A tunnel deformation monitoring device based on millimeter-wave radar

Abstract

This invention discloses a tunnel deformation monitoring device based on millimeter-wave radar, comprising a main housing, millimeter-wave radar, laser sensor, PCBA, and bubble level. It integrates multimodal data fusion analysis, dynamic measurement range intelligent adjustment, real-time deformation prediction and control, self-calibration, and fault diagnosis modules. The multimodal fusion system fuses radar and laser data using a Kalman filter algorithm, and utilizes median filtering, Gaussian fitting preprocessing, and dynamic weight allocation to achieve high-precision monitoring across the entire measurement range. The dynamic measurement range module, based on a decision tree algorithm, intelligently switches the measurement and frequency sweep range according to mining progress, geological conditions, and distance fluctuation coefficient. The real-time prediction module uses an ARIMA model to predict deformation in advance and automatically adjusts the scanning angle and acquisition frequency. The self-calibration module corrects errors daily using a high-precision reflector. Supporting 5G communication and remote control, it combines high precision, adaptability, and high reliability, making it suitable for long-term deformation monitoring of mine tunnels.

Description

Technical Field

[0001] This invention relates to the field of tunnel safety monitoring technology, and in particular to a tunnel deformation monitoring device based on millimeter-wave radar. Background Technology

[0002] In underground engineering projects such as mining and tunnel construction, deformation of the surrounding rock in roadways is a key hidden danger threatening safe production. Traditional monitoring methods, such as mechanical measuring rods and total stations, have problems such as low single-point measurement coverage, limited frequency of manual inspections, and insufficient real-time performance, making it difficult to meet the dynamic deformation monitoring needs under complex geological conditions.

[0003] Currently, most tunnel deformation monitoring devices are based on radar or optical sensors. However, single sensors have significant limitations. Although millimeter-wave radar has the advantages of all-weather operation and strong penetration, its accuracy at close range is greatly affected by multipath effects. Laser rangefinders have high accuracy at close range, but are significantly affected by dust and water vapor at long range, and their fixed scanning angle makes it impossible to dynamically focus on the deformed area.

[0004] In addition, existing devices generally lack an adaptive measurement range adjustment mechanism, which can easily lead to monitoring blind spots or data redundancy when the mining progress changes or geological conditions become complex. At the same time, the sensor drift problem during long-term operation lacks automatic calibration and fault diagnosis capabilities, resulting in a decrease in monitoring accuracy over time and posing a risk of delayed safety warnings.

[0005] Therefore, how to provide a tunnel deformation monitoring device based on millimeter-wave radar is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a tunnel deformation monitoring device based on millimeter-wave radar. This invention achieves high-precision, adaptive, and high-reliability tunnel deformation monitoring and early warning through multi-modal data fusion, intelligent adjustment of dynamic measurement range, real-time deformation prediction, and self-calibration technology.

[0007] According to an embodiment of the present invention, a tunnel deformation monitoring device based on millimeter-wave radar includes a main housing, a rear cover, a millimeter-wave radar, a laser, a PCBA, and a bubble level. The millimeter-wave radar and laser are disposed inside the main housing and face the monitoring area. The PCBA integrates a data processing and control module. The side of the main housing is provided with a gland for the insertion and sealing of external cables. The bubble level is embedded in the top surface of the main housing. The rear cover is detachably installed on the rear side of the main housing. The device also includes a multimodal data fusion analysis system, a dynamic measurement range intelligent adjustment module, a real-time feedback and predictive control module, and a self-calibration and fault diagnosis module integrated on the PCBA.

[0008] The multimodal data fusion analysis system uses the processor built into the PCBA to fuse distance data acquired by millimeter-wave radar and laser, and generates fused distance data by combining the measurement advantages of both through a Kalman filter algorithm. The fusion process satisfies the following state equation: and observation equations ,in Let k be the state vector at time k, containing distance and velocity. The observation vector includes measurements from millimeter-wave radar and laser. , Here is the state transition matrix. and For process and observation noise;

[0009] The dynamic measurement range intelligent adjustment module utilizes the processor to calculate the distance fluctuation coefficient based on initial measurement data, mining progress data, and geological condition data using a decision tree algorithm. in, For the nth distance measurement, This is the average of the first three measurements, which can be determined according to... Dynamically adjust the measurement range of the millimeter-wave radar;

[0010] The real-time feedback and predictive control module constructs a deformation prediction model based on the ARIMA model, predicts deformation trends in real time, and adjusts the measurement strategy accordingly. The prediction model expression is: in, Predict the deformation at time t. and These are the autoregressive and moving average coefficients, respectively. Let be the residual at time t. The order of the autoregression determines the number of time steps of past data considered in the autoregressive model. express A data sequence value related to deformation at a given time. The order of the moving average determines the number of time steps of the past residuals considered in the moving average model.

[0011] The self-calibration and fault diagnosis module periodically calls the built-in 2-meter standard distance reflector to emit signals, based on the correction formula: Calibrate measurement errors;

[0012] in, For radar measurement error rate, The error rate of laser measurement is used to monitor equipment malfunctions through sensor operating parameters.

[0013] Furthermore, the multimodal data fusion analysis system includes a preprocessing unit, a weight allocation unit, and a Kalman filter unit. The preprocessing unit can use median filtering to remove data from millimeter-wave radar data. External impulse noise, To measure the standard deviation of the data, Gaussian fitting was performed on the laser data using the least squares method. Determine the center of the light spot ,in, For light spot intensity, The coordinates of the center of the light spot are The standard deviation of the light spot is denoted as . The fitting error is verified by the root mean square error (RMSE) and satisfies RMSE ≤ 0.5 mm. The weight allocation unit dynamically allocates weights according to the measurement distance. When the distance is ≤ 5 meters, the weight of laser data accounts for ≥ 70%. When the distance is > 5 meters, the weight of millimeter-wave radar data accounts for ≥ 60%. The Kalman filter unit iteratively calculates the fused distance measurement value through the state equation and the observation equation.

