A grouting real-time monitoring system
Patent Information
- Application Number
- CN202610998706.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-06
- Publication Date
- 2026-10-13
AI Technical Summary
1、传统传感器多为单一参数检测，需集成多个独立设备，安装复杂且成本高，缺乏高精度校准机制，尤其在温度变化（如-20℃~80℃宽温域）、高粘度浆液流动环境下，传感器输出易受噪声干扰，测量误差显著；
通过设置监测系统，高精度传感器的使用，配合合理的初始化参数设置，可确保数据采集的准确性和稳定性；定制化密度匹配颗粒的应用，使得传感器能随浆液均匀流动分布，精准记录各传感器的投放时间与位置坐标，为后续的数据处理与分析提供了可靠基础；数据处理模块运用先进的滤波算法和时间同步校准技术，能有效去除噪声干扰，统一多源数据时间戳，同时对数据进行补偿修正，显著提高了数据的质量和准确性，为后续的分析和决策提供了可靠依据；可视化模块通过构建注浆区域的三维模型，能直观展示浆液的流动和填充情况，以及区域内的压力变化，还能精确计算填充率，有助于工作人员及时掌握注浆状态，做出科学合理的决策；预测模块结合规则引擎和LSTM神经网络，能够准确预测未来的注浆参数趋势，一旦检测到异常便能及时发出警报并采取控制措施，有效避免了潜在的安全风险和质量问题。同时，将数据存入数据库并备份至云端，确保了历史数据的可追溯性，为后续的数据分析和总结提供了有力支持。
Smart Images
Figure CN122900643A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of grouting technology, and in particular to a real-time grouting monitoring system. Background Technology
[0002] In fields such as building construction, tunnel construction, and underground structure reinforcement, grouting is a key technology for controlling ground settlement and enhancing structural stability. Its core lies in filling cavities or fissures with grout to achieve dense reinforcement of the target area. However, existing technologies have the following shortcomings: 1. Traditional sensors mostly detect a single parameter, requiring the integration of multiple independent devices, which is complex and costly to install. They also lack a high-precision calibration mechanism. Especially in environments with temperature variations (such as a wide temperature range of -20℃ to 80℃) and high-viscosity slurry flow, the sensor output is easily affected by noise, resulting in significant measurement errors. 2. During the grouting process, the grout ratio (such as water-cement ratio and admixture dosage) depends on the preset parameters in the construction drawings and cannot be dynamically adjusted according to the real-time detected hole volume (such as large holes >80cm or micro holes <30cm). This results in insufficient grout fluidity and low filling efficiency when grouting large holes, material waste when grouting micro holes, poor sensor positioning accuracy, difficulty in accurately tracking the grout flow path, and rough 3D model construction, which cannot reflect the filling status in real time. The raw data contains high-frequency noise and dynamic interference, which is not effectively filtered, resulting in large fluctuations in the parameters received by the SCADA system. This affects control decisions, lacks a time-series prediction model for grouting parameters (pressure, temperature), and causes a lag in the detection of abnormal states (such as sudden pressure drops and abnormal temperature increases), which can easily lead to accidents such as grout leakage and pipe blockage, increasing construction risks.
[0003] To address the above problems, this invention proposes a real-time grouting monitoring system. Summary of the Invention
[0004] To address the existing technical problems of inaccurate parameters, delayed response, and crude decision-making, this invention proposes a real-time grouting monitoring system.
