Intelligent sensing and adaptive estimation system and method for bearing behavior of pressure-type anchor rod and storage medium
By deploying multi-dimensional sensors and adaptive prediction models on the anchor bolts, the problem of the single dimension in pressure anchor bolt monitoring technology has been solved, enabling real-time and accurate prediction of bearing capacity, thus improving engineering safety and construction efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEFEI UNIV OF TECH
- Filing Date
- 2026-02-28
- Publication Date
- 2026-06-05
AI Technical Summary
Existing pressure-type anchor monitoring technology has a single dimension and cannot predict its bearing capacity in real time and adaptively, resulting in large prediction errors and failing to provide a reliable basis for engineering early warning and decision-making.
The micro-strain, macro-displacement, and attitude signals of the anchor bolt are monitored simultaneously using strain sensors, displacement sensors, and tilt sensors. Combined with a signal processing module and an adaptive prediction model, a machine learning model built using the XGBoost algorithm is used for real-time prediction. The model is then calibrated and optimized through incremental learning and engineering feedback.
It enables real-time and accurate prediction of the load-bearing characteristics of anchor bolts, with a prediction accuracy of 94.8%, significantly reducing the root mean square error of prediction and improving engineering safety and construction efficiency.
Smart Images
Figure CN122153391A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geotechnical engineering anchorage structure monitoring technology, specifically to an intelligent sensing and adaptive prediction system, method, and storage medium for the bearing characteristics of pressure-type anchor bolts. Background Technology
[0002] In geotechnical engineering, pressure anchors serve as a crucial foundation support and reinforcement component, and their load-bearing characteristics and stability during long-term service directly impact the safety of the overall project. However, accurately monitoring and assessing the health status of such concealed engineering structures remains a challenging technical problem.
[0003] Traditional monitoring methods mainly rely on periodic manual inspections and point-based instrument measurements. This approach not only has a low monitoring frequency and produces discrete and discontinuous data, making it difficult to capture the dynamic changes in the stress state of anchor bolts during critical periods, but also fails to achieve real-time assessment of their remaining bearing capacity. Although some solutions have emerged to improve monitoring levels with the development of sensing technology, such as integrating single stress or strain sensors onto anchor bolts or using fiber optic grating technology to monitor axial stress, these methods often only obtain isolated, one-dimensional physical quantity information.
[0004] In recent years, intelligent monitoring devices capable of simultaneously acquiring axial stress and deformation data have emerged. However, their technical functions are still mainly focused on data acquisition and remote transmission. Essentially, they haven't solved the core problem of how to deeply mine information from multi-source, heterogeneous monitoring data and intelligently and accurately assess the real-time bearing capacity of anchor bolts. Existing prediction methods are mostly based on simplified empirical formulas or static models. These models cannot fully reflect the complexity, time-varying nature, and uncertainty of the interaction between soil / rock and anchor bolts, leading to large prediction errors and failing to provide a reliable basis for engineering early warning and decision-making.
[0005] Therefore, engineering practice urgently needs an innovative technical solution that can break through the limitations of single-dimensional monitoring, integrate multi-source information, and be able to adaptively learn and evolve, thereby enabling real-time, online, and intelligent prediction of the bearing characteristics of pressure anchor bolts, in order to bridge the key technological gap between "data acquisition" and "capability assessment". Summary of the Invention
[0006] The technical problem to be solved by this invention is: how to overcome the problem that existing pressure-type anchor monitoring technology has a single dimension and cannot predict its bearing capacity in real time and adaptively.