[0014] Furthermore, the adjustment method of the dynamic measurement range intelligent adjustment module includes the following steps:

[0015] S1. Obtain mining progress data and geological condition data through an external RS485 interface. The mining progress data includes excavation depth and support type, and the geological condition data includes rock compressive strength and joint density.

[0016] S2. If the distance fluctuation coefficient of three consecutive measurements is ≤5% and the excavation speed is ≤0.5m / h, set the measurement range of the millimeter-wave radar to 0-10 meters.

[0017] S3. If the fluctuation coefficient is greater than 5% or the excavation speed is greater than 0.5m / h, the measurement range will be automatically extended to 0-30 meters.

[0018] The parameter thresholds are stored in the PCBA's EEPROM.

[0019] Furthermore, the real-time feedback and predictive control module includes a predictive model building unit, a real-time early warning unit, and a measurement strategy adjustment unit. The predictive model building unit trains an ARIMA model using historical fusion distance data, optimizes the order through the BIC criterion, and establishes a functional relationship between deformation and time. The real-time early warning unit drives an external audible and visual alarm when the predicted deformation of the mining roadway is ≥50mm or the deformation of the transport roadway is ≥80mm in the next 2 hours. The measurement strategy adjustment unit controls the sensor scanning angle to deflect ±15° to focus on the deformation area when the deformation rate is detected to be ≥10mm / h, and increases the acquisition frequency from once every 60 seconds to once every 10 seconds.

[0020] Furthermore, the self-calibration and fault diagnosis module includes a self-calibration unit and a fault diagnosis unit. The self-calibration unit triggers a calibration program at 0:00 daily, controlling the millimeter-wave radar and laser to emit signals to a 2-meter standard reflector built into the main housing. The reflector has a reflectivity of ≥95%, and the reference distance is 2 meters ±0.1 mm. The measurement error is calculated based on the reflection time. as well as The fault diagnosis unit can monitor the transmission power of the millimeter-wave radar and the spot offset of the laser in real time. The normal range of the transmission power of the millimeter-wave radar is 20-25dBm, and the normal range of the spot offset is ±3°. If the transmission power is <18dBm or the offset is >5°, the sensor is determined to be faulty and a fault code is generated and stored in FLASH.

[0021] Furthermore, the PCBA integrates a 5G communication module, which can upload fused distance data, prediction results, and fault codes to a cloud server via the MQTT protocol, and can also receive remote control commands, supporting remote adjustment of measurement range, acquisition frequency, and early warning threshold parameters.

[0022] Furthermore, the preprocessing unit uses a 5×5 window median filter for the millimeter-wave radar data and employs the least squares method for Gaussian fitting of the laser data. The fitting error is verified by the root mean square error and satisfies RMSE≤0.5mm.

[0023] Furthermore, the input parameters of the decision tree algorithm also include the tunnel support material and the monitoring period, and the output is the frequency sweep range of the millimeter-wave radar. The frequency sweep range of the millimeter-wave radar is adjustable from 24GHz to 77GHz. The tunnel support material includes anchor bolts and anchor cables. The monitoring period includes the tunneling period and the stabilization period, wherein the tunneling period automatically expands to 60-77GHz and the stabilization period shrinks to 24-30GHz.

[0024] Furthermore, in the real-time early warning unit, the roadway types of the mining roadway and the transportation roadway are read by an external RFID tag connected to the PCBA or input by a key. The tag stores the roadway type code, and the system automatically matches the early warning threshold according to the code.

[0025] Furthermore, the standard reflector built into the main housing is a high-precision diffuse reflector, installed in the center of the inner wall of the main housing, with a surface roughness Ra≤0.2μm, to ensure stable reflection of the calibration signal and consistency with the measurement reference.

[0026] The beneficial effects of this invention are:

[0027] 1. This invention employs a multi-modal data fusion scheme combining millimeter-wave radar and laser sensors. By dynamically combining the advantages of both through a Kalman filter algorithm and designing a preprocessing unit and a dynamic weight allocation mechanism, it solves the problems of the sharp drop in accuracy of traditional single laser sensors at long distances due to dust interference and the insufficient accuracy of single radar sensors at short distances. After fusion, the accuracy across the entire range can be effectively improved, meeting the millimeter-level deformation monitoring requirements of mine roadways. Furthermore, by removing radar pulse noise through median filtering and correcting laser spot offset through Gaussian fitting, the data efficiency can be greatly improved in underground environments with high dust concentrations, significantly reducing the impact of invalid data on monitoring results.

[0028] 2. Based on the decision tree algorithm, this invention integrates mining progress, geological conditions and real-time measurement data to dynamically adjust the measurement range and frequency sweep range of the millimeter-wave radar. This effectively avoids the defects of fixed-range monitoring in traditional devices. No redundant monitoring is required during the stable period, and the range is automatically expanded to capture key deformation areas during the tunneling period. It balances monitoring accuracy and equipment energy consumption. Furthermore, it automatically switches the frequency sweep strategy according to the support type and monitoring period, improving adaptability to complex geological conditions such as jointed zones and soft rock sections, and achieving "on-demand monitoring and accurate measurement".

[0029] 3. This invention constructs a deformation prediction model based on the ARIMA model, optimizes the model order through the BIC criterion, predicts the deformation amount in the next 2 hours in real time, and adjusts the measurement strategy accordingly. Compared with traditional devices that can only display data in real time, this invention can identify deformation trends in advance, providing an early warning time of 2 hours, thus gaining valuable time for roadway support and reinforcement, effectively reducing the risk of deformation runaway. Furthermore, when the deformation rate is ≥10mm / h, the sensor scanning angle is automatically deflected and the acquisition frequency is increased, increasing the data density in the deformation area from 1 set per minute to 6 sets per minute, improving its ability to capture subtle deformations, thereby effectively identifying local sudden deformations.