[0005] The present invention proposes a real-time grouting monitoring system, which includes the following operating steps: Step 1: Use a high-precision sensor with a stainless steel alkali-resistant encapsulation shell, initialize the sensor parameters, set the data acquisition frequency to 5Hz and adjust the wireless transmission power to 20dBm; Step 2: Run the grouting dosing management system. The grouting dosing management system includes a data acquisition module, a data processing module, a visualization module, a prediction module, and a report generation module. The data acquisition module uses sensors mounted on customized density-matching particles to be dosed into the grouting fluid. The particles flow and distribute evenly with the grout, and the system records the dosing time and location coordinates of each sensor in real time, generating electronic logs in real time. At the same time, the system reads the data from each sensor periodically according to the preset Modbus / TCP protocol. When reading the data, the program will perform a preliminary legality check to check whether the data is within a reasonable range. If abnormal data is found, it will be marked and error information will be recorded. Step 3: The data processing module performs in-depth processing on the collected data. It uses wavelet filtering or Kalman filtering algorithms from the MATLAB algorithm library to filter noise, remove high-frequency noise and dynamic noise, and make the data smoother and more accurate. It also uses the NTP protocol based on the Pandas library to perform time synchronization calibration, unify the timestamps of multi-source data, and ensure the consistency of data time. At the same time, it calls the sensor calibration model to compensate for the data, especially to correct the temperature drift problem and improve the accuracy of the data. Step 4: The visualization module is used to construct a three-dimensional model of the grouting area. Point cloud data obtained by three-dimensional laser scanning is divided into tetrahedral meshes using the Delaunay triangulation algorithm to construct a three-dimensional model of the grouting area. Based on the position data collected by the sensor, the filling state of the grout in the model is updated in real time through the mesh division algorithm, which intuitively displays the flow and filling of the grout. The Kriging interpolation method is used to spatially interpolate the discrete pressure data of the sensor to generate a continuous pressure field. Then, the pressure cloud map is generated through the contour algorithm to intuitively display the pressure changes in the area. Finally, the volume integral method is used to accurately calculate the filling rate. Step 5: The prediction module is used to predict future data. It defines thresholds through a preset rule engine and combines them with an LSTM neural network trained based on TensorFlow to predict parameter trends. Inputting the pressure and temperature sequence of the previous 10 minutes, the model predicts the data for the next 2 minutes. If an anomaly is detected, the SCADA monitoring system sends a control command to the grouting pump and triggers the on-site alarm to remind the staff. If there is no anomaly, the data is stored in the database and key data is backed up to cloud storage to ensure that historical data is traceable. Step 6: The report generation module processes the data using Python's Pandas library, then uses the Matplotlib / Seaborn library to generate visualization charts, integrates all the data and charts, and generates a PDF / Excel report.
[0006] Preferably, in step one, the sensor uses MEMS technology to integrate a three-parameter detection module for pressure, temperature, and flow rate, and is connected to an edge computing gateway via an RS485 bus. The initialization parameters include range calibration, sampling frequency, and wireless transmission power.
[0007] Preferably, in step two, the density of the customized density-matching particles is achieved by adjusting the mixing ratio of hollow glass microspheres and epoxy resin, with a density range of 1.2-2.5 g / cm³. A fluorescent tracer coating is sprayed onto the particle surface, and the position coordinates are tracked in real time using UWB positioning technology, ensuring high positioning accuracy. 10cm.
[0008] Preferably, in step three, the data processing module uses an extended Kalman filter algorithm to reduce noise in the temperature signal and establishes a temperature state equation. ,in For the rate of temperature change, For process noise .
[0009] Preferably, in step four, the sensor calibration model is invoked to compensate the data, including temperature drift correction, and the correction formula is as follows: ,in For temperature coefficient, For real-time temperature, This is a reference temperature.
[0010] Preferably, in step five, the LSTM neural network structure of the prediction module includes an input layer, a hidden layer, and an output layer. The input layer receives preprocessed grouting pressure and temperature time series data and converts the original data into a dimensional structure suitable for LSTM processing. The hidden layer uses two LSTM units and a "forget-store-output" collaborative mechanism to capture the long-term dependency features of grouting parameters. The output layer uses a linear activation function to adapt to the prediction requirements of continuous pressure and temperature values. Based on the features extracted by the hidden layer, the predicted values of grouting pressure and temperature for the next two minutes are output.
[0011] Preferably, in step five, the LSTM neural network structure of the prediction module includes an input layer, a hidden layer, and an output layer. The input layer receives preprocessed grouting pressure and temperature time series data and converts the original data into a dimensional structure suitable for LSTM processing. The hidden layer uses two LSTM units and a "forget-store-output" collaborative mechanism to capture the long-term dependency features of grouting parameters. The output layer uses a linear activation function to adapt to the prediction requirements of continuous pressure and temperature values. Based on the features extracted by the hidden layer, the predicted values of grouting pressure and temperature for the next two minutes are output.