[0007] To achieve the above objectives, on the one hand, the present invention provides an intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts, comprising: a sensing unit, a data acquisition unit, a signal processing module, and a user interaction module; The sensing unit includes strain sensors, displacement sensors, and tilt sensors deployed at different mechanical feature points of the anchor bolt; The strain sensor is rigidly connected to the anchoring section of the anchor rod and is used to monitor the micro-strain signal of the anchor rod. The displacement sensor is connected to the support body of the anchor rod and is used to monitor the overall displacement signal of the anchor rod. The tilt sensor is connected to the anchor head of the anchor rod and is used to monitor the tilt angle signal of the anchor rod. The data acquisition unit is electrically connected to each sensor in the sensing unit via a data cable to acquire the raw signals from each sensor. The signal processing module is connected to the data acquisition unit and is used to preprocess the raw signal. It has an adaptive prediction model built in, which is a machine learning model built based on the gradient boosting decision tree ensemble learning algorithm. It is used to output the estimated bearing capacity of the anchor bolt based on the preprocessed data. The user interaction module is communicatively connected to the signal processing module and is used to display the estimated carrying capacity in real time and trigger an early warning when it exceeds a preset threshold.
[0008] As an optional implementation, the signal processing module includes a signal conditioner and a data processor; the signal conditioner is used to perform preprocessing on the original signal, including wavelet thresholding noise reduction; the data processor is used to extract feature parameters from the preprocessed data and input the feature parameters into the adaptive prediction model.
[0009] As an optional implementation, the adaptive prediction model is a machine learning model built based on the XGBoost algorithm.
[0010] As an optional implementation, the input feature parameters of the adaptive prediction model include the peak strain obtained from the strain signal, the displacement obtained from the displacement signal, and the tilt angle obtained from the tilt angle signal.
[0011] As an optional implementation, the adaptive prediction model is configured with an incremental learning mechanism, which is set to start automatically at a preset period, set exponential decay weights for new monitoring data, and perform retraining or additional boosting iteration training on the adaptive prediction model based on the exponential decay weights to update the model parameters.
[0012] As an optional implementation, the adaptive prediction model is also configured to receive external engineering practice feedback data, including manual inspection records or destructive test data, and to convert the engineering practice feedback data into supervised labeled data samples, and to calibrate and retrain the model or perform additional improvement iterative training based on the data samples.
[0013] In a second aspect, the present invention also provides an intelligent sensing and adaptive prediction method for the bearing characteristics of pressure-type anchor bolts, applied to the intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts as described in the first aspect, the method comprising: S1. The original signals of strain, displacement and tilt angle of the anchor are collected synchronously by strain sensors, displacement sensors and tilt angle sensors installed on the anchor. S2. Preprocess the acquired raw signals to obtain preprocessed data; S3. Extract feature parameters characterizing the anchor bolt state from the preprocessed data; S4. Input the feature parameters into the pre-trained adaptive prediction model, calculate and output the predicted bearing capacity of the anchor bolt, wherein the adaptive prediction model is a machine learning model constructed based on the gradient boosting decision tree ensemble learning algorithm; S5. Compare the estimated load-bearing capacity with a preset threshold. If the load exceeds the threshold, generate an early warning message.
[0014] As an optional implementation, the preprocessing in step S2 includes: denoising the original signal using wavelet thresholding.
[0015] As an optional implementation, the feature parameters extracted in step S3 include at least: peak strain, displacement and tilt angle; and the adaptive prediction model in step S4 is a machine learning model built based on the XGBoost algorithm.
[0016] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the method described in the second aspect.
[0017] First, this invention constructs a comprehensive, collaborative sensing system covering "microscopic strain, macroscopic displacement, and overall attitude" by deploying three types of sensors—strain, displacement, and tilt angle—at key mechanical feature points such as the anchoring section, bearing body, and anchor head of the anchor bolt. This overcomes the shortcomings of existing technologies, which suffer from single monitoring dimensions and incomplete information, laying a comprehensive and reliable data foundation for subsequent accurate analysis. Furthermore, this invention introduces an adaptive prediction model based on the XGBoost machine learning algorithm. This model can deeply integrate and learn the complex nonlinear mapping relationship between multi-dimensional feature parameters and the ultimate bearing capacity of the anchor bolt, achieving a leap from traditional data display to intelligent quantitative output of "predicted bearing capacity." This significantly improves the scientific rigor and accuracy of the prediction, as shown in Table 1 below, with a prediction accuracy rate of up to 94.8%.