[0030] 4. This invention utilizes a built-in 2-meter high-precision diffuse reflector plate to automatically trigger a calibration program at 0:00 daily. By fusing radar and laser error rates through a correction formula, it monitors radar transmission power and laser spot offset in real time, and can generate, store, and upload fault codes. Through daily self-calibration, the error rate of long-term measurements by millimeter-wave radar and laser sensors can be effectively reduced, eliminating the need for regular manual calibration, thus reducing maintenance costs. It is suitable for long-term unattended mine monitoring scenarios, and the fault diagnosis response time is less than 30 seconds. It can accurately identify abnormal radar power and automatically switch to redundant monitoring mode, significantly reducing the fault missed rate and improving system reliability.

[0031] 5. This invention integrates a 5G communication module, utilizes RFID tags or buttons to input the roadway type, automatically matches early warning thresholds, supports remote adjustment of measurement parameters, and significantly shortens on-site configuration time by reading RFID tags or inputting roadway types via buttons, avoiding errors in manual parameter setting. The cloud-based remote control function supports real-time response to complex working conditions, greatly improving equipment deployment efficiency. Furthermore, by constructing a three-dimensional monitoring network through the 5G network, it achieves data synchronization and fusion analysis of multiple devices, providing a wider coverage area than single-point monitoring and offering comprehensive data support for the overall stability assessment of roadways. Attached Figure Description

[0032] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0033] Figure 1 This is a schematic diagram of the overall structure of a tunnel deformation monitoring device based on millimeter-wave radar proposed in this invention.

[0034] Figure 2 This is a flowchart of a multimodal data fusion analysis system for a tunnel deformation monitoring device based on millimeter-wave radar proposed in this invention.

[0035] Figure 3 This is a logic block diagram of the intelligent adjustment of the dynamic measurement range of a tunnel deformation monitoring device based on millimeter-wave radar proposed in this invention.

[0036] Figure 4 This is a flowchart of the real-time feedback and predictive control module of a tunnel deformation monitoring device based on millimeter-wave radar proposed in this invention.

[0037] Figure 5 This is a hardware connection diagram of the self-calibration and fault diagnosis module of a tunnel deformation monitoring device based on millimeter-wave radar proposed in this invention.

[0038] In the diagram: 1. Main housing; 2. Back cover; 3. Millimeter-wave radar; 4. Laser sensor; 5. Gland head; 6. Bubble level; 7. PCBA. Detailed Implementation

[0039] In the description of this invention, the terms "upper," "lower," "inner," and "outer," etc., indicating orientation or positional relationships, are generally based on the orientation of the device during normal installation and use. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Unless otherwise explicitly specified and limited, the terms "set," "install," and "connect," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two elements. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0040] In this application, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art. In case of any inconsistency, the meaning set forth in this specification or derived from the content described herein shall prevail. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0041] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. This embodiment takes a specific mine roadway monitoring scenario as an example to illustrate the working process of the roadway deformation monitoring device based on millimeter-wave radar 3.

[0042] Please see Figures 1-5 A tunnel deformation monitoring device based on millimeter-wave radar 3 includes a main housing 1, a rear cover 2, a millimeter-wave radar 3, a laser sensor 4, a PCBA 7, and a bubble level 6. The main housing 1 and the rear cover 2 are made of high-strength, corrosion-resistant engineering plastics or metal materials, such as aluminum alloy. This material ensures the structural strength of the device, enabling it to adapt to the complex environment of underground mines, while also possessing good corrosion resistance to prevent damage to the equipment caused by factors such as underground humidity and corrosive gases. The main housing 1 and the rear cover 2 are connected by screws or clips, facilitating the installation and maintenance of internal components.

[0043] The millimeter-wave radar 3 and the laser sensor 4 are installed inside the main housing 1, and both of them are directed towards the tunnel area to be monitored, so as to obtain the distance data of the tunnel in real time.

[0044] The millimeter-wave radar 3 is a model with high resolution and high precision measurement capabilities, capable of accurately measuring the distance information of target objects. The laser sensor 4 is a high-precision laser rangefinder sensor, which utilizes the linear propagation characteristics of laser light to calculate distance by measuring the time difference between laser emission and reception, exhibiting high accuracy and stability.

[0045] PCBA7 integrates a data processing and control module, responsible for receiving and processing data from millimeter-wave radar 3 and laser sensor 4, and controlling the operation of various functional modules. PCBA7 includes components such as a microprocessor, memory chips, and a communication module. The microprocessor needs to have high-speed data processing capabilities and abundant interface resources to meet the real-time processing requirements of the data acquired by millimeter-wave radar 3 and laser sensor 4. The memory chip is used to store the acquired data, processing algorithms, and system configuration parameters, such as using a large-capacity Flash memory. The communication module integrates 5G communication functionality, using a standard-compliant 5G communication chip to ensure stable and fast data transmission to the cloud server.

[0046] The main housing 1 has a gland 5 on its side for inserting and sealing external cables. The gland is made of stainless steel, has an IP68 protection rating, and features an internal rubber sealing ring to effectively prevent dust and moisture from entering the housing. After the cable is inserted through the gland, it is secured with a compression nut to ensure a stable and reliable seal, meeting the explosion-proof and moisture-proof requirements for underground equipment in coal mine safety regulations.

[0047] The bubble level 6 uses a high-precision circular or tubular bubble level 6. By observing the position of the bubble in the level, the operator can easily adjust the device to a horizontal state, ensuring that the measurement direction of the millimeter-wave radar 3 and the laser sensor 4 is accurate and avoiding measurement errors caused by device tilt.

