Shield tunneling machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things
Through artificial intelligence Internet of Things technology, real-time collection and analysis of shield machine data, establishing a settlement prediction model, solving the problem of inefficiency of traditional tunnel monitoring methods, real-time monitoring of shield machine and accurate prediction of tunnel settlement, and improving construction safety Sex and efficiency.
Patent Information
- Application Number
- CN202411083350.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-15
AI Technical Summary
Traditional tunnel monitoring methods are inefficient and cannot achieve real-time monitoring of shield machines and effective prediction of tunnel settlement, resulting in construction progress and safety issues.
Using an artificial intelligence Internet of Things method, the shield machine operation parameters and tunnel settlement data are collected in real time, and data transmission and preprocessing are used using AIoT technology, and a variety of machine learning algorithms, such as linear regression, decision trees and convolutional nerves Network, establish settlement prediction models, adjust operating parameters in real time, and continuously optimize the model.
Real-time monitoring and prediction of shield machine operation and tunnel settlement is realized, construction efficiency and safety are improved, construction risks are reduced, and data analysis is improved.
Smart Images

Figure CN118959010A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of artificial intelligence Internet of Things, and specifically relates to a shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things. Background Art
[0002] In the field of tunnel construction, the shield machine is a commonly used tunnel boring equipment that can efficiently advance and excavate tunnels in underground environments. However, the shield machine may encounter various complex geological conditions during the construction process, leading to tunnel settlement, ground deformation and other problems. The emergence of these problems not only affects the construction progress, but may also cause safety accidents. Therefore, real-time monitoring of the operating parameters of the shield machine and prediction of tunnel settlement have become the key to ensuring construction safety.
[0003] Traditional tunnel monitoring methods mainly rely on manual inspections and periodic measurements, which are inefficient and cannot achieve real-time monitoring. With the development of Internet of Things technology, more and more sensors and monitoring devices are integrated into shield machines and tunnel structures, which can collect a large amount of data in real time. However, how to effectively process and analyze this data and extract valuable information from it remains a challenge.
[0004] In addition, with the development of big data technology, how to mine valuable information from massive data has become a hot topic in current research. In the field of real-time monitoring and settlement prediction of shield machines, by analyzing historical data, potential laws and trends can be discovered, providing a scientific basis for construction decisions.
[0005] The application of IoT technology in the field of tunnel construction provides a hardware foundation for real-time monitoring of shield machine operating parameters and tunnel settlement. By installing various sensors on the shield machine, such as temperature, pressure, speed, torque, thrust, etc., operating parameters can be collected in real time. At the same time, laser rangefinders, inclinometers, GPS and other equipment are installed on the tunnel lining and surface to collect settlement data in real time. These data are transmitted to the data processing center in real time through wired or wireless communication interfaces, providing data support for subsequent data analysis and prediction.
[0006] However, how to extract valuable information from these data and achieve real-time monitoring and prediction remains a challenge. Summary of the invention
[0007] In order to solve the deficiency that the above-mentioned technologies cannot realize real-time monitoring and prediction, the present invention provides a real-time monitoring and settlement prediction method of a shield machine based on artificial intelligence Internet of Things. The core of the method is to use the Internet of Things technology to collect the operating parameters of the shield machine and the tunnel settlement data in real time, and analyze and predict the collected data through artificial intelligence algorithms, so as to realize real-time monitoring and prediction of the shield machine operation and tunnel settlement.
[0008] In order to achieve the above object, the present invention is achieved through the following technical solutions:
[0009] The present invention is a shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things, which specifically includes the following steps:
[0010] Step 1: Collect shield machine operating parameters and tunnel settlement data in real time, and conduct all-round real-time monitoring;
[0011] Step 2: Transmit the operating parameters of the shield machine and the tunnel settlement data collected in step 1 to the data processing center in real time through AIoT technology;
[0012] Step 3: pre-process the data transmitted to the data processing center to provide data support for subsequent settlement prediction;
[0013] Step 4: Use a variety of machine learning algorithms including linear regression and decision tree to establish a settlement prediction model to predict tunnel settlement in real time;
[0014] Step 5: According to the prediction result of step 4, the operating parameters of the shield machine are adjusted in real time to control the ground response;
[0015] Step 6: Continuously optimize and update the settlement prediction model through the feedback mechanism.
[0016] A further improvement of the present invention is that: in step 2, transmitting data to a data processing center in real time by using AIoT technology specifically includes the following steps:
[0017] Step 2.1: Data perception: Collect data using signal sampling:
[0018] x[g]=x(t)*δ(t-gT)
[0019] Where x[g] represents the signal value at the g-th sampling point, x(t) represents the continuous time signal, δ(t) represents the unit impulse function, and T represents the sampling period;
[0020] Step 2.2, signal processing: Use y[g]=H(x[g]) to perform signal processing on the data in step 2.1, where y[g] represents the processed signal and H represents the signal processing function; such as digital filter, nonlinear transformation, etc.
[0021] Step 2.3, data transmission: compress and encrypt the data processed in step 2.2 and transmit it to the data processing center;
[0022] Step 2.4: The machine learning algorithm predicts the data transmitted to the data processing center in step 2.3. The linear regression formula for machine learning is as follows: y=wx+b, where y represents the predicted value, x represents the input feature, w represents the weight, and b represents the bias, which is used to determine the optimal weight and bias to accurately predict unknown data.