[0018] Furthermore, the system constructed by this invention is not static. Its built-in incremental learning mechanism and engineering feedback calibration function enable the model to continuously absorb new monitoring data and higher-confidence field feedback information (such as destructive test results) as construction progresses and time goes by, and thus have the "evolution" ability to dynamically adapt to specific engineering conditions and anchor performance degradation. This ensures the reliability and stability of the evaluation accuracy during long-term service.
[0019] Ultimately, these technological advantages converge into a whole, which not only significantly reduces the root mean square error of prediction and provides engineers with high-precision safety warnings, but also guides the optimization of design parameters by gaining insight into the stress state of anchor bolts. This reduces construction risks and blind testing, while improving the overall scientific, economic, and safety level of engineering construction. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.
[0021] Figure 1 This is a schematic diagram of the intelligent sensing unit of the present invention; Figure 2 This is a schematic diagram of the data acquisition and processing module of the present invention; Figure 3 This is a flowchart illustrating the data processing procedure and algorithm of the present invention. Figure 4 This is a flowchart of the adaptive prediction model of the present invention; Figure 5 This is a schematic diagram of the user interface of the present invention; Figure 6 This is a schematic diagram of the overall system architecture of the present invention; In the diagram: 1. Strain sensor; 1a. Strain sensor threaded connector; 2. Displacement sensor; 2a. Displacement sensor threaded connector; 2b. Displacement sensor bolt; 3. Tilt sensor; 3a. Tilt sensor threaded connector; 4. Data cable; 5. Sensor connector; 6. Base; 7. Pad; 8. Anchor; 9. Anchor bolt; 10. Data acquisition unit; 10a. Data acquisition card; 10b. Data receiver; 11. Signal processing module; 11a. Signal conditioner button; 12. User interaction module; 12a. Data processor display screen; 12b. Data processor interaction button; 12c. Data processor warning light one; 12d. Data processor alarm light two; 13. Connecting cable plug. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0023] Example 1: This embodiment provides an intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts. As a comprehensive monitoring platform integrating high-precision sensing technology, signal processing technology and artificial intelligence algorithms, this system aims to solve the technical problems of invisible anchor bolt stress state, fragmented monitoring data and delayed bearing capacity prediction in geotechnical engineering.
[0024] like Figure 1 and Figure 6 As shown, the system mainly consists of four core parts: a sensing unit, a data acquisition unit 10, a signal processing module 11, and a user interaction module 12.
[0025] First, the anchor rod 9, the core load-bearing component, is inserted into the soil and rock mass and fixed by a combination of the pad 7, anchor 8, and base 6. Specifically, the pad 7, anchor 8, and base 6 can be connected by welding or threading to form a rigid anchoring end structure. During anchoring, the bottom surface of the base 6 is in contact with the soil surface, ensuring that the anchor rod 9 can effectively transfer the load to the depth of the soil when subjected to pull-out force. During the system installation preparation phase, the construction personnel need to cut the anchor rod 9 to the appropriate length according to the design drawings.
[0026] The sensing unit includes a strain sensor 1, a displacement sensor 2, and a tilt sensor 3 deployed at different mechanical feature points of the anchor bolt 9. This embodiment uses a "three-in-one" monitoring layout to capture the full-dimensional mechanical information of the anchor bolt.
[0027] like Figure 1 As shown, strain sensor 1 is rigidly connected to the anchoring section of anchor rod 9. Specifically, the anchoring section of anchor rod 9 has a pre-installed strain sensor threaded connector 1a. Strain sensor 1 is tightly screwed onto anchor rod 9 through this threaded connector 1a, ensuring strict coaxiality. This rigid connection and alignment design aims to eliminate slippage errors between the sensor and the measured object, ensuring that strain sensor 1 can sensitively monitor the micro-strain signals generated by the anchor rod under axial force.