[0048] In this embodiment, a suitable installation location is first selected, generally a location on the tunnel wall with a wide field of vision, allowing clear observation of the monitoring area. The main housing 1 is then securely installed on the tunnel wall using expansion bolts or other fixing devices. During installation, the angle of the main housing 1 is adjusted by observing the bubble level 6, ensuring the bubble is centered on the level and the device is horizontal. After installation, the data and power cables between the millimeter-wave radar 3, laser sensor 4, and PCBA7 are connected, ensuring a secure connection to prevent data transmission abnormalities or power outages due to loosening. Finally, the rear cover 2 is placed on top for further protection of the internal components.

[0049] Please see Figure 2 The multimodal data fusion analysis system includes a preprocessing unit, a weight allocation unit, and a Kalman filter unit.

[0050] Specifically, the preprocessing unit can perform preliminary processing on the data collected by millimeter-wave radar 3 and laser sensor 4. For the millimeter-wave radar 3 data, a 5×5 window median filtering algorithm is used to remove external impulse noise. In actual mining environments, there are various sources of electromagnetic interference, such as underground electrical equipment and communication equipment. These interferences may cause abnormal impulse noise in the data collected by millimeter-wave radar 3. The principle of the median filtering algorithm is to arrange the data in a 5×5 window in ascending order and take the median value as the data output at the center of the window.

[0051] For example, suppose the data in the window is:

[0052] [12,5,20,3,25,18,30,8,40,15,50,35,60,22,45,7,32,55,47,52,10,27,37,42,62];

[0053] Sort these data from smallest to largest, and we get:

[0054] [3,5,7,8,10,12,15,18,20,22,25,27,30,32,35,37,40,42,45,47,50,52,55,60,62];

[0055] The 5x5 window contains 25 data points, so the median value is 30. The center of the window can be output with the value 30 to remove outliers such as impulse noise.

[0056] In this way, impulse noise can be effectively removed, improving the stability and reliability of the data.

[0057] For the data from laser sensor 4, Gaussian fitting using the least squares method was employed to determine the center of the laser spot. During laser propagation, factors such as dust and fog within the tunnel can distort the spot shape, affecting the accuracy of distance measurements. The principle of the least squares method is to find a set of parameters that minimizes the sum of squared errors between the measured data and the theoretical model.

[0058] In this embodiment, it is assumed that the theoretical model of the laser spot is a Gaussian function. ,in For light spot intensity, The coordinates of the center of the light spot are The standard deviation of the light spot is given. The standard deviation is continuously adjusted by performing least-squares fitting on multiple light spot data points collected by laser sensor 4. , , and The value of minimizes the sum of squared errors between the fitted curve and the actual data points. Extensive experimental verification shows that this fitting method achieves a root mean square error (RMSE) of ≤0.5mm, accurately determining the center position of the light spot and providing a reliable data foundation for subsequent distance calculations.

[0059] The weighting unit dynamically assigns weights to the millimeter-wave radar 3 data and the laser sensor 4 data based on the measurement distance. When the measurement distance is ≤5 meters, the laser sensor 4 has higher measurement accuracy, and the laser data weight accounts for ≥70%. This is because, at close range, the laser sensor 4 is less affected by environmental factors and can provide more accurate distance measurements.

[0060] For example, at a distance of 3 meters, after multiple experimental verifications, the measurement error of the laser sensor 4 is within ±1mm, while the measurement error of the millimeter-wave radar 3 is around ±3mm. Therefore, when fusing data, giving higher weight to the laser data can improve the accuracy of the fusion results.

[0061] When the distance is greater than 5 meters, the millimeter-wave radar 3 has certain advantages in long-distance measurement, with its data weighting accounting for ≥60%. As the measurement distance increases, the signal strength of the laser sensor 4 will gradually weaken and become more susceptible to environmental factors. However, the millimeter-wave radar 3, by emitting millimeter-wave signals, can maintain good measurement performance at a greater distance.

[0062] For example, at a distance of 10 meters, the measurement error of millimeter-wave radar 3 is around ±5mm, while the measurement error of laser sensor 4 may increase to more than ±10mm. In this case, increasing the weight of millimeter-wave radar 3 data can better reflect the actual distance situation.

[0063] The Kalman filter unit is based on the state equation as follows: and observation equations ,in Let k be the state vector at time k, containing distance and velocity. and These are process noise and observation noise, respectively. Process noise The main sources are uncertainties within the system, such as the impact of electronic noise and mechanical vibration on measurement results; observation noise. This is mainly caused by the measurement errors of the millimeter-wave radar 3 and the laser sensor 4 themselves; The observation vector encompasses the measurements from millimeter-wave radar 3 and laser sensor 4; This is the state transition matrix, used to describe the relationship between states and time. When acquiring measurements, the data from the millimeter-wave radar 3 and the laser sensor 4 need to be processed synchronously to ensure data fusion occurs at the same time.

[0064] In practical calculations, these two types of noise need to be estimated and processed. Typically, statistical analysis of a large amount of historical data can be used to obtain the statistical characteristics of the noise, such as the mean and variance, and then appropriate processing can be performed in the Kalman filter algorithm to improve the accuracy of the fusion results.

[0065] In practical applications, distance information can be directly obtained from the measurement data of the millimeter-wave radar 3 and the laser sensor 4, while velocity information can be obtained by differential calculation of distance data at multiple consecutive time points. For example, The distance of time is ,exist The distance of time is The sampling time interval is Then speed .

[0066] In this embodiment, it is assumed that the state transition matrix is... ,in The sampling time interval is represented by this matrix. This matrix indicates that within one sampling period, the change in distance is related to the velocity, while the velocity remains constant.