[0012] The beneficial effects of this invention are as follows: By setting up a monitoring system and using high-precision sensors, along with reasonable initialization parameter settings, the accuracy and stability of data acquisition can be ensured. The application of customized density-matching particles allows sensors to be evenly distributed with the grout flow, accurately recording the deployment time and location coordinates of each sensor, providing a reliable foundation for subsequent data processing and analysis. The data processing module utilizes advanced filtering algorithms and time synchronization calibration technology to effectively remove noise interference, unify timestamps from multiple data sources, and compensate and correct the data, significantly improving data quality and accuracy, providing a reliable basis for subsequent analysis and decision-making. The visualization module, by constructing a 3D model of the grouting area, can intuitively display the flow and filling of the grout, as well as pressure changes within the area, and can accurately calculate the filling rate, helping staff to promptly grasp the grouting status and make scientific and reasonable decisions. The prediction module, combining a rule engine and LSTM neural network, can accurately predict future trends in grouting parameters. Once an anomaly is detected, it can promptly issue alarms and take control measures, effectively avoiding potential safety risks and quality problems. Simultaneously, storing data in a database and backing it up to the cloud ensures the traceability of historical data, providing strong support for subsequent data analysis and summarization. Attached Figure Description
[0013] Figure 1 This is a schematic diagram of a real-time grouting monitoring system proposed in this invention; Figure 2 This is a schematic diagram of grouting in irregular holes for a real-time grouting monitoring system proposed in this invention. Detailed Implementation
[0014] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0015] Reference Figures 1-2 A real-time grouting monitoring system includes the following operating steps: Step 1: A high-precision sensor with a stainless steel alkali-resistant encapsulation is initialized. The data acquisition frequency is set to 5Hz and the wireless transmission power is adjusted to 20dBm. The sensor integrates pressure, temperature, and flow rate detection modules using MEMS technology and connects to an edge computing gateway via an RS485 bus. Initialization parameters include range calibration, sampling frequency, and wireless transmission power. Specifically, a MEMS integrated sensor is selected and initialized. A high-precision pressure calibrator is used. The sensor is connected to the calibrator, and standard pressure values of 0%, 25%, 50%, 75%, and 100% of the range are applied. The electrical signal values output by the sensor are recorded. A linear fit is performed between the standard pressure values from the calibrator and the sensor output values to obtain the calibration coefficient. Through calibration, the sensor's pressure measurement accuracy reaches [the required level]. FS; Place the sensor in a high-precision constant temperature bath, setting the bath temperature to different points: -20℃, 0℃, 20℃, 40℃, 60℃, and 80℃. Record the sensor's output value at each temperature point. Establish a temperature-output value correspondence based on the standard temperature value and the sensor output value, and perform calibration compensation to control the temperature measurement error within a certain range. Within a certain range; using a standard flow calibration device, the sensor's output signal is measured at different flow rates, such as 0 m³ / h, 5 m³ / h, 10 m³ / h, 15 m³ / h, and 20 m³ / h. Based on the calibrator's standard flow rate and the sensor's output signal, a flow calibration curve is determined to achieve the required flow measurement accuracy. .
[0016] Step 2: Run the grouting dosing management system. This system includes a data acquisition module, a data processing module, a visualization module, a prediction module, an acoustic monitoring module, and a report generation module. The data acquisition module uses sensors mounted on customized density-matching particles to be added to the grouting fluid. These particles flow and distribute evenly with the grout, recording the dosing time and location coordinates of each sensor in real time and generating an electronic log. Simultaneously, it periodically reads data from each sensor according to a preset Modbus / TCP protocol. During data reading, the program performs a preliminary validity check, verifying that the data is within a reasonable range. Abnormal data is marked and error messages are recorded. The density of the customized density-matching particles is achieved by adjusting the mixing ratio of hollow glass microspheres and epoxy resin, with a density range of 1.2-2.5 g / cm³. A fluorescent tracer coating is sprayed onto the particle surface, and the position coordinates are tracked in real time using UWB positioning technology, ensuring high positioning accuracy. 10cm.