[0028] Displacement sensor 2 is connected to the bearing body of anchor bolt 9 and is specifically used to monitor the overall displacement signal of anchor bolt 9, i.e., macroscopic deformation. For example... Figure 1As shown, the displacement sensor 2 can be connected to the anchor rod 9 via a displacement sensor threaded connector 2a, maintaining a coaxial connection. Considering the complexity of the on-site construction environment and the ease of sensor maintenance, the displacement sensor 2 can also be a displacement sensor 2 with a bolt fastening structure. The displacement sensor 2 body is securely locked to the anchor rod 9 by passing the displacement sensor bolt 2b through the matching bolt hole on the displacement sensor 2 fitted onto the anchor rod 9. This design allows technicians to quickly install and remove the sensor without damaging the main structure of the anchor rod.
[0029] An inclination sensor 3, connected to the anchor head of the anchor bolt 9 (i.e., the outermost end), is used to monitor the inclination angle signal of the anchor bolt. Figure 1 As shown, it can also be installed coaxially at the end of the anchor bolt 9 via the tilt sensor threaded connector 3a. The sensor is securely mounted on the end of the anchor bolt 9. The setting of this monitoring point is crucial because if the anchor bolt is subjected to eccentric tension, uneven settlement of the foundation, or potential shear failure during the bearing process, it will often be first manifested in the attitude deflection of the anchor head.
[0030] It should be noted that the specific installation methods of the aforementioned sensors, such as the connection of strain sensor 1, displacement sensor 2, and tilt sensor 3 to anchor rod 9 via specific threaded holes or bolt structures, are merely preferred examples provided to facilitate understanding of the technical solution of this invention, and not the sole limitation on the installation methods. In practical engineering applications, those skilled in the art can flexibly match the corresponding installation methods according to the specific model, dimensions, and on-site construction conditions of the selected sensors. As long as the monitoring points of each sensor meet the preset mechanical characteristic point distribution requirements, and the main axis of the sensor and the axis of anchor rod 9 meet the coaxiality requirements to ensure the accuracy of signal transmission, it is acceptable. For example, for sensors with non-threaded structures, high-strength structural adhesive bonding, special clamping, or embedded encapsulation can be used for fixing; similarly, displacement sensor 2 can also achieve displacement monitoring through non-contact installation methods such as laser beam or electromagnetic induction. This flexibility in adjusting the installation method does not change the essence of this invention, which obtains anchor rod bearing capacity information through multi-dimensional sensing units.
[0031] At the signal transmission and acquisition level, such as Figures 1 to 6 As shown, the strain sensor 1, displacement sensor 2, and tilt sensor 3 are all equipped with sensor connectors 5, and are connected to the data acquisition unit 10 via data cables 4. Given the harsh conditions often encountered in geotechnical engineering sites, such as groundwater corrosion, mud intrusion, and mechanical compaction, the data cables 4 of this system are externally sheathed with high-strength cable protection sleeves. These sleeves are then sealed and fixed to the sensor tails using dedicated cable protection connectors, thus constructing a robust physical protective barrier to ensure the long-term stability of the signal transmission link.
[0032] The data acquisition unit 10, serving as a signal aggregation node, includes a data receiver 10b and a data acquisition card 10a plugged into it. The data acquisition unit 10 is electrically connected to the sensing unit via a data cable 4, synchronously acquiring the analog raw signals from each sensor at a preset sampling frequency (e.g., 10Hz-100Hz) and converting them into digital signals. The data acquisition unit 10 is also connected to the subsequent signal processing module 11 via a data cable 4, and to improve connection reliability, both ends of the cable are equipped with industrial-grade connectors 13.
[0033] The signal processing module 11 is the core of the system's computation; it is connected to the data acquisition unit 10 and is used to receive and process raw data. In terms of hardware configuration, this module can be represented as a signal conditioner with an embedded high-performance processing chip (such as...). Figure 2 As shown in the diagram, its panel has a signal conditioner button 11a for technicians to configure parameters or switch modes. The signal processing module 11 integrates an advanced adaptive prediction model, which is built based on the XGBoost algorithm and can output the predicted bearing capacity of the anchor bolt 9 in real time based on the input feature data.