[0067] Through continuous iteration, the Kalman filter unit outputs a fused, high-precision distance measurement. In each iteration, the state vector from the previous time step is first used as the basis for the calculation. The state transition matrix F predicts the state vector at the current time step. ,Right now Then, based on the observation vector and predicted state vector Calculate Kalman gain The calculation formula is: ,in, The covariance matrix of the predicted state vector, For the observation matrix, This is the covariance matrix of the observation noise. Finally, based on the Kalman gain... For the predicted state vector After making corrections, the optimal estimated state vector for the current time step is obtained. ,Right now Through this iterative process, the fusion results are continuously optimized to make them closer to the true distance value.

[0068] The dynamic measurement range intelligent adjustment module acquires mining progress and geological condition data via an external RS485 interface. The RS485 interface is a commonly used industrial communication interface with advantages such as long transmission distance and strong anti-interference capabilities, making it ideal for use in underground mining environments. Mining progress data includes key information such as excavation depth and support type. Excavation depth can be obtained through depth sensors installed on the excavating equipment, while support type can be manually entered or obtained from the mine's production management system. Geological condition data covers rock compressive strength and joint density. Rock compressive strength can be obtained through on-site sampling and laboratory compressive strength testing; joint density can be obtained through on-site observation and statistical analysis by geological surveyors in the tunnel. This data is transmitted to the PCBA7 data processing and control module via the RS485 interface using a specific communication protocol.

[0069] The adjustment strategy can first be based on the distance fluctuation coefficient calculated using the decision tree algorithm. ,in, For the nth distance measurement, This is the average of the first three measurements, which can be determined according to... The measurement range of the millimeter-wave radar 3 is dynamically adjusted. The decision tree algorithm is a classification and prediction algorithm based on a tree structure. In this embodiment, it takes initial measurement data, mining progress data, geological condition data, tunnel support materials, and monitoring period as input parameters and outputs the frequency sweep range of the millimeter-wave radar 3.

[0070] If the distance fluctuation coefficient is ≤5% for three consecutive measurements and the excavation speed is ≤0.5m / h, it indicates that the tunnel deformation is relatively stable. In this case, the measurement range of the millimeter-wave radar 3 should be set to 0-10 meters. This is because a smaller measurement range is sufficient to meet monitoring needs when the tunnel deformation is stable, while also reducing the power consumption and data processing load of the equipment. For example, in a stable tunnel section, after multiple consecutive measurements, the distance fluctuation coefficient is around 3%, and the excavation speed is 0.3m / h. In this case, setting the measurement range of the millimeter-wave radar 3 to 0-10 meters can ensure effective monitoring of tunnel deformation while reducing unnecessary resource consumption.

[0071] Conversely, if the fluctuation coefficient is greater than 5% or the excavation speed is greater than 0.5 m / h, it indicates that the roadway deformation may be large or changing rapidly, and the measurement range is automatically expanded to 0-30 meters. In this case, a larger measurement range allows for more comprehensive monitoring of roadway deformation and timely detection of potential safety hazards. For example, during roadway excavation, if the excavation speed is high, reaching 0.8 m / h, and the distance fluctuation coefficient is also large, exceeding 5%, expanding the measurement range to 0-30 meters allows for real-time monitoring of roadway deformation at a greater distance, ensuring construction safety.

[0072] Relevant parameter thresholds are stored in the EEPROM of PCBA7. EEPROM is characterized by data retention even when power is off, ensuring that previously set parameter thresholds are preserved even after device restarts or battery replacements. Operators can modify and adjust the parameter thresholds in the EEPROM using the debugging interface on PCBA7 or remote control commands, as needed.

[0073] The input parameters for the decision tree algorithm also include roadway support materials such as anchor bolts and cables, and monitoring periods during the excavation and stabilization phases. During the excavation phase, due to continuous roadway excavation and significant changes in surrounding rock stress, the risk of deformation is high. Therefore, the sweep frequency range of the millimeter-wave radar 3 is automatically extended to 60-77 GHz. A higher sweep frequency range provides higher resolution and more accurate measurement results, enabling timely detection of minor roadway deformations during excavation. During the stabilization phase, roadway deformation is relatively small, and to reduce equipment power consumption and data processing load, the sweep frequency range is narrowed to 24-30 GHz.

[0074] The real-time feedback and predictive control module includes a predictive model building unit, a real-time early warning unit, and a measurement strategy adjustment unit.

[0075] The prediction model building unit trains the ARIMA model using historical fusion distance data. The ARIMA model is a commonly used time series forecasting model that effectively captures the trend, seasonality, and randomness characteristics of time series data.

[0076] In this implementation, historical fusion distance data over a period of time is first collected, such as fusion distance data collected at regular time intervals over the past month, such as every hour. Then, this data is preprocessed, including outlier removal and stabilization. Stabilization can be achieved through differencing, that is, performing first-order or multi-order differencing on the time series data to ensure it meets the stationarity condition.

[0077] Model order is optimized using the BIC criterion. The BIC criterion is a commonly used criterion in model selection, considering both goodness of fit and model complexity. When training an ARIMA model, different autoregressive orders p, differencing orders d, and moving average orders q are tried, and the BIC value of each model is calculated. The model with the lowest BIC value is selected as the optimal model. This ensures good model fit while avoiding overfitting. For example, after multiple experiments, it was found that when... When the model's BIC value is at its minimum, the ARIMA(2,1,1) model established at this time can fit the historical fusion distance data well and accurately predict future deformation trends.

