy, including combining multiple models for ensemble learning to improve prediction accuracy.

4. The urban underground space structure health detection method according to claim 1, characterized in that: The S4 includes: Dynamically adjusting thresholds for health assessments based on environmental parameters, including dynamically adjusting criteria for health assessments based on environmental parameters; Use ARIMA models for time series forecasting, including using the autoregressive integrated moving average model ARIMA to predict the health of the structure over a period of time in the future; Survival analysis techniques are used to predict the remaining life of structures, including estimating the remaining life of structures through survival analysis techniques, providing a basis for maintenance decisions.

5. A detection system for a method for detecting the health of an urban underground space structure according to any one of claims 1 to 4, characterized in that: include: A data acquisition module, used to collect multimodal data by deploying diverse sensors; Data processing module, used for preprocessing the collected data, including filtering, denoising and feature extraction; A deep learning module for pattern recognition and health status assessment of processed data using a deep learning model; An adaptive assessment module, which is used to dynamically adjust the threshold of health assessment according to environmental parameters and predict future health trends; Visual management platform, used to display assessment results and generate corresponding health check reports.

6. The detection system of the urban underground space structure health detection method according to claim 5 is characterized in that: The data acquisition module comprises: Vibration sensors to monitor structural vibrations; Temperature sensor, used to monitor ambient temperature changes; Humidity sensor, used to monitor changes in ambient humidity; Fiber Bragg grating (FBG) sensors are used to monitor stress changes inside structures; LiDAR, for non-contact monitoring of structural surface deformation; Infrared sensors to monitor the temperature distribution on the surface of the structure; Groundwater quality monitoring sensors to assess corrosion risk.

7. The detection system of the urban underground space structure health detection method according to claim 5 is characterized in that: The data processing module comprises: Multi-scale signal processing unit, used to perform multi-resolution analysis on the collected vibration signals and extract useful information in different frequency bands; Adaptive filtering unit, used to dynamically adjust filtering parameters according to signal characteristics and remove noise interference; Feature extraction unit, used to extract the most representative features from high-dimensional data and simplify subsequent processing steps; The feature selection unit is used to select the most representative features through regularization methods.

8. The detection system of the urban underground space structure health detection method according to claim 5 is characterized in that: The deep learning module includes: An image recognition unit, used to extract features and classify the collected image data and identify cracks; The time series analysis unit is used to process the collected time series data and identify the trend of structural health changes; The integrated learning unit is used to combine multiple models for integrated learning to improve prediction accuracy.

9. The detection system of the urban underground space structure health detection method according to claim 5, characterized in that: The adaptive evaluation module comprises: A dynamic threshold adjustment unit, used to dynamically adjust the health assessment criteria according to environmental parameters; The time series forecasting unit is used to predict the health status of the structure in the future using the autoregressive integrated moving average model ARIMA; The remaining life prediction unit is used to estimate the remaining life of the structure through survival analysis technology to provide a basis for maintenance decisions.

10. The detection system of the urban underground space structure health detection method according to claim 5, characterized in that: The visual management platform includes: A three-dimensional visualization unit, which is used to present the structural layout and health status of underground spaces through 3D modeling technology; Immersive management unit, used to provide immersive management experience through virtual reality VR or augmented reality AR technology; Interactive interface unit, used to design user interfaces, supports drag and drop, zoom interaction, and supports remote access to the system; The report generation unit is used to automatically generate regular health check reports and support exporting to PDF format.

Images