[0034] The user interaction module 12 is communicatively connected to the signal processing module 11 and is responsible for human-computer interaction and safety warnings. For example... Figure 5 As shown, the module is equipped with a data processor display screen 12a, which can display the stress state, deformation characteristics, and estimated bearing capacity of the anchor bolt in real time in the form of graphs, curves, or numbers. Users can switch display modes via the data processor interaction button 12b. More importantly, the module has a built-in threshold warning mechanism, with a data processor warning light 12c (e.g., yellow) and a data processor alarm light 12d (e.g., red) on the panel. When the monitored data or estimated results exceed the preset safety threshold, the system will immediately trigger an audible and visual alarm to remind engineers to take grouting reinforcement, unloading, or other emergency measures.
[0035] This embodiment, through scientific hardware selection and layout, constructs a full-dimensional monitoring hardware platform covering "microscopic strain-macroscopic displacement-attitude deflection", laying a solid material foundation for subsequent data processing and intelligent evaluation.
[0036] Example 2: This embodiment, based on the hardware system described in Embodiment 1, provides a specific execution flow for an intelligent perception and adaptive prediction method for the bearing characteristics of pressure-type anchor bolts. This method follows a data-driven logic, achieving accurate judgment of the anchor bolt's working characteristics through rigorous signal processing and feature extraction. Please refer to... Figure 3 Data processing flow and algorithm flowchart and Figure 4 To understand the adaptive prediction model flowchart.
[0037] The first step of this method is S1 (synchronous acquisition). After the system starts, the data acquisition unit 10 synchronously acquires the raw signals of strain, displacement, and tilt of the anchor bolt 9 through the strain sensor 1, displacement sensor 2, and tilt sensor 3 deployed on the anchor bolt 9. Synchronization is the key to this step. The high-precision data acquisition card 10a ensures that the signals of the three different physical quantities are strictly aligned on the time axis, thereby reflecting the true stress state of the anchor bolt at the same moment.
[0038] The next step is step S2 (signal preprocessing). Due to various electromagnetic interferences and mechanical vibrations in the field environment, the acquired raw signals are often mixed with noise. The signal conditioner in signal processing module 11 first amplifies and performs hardware filtering on the signal, and then uses a built-in algorithm to preprocess the raw signal. This step can use wavelet threshold denoising to remove noise. This method utilizes the multi-resolution analysis feature of wavelet transform to decompose the signal into wavelet coefficients of different frequency bands. Since noise usually manifests as high-frequency wavelet coefficients with small amplitudes, while useful signals manifest as coefficients with large amplitudes, by setting a reasonable threshold function, coefficients below the threshold are set to zero or contracted, and then inverse wavelet transform is performed, which can effectively filter out background noise, retain the true characteristics of the anchor bolt's force response, and obtain preprocessed data with a high signal-to-noise ratio.
[0039] The process then proceeds to step S3 (feature extraction). To transform the massive time-series data into an input vector understandable to the model, the data processor of signal processing module 11 extracts feature parameters characterizing the anchor bolt's state from the preprocessed data. These feature parameters include at least the strain peak value obtained from the strain signal. (Reflecting the point of maximum force), the displacement amount obtained from the displacement signal (Reflecting cumulative deformation), and the tilt angle value obtained from the tilt angle signal. (Reflecting attitude changes). As a preferred implementation, to more comprehensively describe the dynamic mechanical behavior of the anchor bolt, this system also extracts the strain gradient. and displacement rate Among these parameters, the strain gradient can sensitively reflect the degree of stress concentration, helping to identify local yielding or crack propagation in the anchor rod at an early stage; the displacement rate quantifies the rate of anchor rod deformation, and is of great significance in determining whether the anchor rod has entered the unstable creep stage. Statistical analysis of these characteristic parameters using data analysis tools can preliminarily identify the stress pattern and deformation trend of anchor rod 9.