[0078] By establishing a functional relationship between deformation and time, future deformation trends can be predicted. Assume the established ARIMA model expression is as follows: ,in, Predict the deformation at time t. and These are the autoregressive and moving average coefficients, respectively. Let be the residual at time t. In practical applications, the fused distance data up to the current time is substituted into the model to calculate the model parameters. and Then, based on these parameters, the deformation at different future times is predicted. Real-time early warning unit: When the predicted deformation of the mining roadway is ≥50mm or the deformation of the haulage roadway is ≥80mm within the next 2 hours, the external audible and visual alarm is immediately activated. Different warning thresholds are set for different types of roadways because mining roadways and haulage roadways differ in function, structure, and frequency of use. Mining roadways are directly connected to the coal face during coal mining; their deformation is directly affected by mining activities, and the deformation rate may be faster, requiring higher safety standards. Therefore, a lower warning threshold is set. Harvesting roadways, mainly used for transporting coal and equipment, deform relatively slowly, but large deformations can still affect transportation safety, so a relatively higher warning threshold is set.

[0079] The audible and visual alarm is selected with high brightness and high volume to ensure that workers can hear the alarm and see the warning light in a timely manner in the noisy underground mining environment. When the warning conditions are met, PCBA7 sends a trigger signal to the audible and visual alarm through the control interface, causing it to sound an alarm and reminding workers to take appropriate measures in time, such as strengthening support or stopping work.

[0080] When the measurement strategy adjustment unit detects a deformation rate ≥10mm / h, it controls the sensor scanning angle to deflect ±15° to focus on the deformed area in order to more accurately monitor the deformation region. In this embodiment, the scanning angles of the millimeter-wave radar 3 and the laser sensor 4 are controlled by a motor drive device. The motor drive device receives a control signal from the PCBA7 and drives the motor to rotate according to the signal's instructions, thereby driving the sensor's scanning mechanism to rotate and adjusting the scanning angle. For example, when the deformation rate of a certain area reaches 12mm / h, the PCBA7 sends a command to the motor drive device to deflect the sensor scanning angle 15° toward the deformed area, so as to more effectively acquire distance data for that area.

[0081] Simultaneously, the sampling frequency was increased from once every 60 seconds to once every 10 seconds. This was achieved by adjusting the sampling time interval of the data acquisition module in PCBA7. When the deformation rate is high, increasing the sampling frequency allows for the acquisition of more dense measurement data, enabling more timely understanding of changes in the deformation area. For example, during periods of high deformation rate, collecting data every 10 seconds can more accurately capture the details of roadway deformation, providing more comprehensive data support for subsequent analysis and decision-making.

[0082] The roadway type is read via an external RFID tag connected to the PCBA7 or entered via a keypad. The RFID tag stores the roadway type code. When the tag approaches the RFID reader module on the PCBA7, the module emits a radio frequency signal through its antenna, activating the tag and reading its stored roadway type code, such as an 8-bit binary code: "00000001" represents a mining roadway, "00000010" represents a transport roadway, and "00000011" represents a return air roadway, etc. The RFID reader module uses a 13.56MHz high-frequency band with a reading distance ≤5cm, ensuring that only designated tags can be identified when they are close, avoiding misreads.

[0083] If the button input method is selected, the PCBA7 panel has three function buttons: "Type+", "Type-", and "Confirm". The roadway type code is entered by long-pressing or short-pressing in combination. For example, long-pressing the "Type+" button for 3 seconds enters the setting mode, and a short press switches the code. The screen simultaneously displays the icon of the currently selected roadway type; for example, a longwall roadway displays the "Coal Face" icon, and a transport roadway displays the "Track" icon. After pressing the "Confirm" button, the code is stored in the PCBA7's EEPROM.

[0084] The system prioritizes RFID tag input. When a valid tag is detected, it automatically overwrites the currently stored roadway type, regardless of whether it was previously entered via button or the default setting. If no tag is detected, the code entered via button is used as the roadway type. Both input methods support dynamic modification during device operation. For example, when a roadway transitions from the excavation phase to the stable phase, the system can automatically match the corresponding early warning threshold and measurement strategy based on the new roadway type by replacing the RFID tag or resetting via button.

[0085] To prevent accidental operation, button input requires continuous and correct verification codes. If input is not completed within 30 seconds in setting mode, the system will automatically exit, preventing accidental touches that could corrupt parameters. RFID tags employ a unique coding binding mechanism; each tunnel type corresponds to a unique tag ID. The mapping relationship between tags and tunnel types can be configured on the cloud server via management software. For example, the tag ID "ABC123" can be bound to "Transportation Tunnel - East Section" through the backend management system, and associated with the geological parameters and historical deformation data of that area, enabling intelligent configuration and management of the monitoring system.

[0086] Through these two input methods, the system can quickly and accurately identify roadway types, avoiding manual configuration errors and ensuring that early warning thresholds and measurement strategies precisely match the actual roadway attributes, thus improving the adaptability and reliability of the monitoring system. For example, when the device is moved from a mining roadway to a transport roadway, it only needs to be placed near the corresponding RFID tag or the new type can be entered via a button; there is no need to readjust the sensor parameters, significantly reducing the difficulty of on-site operation and maintenance costs.

[0087] The self-calibration and fault diagnosis module includes a self-calibration unit and a fault diagnosis unit.

[0088] The self-calibration unit utilizes a built-in high-precision diffuse reflector to periodically calibrate the millimeter-wave radar 3 and laser sensor 4, ensuring long-term measurement accuracy. The calibration procedure is automatically triggered daily at 00:00 by the real-time clock module of PCBA7, and the specific process is as follows:

[0089] Reflector hardware design: The standard reflector is a diffuse reflector, made of polytetrafluoroethylene (PTFE) or a high-reflectivity coating material with a reflectivity ≥95%. The reference distance is strictly calibrated to 2 meters ± 0.1 mm, and the surface roughness Ra ≤ 0.2 μm to ensure uniform and stable reflected signals. The reflector is fixedly installed in the center of the inner wall of the main housing 1, perpendicular to the emission direction of the millimeter-wave radar 3 and the laser sensor 4. A precision mechanical structure ensures that the installation deviation is ≤0.05°, avoiding calibration reference errors caused by installation tilt.