[0017] Specifically, acoustic monitoring includes acoustic signature extraction, anomaly classification models, and location assistance. Acoustic signature extraction transforms the raw signal into computable features, reducing the complexity of model processing. The anomaly classification model automatically identifies abnormal states during the grouting process and outputs the probability of anomalies. Location assistance determines the specific location of abnormal sound sources, providing precise location for grouting process adjustments or fault repairs. The data acquisition module and acoustic monitoring module work together to construct a three-dimensional pore model of the grouting area in real time and automatically calculate the target filling volume. Based on the geometric dimensions, the concrete grout mix ratio and grouting parameters are dynamically adjusted according to the following rules: When the volume of the hole If the minimum cross-sectional diameter is >80cm, it is determined to be large-diameter grouting. The water-cement ratio is adjusted to 1:0.7, 0.3% water-reducing agent is added, the slump is controlled at 180-200mm, and the number of particles carried by the sensor is increased to 20-30 per m³.
[0018] When the volume of the hole or minimum cross-sectional diameter When the depth is 30cm, it is determined to be microporous grouting. The water-cement ratio is adjusted to 1:0.5, 5% silica fume is added, the slump is controlled at 120-150mm, and the number of particles on the sensor is reduced to 10-15 per m³.
[0019] Prepared according to intelligent proportions The concrete slurry is uniformly mixed with the corresponding number of sensor particles and mixed for ≥3 minutes using a dual-shaft forced mixer to ensure dispersion. It is then injected into the holes through a grouting pump. The SCADA system monitors the pressure, flow rate, and temperature feedback from the sensors in real time and updates the grouting parameters every 5 seconds to achieve closed-loop control of "detection-proportioning-grouting".
[0020] Step 3: The data processing module performs in-depth processing on the collected data. It uses wavelet filtering or Kalman filtering algorithms from the MATLAB algorithm library to filter noise, removing high-frequency and dynamic noise to make the data smoother and more accurate. Furthermore, it performs time synchronization calibration using the NTP protocol based on the Pandas library, unifying the timestamps of multi-source data to ensure data time consistency. Simultaneously, it calls the sensor calibration model to compensate for the data, especially correcting temperature drift issues to improve data accuracy. The data processing module uses the extended Kalman filter algorithm to reduce noise in the temperature signal and establishes the temperature state equation. ,in For temperature change rate, process noise .
[0021] Specifically, the noise filtering operation steps are as follows: S1. Read the raw temperature signal Remove obvious outliers and initialize EKF modeling parameters. , .
[0022] S2. Real-time filtering processing, pseudocode as follows: # Initialize parameters alpha = 0.01# Damping coefficient delta_t = 0.2 # Sampling interval Q = np.diag([0.1**2, 0.05**2]) # Process noise covariance R = 0.25# Observation noise covariance x_hat = np.array([z[0], 0]) # Initial state P = np.diag([1, 0.1]) # Initial covariance for k in range(1, len(z)): # Prediction Phase F = np.array([[1, delta_t], [-alpha*delta_t, 1 - alpha*delta_t]]) x_pred = np.dot(F, x_hat) P_pred = np.dot(F, np.dot(P, FT)) + Q # Update Phase H = np.array([[1, 0]]) K = np.dot(P_pred, HT) / (np.dot(H, np.dot(P_pred, HT)) + R) x_hat = x_pred + K * (z[k] - x_pred[0]) P = np.dot((np.eye(2) - np.outer(K, H)), P_pred) # Output filtered temperature value T_filtered[k] = x_hat[0]; S3. Compare the signals before and after filtering, and calculate the improvement in signal-to-noise ratio (SNR). .