[0040] The core step is S4 (intelligent computation). The multidimensional feature parameters extracted in step S3 are input into a pre-trained adaptive prediction model. This adaptive prediction model is a machine learning model built based on the XGBoost (Extreme Gradient Boosting) algorithm. The XGBoost algorithm constructs a strong classifier by integrating a large number of weak decision tree classifiers, exhibiting extremely strong nonlinear mapping ability and anti-overfitting ability. In this embodiment, the model's input vector... The output is the estimated bearing capacity of the anchor bolt. The model internally employs a gradient boosting strategy to continuously optimize the tree structure and the weights of the leaf nodes, thereby accurately fitting the complex nonlinear relationship between input features and carrying capacity.
[0041] Finally, there is step S5 (early warning decision). The signal processing module 11 will calculate the estimated load capacity of the output. The data is transmitted to the user interaction module 12. The system compares this estimated value with a preset safety threshold (e.g., 80% of the design load capacity is a warning value, and 100% is an alarm value). If... If the threshold is exceeded, the user interaction module 12 immediately generates an early warning message, lights up the corresponding indicator light and emits a buzzer sound, realizing unattended monitoring around the clock.
[0042] The method in this embodiment ensures data quality through wavelet denoising, uncovers deep mechanical laws through multidimensional feature extraction, and achieves high-precision load-bearing capacity mapping using the XGBoost algorithm, significantly improving the reliability and timeliness of monitoring results.
[0043] Example 3: This embodiment provides a full lifecycle management mechanism for adaptive prediction models in the system, including initial model construction, incremental learning and updating, and a calibration strategy based on engineering feedback. This part addresses the problem of accuracy degradation in traditional models after changes in the engineering environment.
[0044] First, in the model building phase, the adaptive prediction model is constructed based on the gradient boosting decision tree ensemble learning algorithm. In this embodiment, the system uses the XGBoost algorithm to build the initial model. The training process uses the root mean square error (RMSE) as the loss function, and its calculation formula is as follows: ,in For actual load-bearing capacity, To estimate carrying capacity, the system minimizes the RMSE (Real-Time Error) to ensure that the model's predicted values approximate the true values as closely as possible. To determine the model parameter configuration, the system employs a hyperparameter optimization method based on cross-validation to iteratively optimize and select the model. Simultaneously, to prevent overfitting on the training set, an L2 regularization term is introduced into the objective function, and five-fold cross-validation is used to evaluate and select the model, ensuring good generalization ability on unseen data.
[0045] However, geotechnical engineering is a dynamic evolutionary process, and soil properties change with time, precipitation, excavation, and other factors. To adapt to these changes, the adaptive prediction model in this embodiment is configured with an incremental learning mechanism. The system is set to automatically initiate the model update process at a preset cycle (e.g., every 24 hours). At the end of each update cycle, the system collects newly monitored data from that cycle, assigns exponentially decaying weights to the new monitored data, and performs retraining or additional enhancement iterations on the model based on these exponentially decaying weights to update the model parameters. This mechanism enables the model to reduce the influence of historical outdated data and enhance its response to the latest working conditions, thereby achieving dynamic updates of model parameters and maintaining high predictive sensitivity.
[0046] More importantly, this system also constructs a closed-loop calibration system based on engineering practice feedback. The adaptive prediction model is configured to receive external engineering practice feedback data. This feedback data includes, but is not limited to, manual inspection records from on-site technicians and data from destructive tests (pull-out tests) conducted at specific stages. Compared to routine monitoring data, destructive tests obtain the true ultimate bearing capacity of the anchor bolts, which is considered "high-confidence" data. When the system receives such data, it uses it as a strong supervisory signal to perform specialized calibration training on the model. Specifically, the engineering practice feedback data is converted into supervised labeled data samples, and the model is calibrated and retrained or further iteratively trained based on these data samples to finely correct the model parameters using real values. This closed-loop calibration method allows the model to continuously evolve as the project progresses, becoming more accurate with each use.