[0090] Calibration signal transmission and error calculation: After calibration is triggered, the millimeter-wave radar 3 transmits a continuous wave signal at a specific frequency, and the laser sensor 4 transmits a pulsed laser signal; both transmit signals synchronously towards the reflector. The measured distance is calculated based on the signal round-trip time.

[0091] Millimeter-wave radar 3 measured distance ,in At the speed of light, This refers to the round-trip time of the signal.

[0092] Laser sensor 4 measured distance ,in This refers to the time of flight of the laser signal.

[0093] Calculate radar measurement error rate Laser measurement error rate .

[0094] Sensor parameter correction: based on the correction formula The parameters of the distance measurement models for millimeter-wave radar 3 and laser sensor 4 are corrected. For example, if the actual measurement error rate of radar is 2% and that of laser is 1%, then in subsequent measurements, radar data needs to be multiplied by 0.98 and laser data by 0.99 before fusion calculation. The corrected parameters are stored in the EEPROM of PCBA7 and are not lost when power is off.

[0095] The fault diagnosis unit assesses equipment status by monitoring key sensor parameters in real time. The specific mechanism is as follows:

[0096] Millimeter-wave radar 3 monitoring: Real-time reading of the radar chip's transmit power register value, with a normal range of 20-25dBm. When the transmit power is <18dBm, signal attenuation may be caused by a power amplifier module malfunction or a loose antenna interface. In this case, a radar fault is determined, generating a fault code "0x01" and storing it in the FLASH memory. Operators can read the code remotely or through a local debugging interface to troubleshoot hardware connections or replace the radar module.

[0097] Laser sensor 4 monitors the offset of the laser spot using a built-in angle sensor. The reference direction of the laser emitting module is aligned with the axis of the main housing 1. When the spot offset is greater than 5°, it indicates that the laser emitting assembly may have undergone mechanical displacement, generating a fault code "0x02". The system simultaneously triggers an audible and visual alarm to indicate "equipment failure" and prevent measurement data from becoming invalid due to sensor pointing deviation.

[0098] Fault code management: The FLASH stores the 50 most recent fault records, including fault codes, occurrence time, sensor status parameters, etc., and supports remote export via 5G communication module for convenient preventive maintenance.

[0099] The PCBA7 integrates a 5G communication module, enabling data transmission and remote control. This module supports NSA / SA dual-mode and operates on mainstream frequency bands including n1 / n3 / n5 / n8 / n41 / n78 / n79, ensuring network connectivity stability in the complex underground mining environment. Data transmission and control functions are implemented as follows:

[0100] Uplink data transmission:

[0101] The data, including distance data, ARIMA model predictions of deformation over the next two hours, and fault codes, are encapsulated via the MQTT protocol. The topic format is " / mine / device / {device ID} / data", where the device ID is a unique 12-digit hexadecimal identifier built into PCBA7.

[0102] Data encryption: The TLS / DTLS protocol is used to encrypt transmitted data to prevent data eavesdropping in the mining network and ensure the security of monitoring data.

[0103] Downlink control commands:

[0104] Remote parameter adjustment: Supports cloud server to send commands via MQTT to modify the measurement range from 0-10 meters to 0-30 meters, the acquisition frequency from 60 seconds / time to 10 seconds / time, and the tunnel deformation early warning threshold from 50mm to 40mm. The command parsing module is located in the PCBA7 control software, and takes effect immediately upon receiving the command, returning parameter confirmation information.

[0105] Remote calibration and fault reset: Supports manual triggering of calibration procedures, remote sending of the "CALIBRATE" command, or resetting the fault status to clear fault codes in FLASH, facilitating remote maintenance.

[0106] It is important to note that the bubble level 6 is embedded in the top surface of the main housing 1. It employs a high-precision tubular level with a graduation of 0.5 mm / m, meaning that a 0.5 mm tilt per meter corresponds to one bubble offset. During installation, the bubble is centered by adjusting the bracket screws to ensure the sensor's detection direction is perpendicular to the tunnel wall. Actual measurements show that when the device's tilt angle is <0.1°, the distance measurement error is <0.3 mm, meeting the millimeter-level monitoring accuracy requirements.

[0107] 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 millimeter-wave radar-based roadway deformation monitoring device, characterized by, The device comprises a main shell, a rear cover, a millimeter wave radar, a laser, a PCBA and a bubble level, the millimeter wave radar and the laser are arranged in the main shell and face the monitoring area, the PCBA is integrated with a data processing and control module, the side surface of the main shell is provided with a gland for the penetration and sealing of external cables, the bubble level is embedded in the top surface of the main shell, the rear cover is detachably mounted on the rear side of the main shell, the device further comprises a multi-modal data fusion analysis system integrated on the PCBA, a dynamic measurement range intelligent adjustment module, a real-time feedback and prediction control module and a self-calibration and fault diagnosis module; The multi-modal data fusion analysis system fuses the distance data collected by the millimeter wave radar and the distance data collected by the laser through the processor built in the PCBA, and generates fused distance data by combining the measurement advantages of both through the Kalman filtering algorithm, and the fusion process satisfies the state equation and the observation equation , wherein is the state vector at time k, including distance, velocity, is the observation vector, including millimeter wave radar and laser measurement values, , is the state transition matrix, and is the process and observation noise; The dynamic measurement range intelligent adjustment module utilizes the processor to calculate a distance fluctuation coefficient according to initial measurement data, mining progress data and geological condition data through a decision tree algorithm: Wherein, is the n-th measured distance, is the average value of the previous three measurements, which can be calculated according to The millimeter wave radar measurement range is dynamically adjusted. The real-time feedback and predictive control module constructs a deformation prediction model based on the ARIMA model, predicts deformation trends in real time, and adjusts the measurement strategy accordingly. The prediction model expression is: in, Predict the deformation at time t. and These are the autoregressive and moving average coefficients, respectively. Let be the residual at time t. The order of the autoregression determines the number of time steps of past data considered in the autoregressive model. express A data sequence value related to deformation at a given time. The order of the moving average determines the number of time steps of the past residuals considered in the moving average model. The self-calibration and fault diagnosis module periodically calls the built-in 2-meter standard distance reflector to emit signals through the processor, and corrects the formula based on the following formula: Calibration measurement error; wherein is the radar measurement error rate, is the laser measurement error rate, and monitors equipment failure through sensor operating parameters.