[0023] Step 4: The visualization module is used to construct a 3D model of the grouting area. Point cloud data acquired using 3D laser scanning is divided into tetrahedral meshes using the Delaunay triangulation algorithm to construct the 3D model of the grouting area. Based on the location data collected by sensors, the grout filling status in the model is updated in real time using a mesh generation algorithm, visually displaying the grout flow and filling situation. Kriging interpolation is used to spatially interpolate the discrete pressure data from the sensors to generate a continuous pressure field. Then, a pressure cloud map is generated using a contour algorithm to visually display the pressure changes within the area. Finally, the volume integral method is used to accurately calculate the filling rate, and the sensor calibration model is called to compensate for the data, including temperature drift correction. The correction formula is as follows: ,in For temperature coefficient, For real-time temperature, This is a reference temperature.
[0024] Specifically, the system receives the location coordinates collected by sensors in real time and quickly locates the corresponding tetrahedral mesh element using a spatial indexing structure. When a sensor enters a mesh element and stays for ≥2 seconds, the element is marked as "filling" and displayed as a semi-transparent blue in the 3D model. When ≥3 sensors within the same element continuously provide data for 10 seconds, the element is marked as "filled" and displayed as an opaque gray in the 3D model. Unfilled areas are shown transparently. Based on the sensor's movement trajectory, the flow direction of the slurry is simulated using Bézier curves, and arrow icons indicate the flow rate update status in the 3D model. Simultaneously, the volume integral method is used to accurately calculate the filling rate. , The total volume of the grouting area. For cells that have already been grouted.
[0025] Step 5: The prediction module is used to predict future data. It defines thresholds through a preset rule engine and combines them with an LSTM neural network trained using TensorFlow to predict parameter trends. Inputting the pressure and temperature sequence of the previous 10 minutes, the model predicts data for the next 2 minutes. If an anomaly is detected, the SCADA monitoring system sends control commands to the grouting pump and simultaneously triggers an on-site alarm to alert personnel. If no anomalies are detected, the data is stored in the database, and key data is backed up to cloud storage to ensure historical data traceability. The LSTM neural network structure of the prediction module includes an input layer, hidden layers, and an output layer. The input layer receives preprocessed grouting pressure and temperature time series data and converts the raw data into a dimensional structure suitable for LSTM processing. The hidden layer uses two LSTM units and a "forget-store-output" collaborative mechanism to capture the long-term dependency characteristics of grouting parameters. The output layer uses a linear activation function to adapt to the prediction requirements of continuous pressure and temperature values. Based on the features extracted from the hidden layer, it outputs the predicted grouting pressure and temperature values for the next 2 minutes.
[0026] Specifically, step five integrates time-series data prediction with rule-driven decision-making to achieve forward-looking control and abnormal risk management of the grouting process. It also combines preset rules to respond to anomalies in real time, ensuring grouting quality and safety, and providing a 2-minute early warning to allow time for manual or automatic intervention and reduce the accident rate.
[0027] Step Six: The report generation module uses Python's Pandas library to process the data, and then uses the Matplotlib / Seaborn library to generate visualization charts. It integrates all the data and charts to generate a PDF / Excel report. Specifically, it uses Matplotlib to draw line graphs of parameters such as pressure and temperature over time to visually display the dynamic changes during the grouting process.