[0047] To verify the effectiveness of the model and system described in this embodiment, we conducted rigorous comparative experiments. The comparison results are shown in Table 1 below: Table 1: Comparison of Predictive Performance of Different Monitoring Schemes As shown in Table 1 above, the traditional empirical formula method, relying solely on anchor diameter and burial depth for estimation, has an average RMSE as high as 61.6 and an average accuracy of only 72.5%, which is insufficient to meet the needs of precision engineering. Using only strain sensors, based on strain peak value and gradient calculations, the average RMSE is 52.3, and the accuracy is improved to 85.2%. Using only displacement sensors, based on cumulative displacement and velocity calculations, the average RMSE is 48.7, and the accuracy is 86.1%. Using both strain and displacement sensors simultaneously, this is currently a more advanced conventional monitoring method, with an average RMSE reduced to 37.7 and an accuracy of 88.2%. The solution of this invention simultaneously integrates strain, displacement, and tilt angle sensor data, combined with the aforementioned adaptive XGBoost prediction model. Input features include strain peak value, strain gradient, cumulative displacement, displacement rate, and tilt angle change. Experimental results show that the average RMSE of this invention is significantly reduced to 12.4, and the average accuracy reaches 94.8%.
[0048] The above data strongly demonstrates that this invention, through the deep fusion of multi-source heterogeneous information and continuous optimization of the adaptive model, generates a synergistic effect of "1+1+1>3". Compared with traditional solutions, this invention not only reduces prediction errors by more than 30%, but also elevates accuracy to a new level, providing strong technical support for engineering safety.
[0049] Example 4: This embodiment provides a computer-readable storage medium on which a computer program is stored. The storage medium may be a non-volatile memory (such as flash memory, hard disk, optical disk) or a volatile memory (such as random access memory). The computer program is processed by a processor (e.g., integrated into...). Figure 2 When the signal processing module 11 or the microprocessor (DSP or FPGA) in the data acquisition unit 10 is executed, it can realize all the steps of the intelligent perception and adaptive prediction method of the bearing characteristics of the pressure anchor described in Embodiments 2 and 3.
[0050] Specifically, when the processor executes the computer program, it will trigger the following operations in sequence: control the data acquisition device to acquire the raw signal; call the preprocessing algorithm (wavelet threshold denoising) to clean the signal; run the feature extraction algorithm to calculate parameters such as strain peak value, displacement, and tilt angle; load the feature vector into the adaptive prediction model in memory for inference calculation, the adaptive prediction model is constructed based on the gradient boosting decision tree ensemble learning algorithm, and in this embodiment, the XGBoost algorithm is used; output the calculation results to the display interface; and start the incremental learning subroutine when the periodic condition is met, or start the model calibration subroutine when external engineering practice feedback data is received.
[0051] The existence of this embodiment enables the core algorithm logic of the present invention to be independently distributed and deployed in the form of portable software or firmware, which facilitates reuse on different hardware platforms and greatly expands the application flexibility and commercial value of the present invention.
[0052] In summary, this invention constructs a comprehensive monitoring system capable of capturing "microscopic deformation, macroscopic displacement, and attitude deflection" by strategically placing specific sensors at key mechanical feature points on anchor bolts. Based on the prediction results of an adaptive prediction model, the system not only provides real-time early warnings but also guides the optimization of anchor bolt design parameters, improving the scientific rigor and rationality of engineering design. Simultaneously, through precise real-time monitoring and prediction, unnecessary blind testing and destructive experiments are significantly reduced, substantially improving construction efficiency and lowering costs. This system, integrating intelligent sensing, real-time processing, and adaptive evolution, represents the forefront of geotechnical engineering monitoring technology towards intelligence and digitalization.