2. The millimeter wave radar-based roadway deformation monitoring device according to claim 1, characterized in that, The multi-modal data fusion analysis system comprises a preprocessing unit, a weight distribution unit and a Kalman filter unit, the preprocessing unit can adopt median filtering to remove external impulse noise, For measuring data standard deviation, Gaussian fitting is performed on laser data based on least square method determine the center of the light spot wherein, for the intensity of the light spot, for the light spot center coordinates, for the standard deviation of the light spot, the fitting error is verified by root mean square error RMSE and satisfies RMSE≤0.5mm, the weight distribution unit dynamically distributes weights according to the measured distance, when the distance≤5m, the weight proportion of laser data≥70%, when the distance>5m, the weight proportion of millimeter wave radar data≥60%, the Kalman filter unit iteratively calculates through state equation and observation equation, and outputs the fused distance measurement value.

3. The millimeter wave radar-based roadway deformation monitoring device of claim 1, wherein, The adjustment method of the dynamic measurement range intelligent adjustment module comprises the following steps: S1, obtaining mining progress data and geological condition data through an external RS485 interface, the mining progress data comprising excavation depth and support type, and the geological condition data comprising rock compressive strength and joint density; S2, if the distance fluctuation coefficient of continuous three times of measurement is less than or equal to 5% and the excavation speed is less than or equal to 0.5 m / h, setting the millimeter wave radar measurement range to 0-10 meters; S3, if the fluctuation coefficient is greater than 5% or the excavation speed is greater than 0.5 m / h, automatically expanding the measurement range to 0-30 meters; Wherein the parameter threshold is stored in the EEPROM of the PCBA.

4. The millimeter wave radar-based roadway deformation monitoring device of claim 1, wherein, The real-time feedback and prediction control module comprises a prediction model construction unit, a real-time early warning unit and a measurement strategy adjustment unit, the prediction model construction unit trains an ARIMA model using historical fusion distance data, optimizes the order through BIC criterion, and establishes a functional relationship between deformation and time, the real-time early warning unit drives an external audible and visual alarm when predicting that the deformation of the recovery roadway in the next 2 hours is greater than or equal to 50 mm or the deformation of the transportation roadway is greater than or equal to 80 mm, and the measurement strategy adjustment unit controls the sensor scanning angle to deflect ±15° to focus on the deformation area when the deformation rate is greater than or equal to 10 mm / h, and increases the collection frequency from 60 seconds to 10 seconds.

5. The millimeter wave radar-based roadway deformation monitoring device of claim 1, wherein, The self-calibration and fault diagnosis module comprises a self-calibration unit and a fault diagnosis unit, the self-calibration unit triggers a calibration program at 0 o'clock every day, controls the millimeter wave radar and the laser to emit signals to a 2-meter standard reflector plate built in the main shell, the reflectivity of the reflector plate is greater than or equal to 95%, the reference distance is 2 meters plus or minus 0.1 mm, and the measurement error is calculated according to the reflection time and The fault diagnosis unit can monitor the transmission power of the millimeter wave radar and the spot offset amount of the laser in real time, the normal range of the transmission power of the millimeter wave radar is 20-25 dBm, the normal range of the spot offset amount is plus or minus 3°, if the transmission power is less than 18 dBm or the offset amount is greater than 5°, it is determined that the sensor is faulty and a fault code is generated and stored in the FLASH.

6. The millimeter wave radar-based roadway deformation monitoring device of claim 1, wherein, The PCBA is integrated with a 5G communication module, which can upload the fusion distance data, prediction results and fault codes to the cloud server in MQTT protocol, and can receive remote control instructions to support remote adjustment of the measurement range, collection frequency and early warning threshold parameters.

7. The millimeter wave radar-based roadway deformation monitoring device of claim 2, wherein, The preprocessing unit adopts 5*5 window median filtering for millimeter wave radar data, and adopts least square method for Gaussian fitting of laser data, and the fitting error is verified by root mean square error, which satisfies RMSE≤0.5mm.

8. The millimeter wave radar-based roadway deformation monitoring device of claim 1, wherein, The input parameters of the decision tree algorithm further comprise roadway support material and monitoring period, and the output is the sweep frequency range of the millimeter wave radar, the sweep frequency range of the millimeter wave radar is adjustable between 24GHz-77GHz, the roadway support material comprises anchor rod and anchor cable, and the monitoring period comprises driving period and stable period, wherein the driving period is automatically expanded to 60-77GHz, and the stable period is contracted to 24-30GHz.

9. The millimeter wave radar-based roadway deformation monitoring device of claim 4, wherein, The roadway type of the recovery roadway and the transportation roadway in the real-time early warning unit is read by an RFID tag connected to the PCBA or input by a button, the tag stores the roadway type code, and the system automatically matches the early warning threshold according to the code.

10. The millimeter wave radar-based roadway deformation monitoring device of claim 5, wherein, The standard reflection plate built in the main shell is a high-precision diffuse reflection plate, which is installed at the central inner wall of the main shell, and has a surface roughness Ra≤0.2 μm, thereby ensuring stable reflection of the calibration signal and consistency of the measurement reference.

Images