[0028] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A real-time grouting monitoring system, characterized in that: The following steps are included: Step 1: Use a high-precision sensor with a stainless steel alkali-resistant encapsulation shell, initialize the sensor parameters, set the data acquisition frequency to 5Hz and adjust the wireless transmission power to 20dBm; Step 2: Run the grouting delivery management system. The grouting delivery management system includes a data acquisition module, a data processing module, a visualization module, a prediction module, and a report generation module. The data acquisition module uses sensors mounted on customized density-matching particles to be delivered into the grouting fluid. The particles flow and distribute evenly with the grout, and the system records the delivery time and location coordinates of each sensor in real time, generating electronic logs in real time. At the same time, the system reads the data from each sensor periodically according to the preset Modbus / TCP protocol. When reading the data, the program will perform a preliminary legality check to check whether the data is within a reasonable range. If abnormal data is found, it will be marked and error information will be recorded. Step 3: The data processing module performs in-depth processing on the collected data. It uses wavelet filtering or Kalman filtering algorithms from the MATLAB algorithm library to filter noise, remove high-frequency noise and dynamic noise, and make the data smoother and more accurate. It also uses the NTP protocol based on the Pandas library to perform time synchronization calibration, unify the timestamps of multi-source data, and ensure the consistency of data time. At the same time, it calls the sensor calibration model to compensate for the data, especially to correct the temperature drift problem and improve the accuracy of the data. Step 4: The visualization module is used to construct a three-dimensional model of the grouting area. Point cloud data obtained by three-dimensional laser scanning is divided into tetrahedral meshes using the Delaunay triangulation algorithm to construct a three-dimensional model of the grouting area. Based on the position data collected by the sensor, the filling state of the grout in the model is updated in real time through the mesh division algorithm, which intuitively displays the flow and filling of the grout. The Kriging interpolation method is used to spatially interpolate the discrete pressure data of the sensor to generate a continuous pressure field. Then, the pressure cloud map is generated through the contour algorithm to intuitively display the pressure changes in the area. Finally, the volume integral method is used to accurately calculate the filling rate. Step 5: The prediction module is used to predict future data. It defines thresholds through a preset rule engine and combines them with an LSTM neural network trained based on TensorFlow to predict parameter trends. Inputting the pressure and temperature sequence of the previous 10 minutes, the model predicts the data for the next 2 minutes. If an anomaly is detected, the SCADA monitoring system sends a control command to the grouting pump and triggers the on-site alarm to remind the staff. If there is no anomaly, the data is stored in the database and key data is backed up to cloud storage to ensure that historical data is traceable. Step 6: The report generation module processes the data using Python's Pandas library, then uses the Matplotlib / Seaborn library to generate visualization charts, integrates all the data and charts, and generates a PDF / Excel report.
2. The grouting real-time monitoring system according to claim 1, characterized in that: In step one, the sensor uses MEMS technology to integrate pressure, temperature and flow rate detection modules, and connects to the edge computing gateway via RS485 bus. The initialization parameters include range calibration, sampling frequency and wireless transmission power.
3. The grouting real-time monitoring system according to claim 1, characterized in that: In step two, the density of the customized density-matching particles is achieved by adjusting the mixing ratio of hollow glass microspheres and epoxy resin, with a density range of 1.2-2.5 g / cm³. A fluorescent tracer coating is sprayed onto the particle surface, and the position coordinates are tracked in real time using UWB positioning technology, ensuring high positioning accuracy. 10cm.
4. The grouting real-time monitoring system according to claim 1, characterized in that: In step three, the data processing module uses an extended Kalman filter algorithm to reduce noise in the temperature signal and establishes a temperature state equation. ,in For the rate of temperature change, For process noise, .
5. The grouting real-time monitoring system according to claim 1, characterized in that: In step four, the sensor calibration model is invoked to compensate the data, including temperature drift correction. The correction formula is as follows: ,in For temperature coefficient, For real-time temperature, This is a reference temperature.
6. The grouting real-time monitoring system according to claim 1, characterized in that: In step five, the LSTM neural network structure of the prediction module includes an input layer, a hidden layer, and an output layer. The input layer receives preprocessed grouting pressure and temperature time series data and converts the original data into a dimensional structure suitable for LSTM processing. The hidden layer uses two LSTM units and a "forget-store-output" collaborative mechanism to capture the long-term dependency features of grouting parameters. The output layer uses a linear activation function to adapt to the prediction requirements of continuous pressure and temperature values. Based on the features extracted by the hidden layer, the predicted values of grouting pressure and temperature for the next two minutes are output.
7. The grouting real-time monitoring system according to claim 1, characterized in that: The grouting management system includes an acoustic monitoring module, specifically comprising acoustic signature extraction, an anomaly classification model, and location assistance. The acoustic signature extraction transforms the raw signal into computable features, reducing the complexity of the model processing. The anomaly classification model automatically identifies abnormal states during the grouting process and outputs the anomaly probability. The location assistance determines the specific location of the abnormal sound source, providing precise positioning for grouting process adjustments or fault repairs.