[0053] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A system for intelligent sensing and adaptive prediction of the bearing characteristics of pressure-type anchor bolts, characterized in that, include: The sensor unit, data acquisition unit (10), signal processing module (11), and user interaction module (12) are included. The sensing unit includes a strain sensor (1), a displacement sensor (2), and an inclination sensor (3) deployed at different mechanical feature points of the anchor rod (9). The strain sensor (1) is rigidly connected to the anchoring section of the anchor rod (9) and is used to monitor the micro-strain signal of the anchor rod; The displacement sensor (2) is connected to the bearing of the anchor rod (9) and is used to monitor the overall displacement signal of the anchor rod; The tilt sensor (3) is connected to the anchor head of the anchor rod (9) and is used to monitor the tilt angle signal of the anchor rod. The data acquisition unit (10) is electrically connected to each sensor in the sensing unit via a data cable (4) to acquire the raw signals of each sensor; The signal processing module (11) is connected to the data acquisition unit (10) and is used to preprocess the original signal. It has an adaptive prediction model built in, which is a machine learning model built based on the gradient boosting decision tree ensemble learning algorithm. It is used to output the estimated bearing capacity of the anchor (9) based on the preprocessed data. The user interaction module (12) is communicatively connected to the signal processing module (11) and is used to display the estimated carrying capacity in real time and trigger an early warning when it exceeds a preset threshold.
2. The intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts according to claim 1, characterized in that: The signal processing module (11) includes a signal conditioner and a data processor; the signal conditioner is used to perform preprocessing on the original signal, including wavelet threshold noise reduction; the data processor is used to extract feature parameters from the preprocessed data and input the feature parameters into the adaptive prediction model.
3. The intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts according to claim 1, characterized in that: The adaptive prediction model is a machine learning model built based on the XGBoost algorithm.
4. The intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts according to claim 3, characterized in that: The input feature parameters of the adaptive prediction model include the strain peak value obtained from the strain signal, the displacement amount obtained from the displacement signal, and the tilt angle value obtained from the tilt angle signal.
5. The intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts according to claim 1, characterized in that: The adaptive prediction model is configured with an incremental learning mechanism, which is set to start automatically at a preset period, and to set exponential decay weights for new monitoring data. Based on the exponential decay weights, the adaptive prediction model is retrained or additional boosting iterative training is performed to update the model parameters.
6. The intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts according to claim 5, characterized in that: The adaptive prediction model is also configured to receive external engineering practice feedback data, including manual inspection records or destructive test data, and to convert the engineering practice feedback data into supervised labeled data samples, and to calibrate and retrain the model or perform additional improvement iterative training based on the data samples.
7. A method for intelligent sensing and adaptive prediction of the bearing characteristics of pressure-type anchor bolts, characterized in that, The method, applied to the intelligent sensing and adaptive prediction system for the bearing characteristics of pressure-type anchor bolts as described in any one of claims 1-6, comprises: S1. The original signals of strain, displacement and tilt angle of the anchor rod are collected synchronously by the strain sensor (1), displacement sensor (2) and tilt sensor (3) installed on the anchor rod (9); S2. Preprocess the acquired raw signals to obtain preprocessed data; S3. Extract feature parameters characterizing the anchor bolt state from the preprocessed data; S4. Input the feature parameters into the pre-trained adaptive prediction model, calculate and output the estimated bearing capacity of the anchor (9), wherein the adaptive prediction model is a machine learning model constructed based on the gradient boosting decision tree ensemble learning algorithm; S5. Compare the estimated load-bearing capacity with a preset threshold. If the load exceeds the threshold, generate an early warning message.
8. The intelligent sensing and adaptive prediction method for the bearing characteristics of pressure-type anchor bolts according to claim 7, characterized in that, The preprocessing in step S2 includes: denoising the original signal using wavelet threshold denoising.
9. The intelligent sensing and adaptive prediction method for the bearing characteristics of pressure-type anchor bolts according to claim 7, characterized in that, The feature parameters extracted in step S3 include at least: peak strain, displacement, and tilt angle; and the adaptive prediction model in step S4 is a machine learning model built based on the XGBoost algorithm.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 7 to 9.