[0023] Step 2.5: Use a convolutional neural network to perform deep learning processing on the data processed in step 2.4. The formula of the convolutional neural network is: y = f(Gx+b), where G represents weight, b represents bias, and f represents activation function, such as Sigmoid, ReLU, etc. Input x and get y, that is, input the data processed in step 2.4, and after deep learning processing, get the result of data processing;
[0024] Step 2.6: Use the PID controller to perform negative feedback on the deep learning processed data obtained in step 2.5 to obtain the adjusted control output data: adjust the control strategy according to the control output u(t) and error e(t), as well as the proportional Kp, integral Ki, and differential Kd coefficients. The PID controller formula is as follows:
[0025] u(t)=Kp*e(t)+Ki*∫e(t)dt+Kd*de(t) / dt
[0026] By comparing the difference between the actual output of the system and the expected output, i.e. the error e(t), and using the three control functions of proportional p, integral Ki and differential Kd to adjust the control output u(t), the purpose of reducing the error and stabilizing the system is achieved;
[0027] Step 2.7: Use the decision tree model for the adjusted control output obtained in step 2.6, and obtain the decision result D(x) through the feature probability P(x|Ck) and category probability P(Ck) in the adjusted control output data in step 2.6:
[0028] D(x)=argmax(P(x|Ck)*P(Ck))
[0029] Where D(x) represents the decision result, P(x|Ck) represents the probability of feature x given category Ck, and P(Ck) represents the probability of category Ck. Intelligent decision-making is added to improve the automation and intelligence level of the system. Through the decision tree algorithm, data is analyzed and processed to make more accurate and adaptable decisions. Based on the real-time collected data and the preset model, it can independently judge and select the best action plan, and can adapt to the complex and changing actual environment.
[0030] AIoT technology realizes real-time data collection and transmission through the above steps, ensuring that the data processing center can receive the latest shield machine operating parameters and tunnel settlement data in real time, so as to make quick responses and decisions.
[0031] A further improvement of the present invention is that the data transmission in step 2.3 specifically includes the following steps:
[0032] Step 2.3.1, data compression: using a coding formula such as Huffman coding, assigning shorter coding lengths to characters with high frequency of occurrence and longer coding lengths to characters with low frequency of occurrence, in order to reduce the amount of data transmission;
[0033] Step 2.3.2, data encryption: Use the following encryption formula to encrypt the data:
[0034] E(m)=m^e mod N
[0035] Where E(m) represents the encrypted information, m represents the original information, and e and N represent the two components of the public key respectively;
[0036] Step 2.3.3, transmit the encrypted data to the data processing center.
[0037] A further improvement of the present invention is that step 3 pre-processes the data transmitted to the data processing center, specifically comprising the following steps:
[0038] Step 3.1, data cleaning: denoising, outlier processing and incomplete data supplementation;
[0039] Step 3.2, integrate the shield machine parameters and tunnel settlement data, and fuse data from different sources to provide comprehensive prediction information for establishing a settlement prediction model in step 4;
[0040] Step 3.3: Data normalization: normalize the cleaned data in step 3.1 to unify the format and scale;
[0041] Step 3.4: Feature engineering the data processed in step 3.1;
[0042] Step 3.5: Segment the data processed by feature engineering in step 3.4 into a training set, a validation set, and a test set to facilitate the training and evaluation of the settlement prediction model;
[0043] Step 3.6: Perform anomaly detection on the training set, validation set, and test set in step 3.5, and identify potentially important information in the preprocessing stage;
[0044] Step 3.7, perform time series analysis on the data transmitted to the data processing center, and perform smoothing, difference and other processing on the time series data to reveal the time dynamic characteristics;
[0045] Step 3.8: Visualize the data features processed by time series analysis in step 3.7 to help understand the data characteristics;
[0046] Step 3.9: During the pre-processing of the data transmitted to the data processing center, data is stored and managed to ensure the convenience of subsequent data processing.
[0047] A further improvement of the present invention is that the feature engineering of step 3.4 specifically includes the following steps:
[0048] Step 3.4.1, select the features related to the target variable in the settlement prediction model, i.e., the settlement value, and eliminate the features irrelevant to the settlement value to reduce the complexity of the settlement prediction model;
[0049] Step 3.4.2, normalize or standardize the numerical features of the data input into the data processing center to eliminate the dimensional effect and improve the convergence speed of the settlement prediction model;
[0050] Step 3.4.3, encode the categorical features obtained by encoding the non-numeric data input to the data processing center, such as one-hot encoding, label encoding, etc., so that the settlement prediction model can process non-numeric data;
[0051] Step 3.4.4: Use correlation analysis, principal component analysis (PCA) and other methods to reduce the dimension of the numerical features in step 3.4.2 and the categorical features in step 3.4.3 to reduce the redundant information between the features;
[0052] Step 3.4.5, create new features, namely polynomial features and interactive features, to enhance the expression ability of the settlement prediction model;
[0053] Step 3.4.6, use feature selection methods (such as recursive feature elimination, model-based feature selection, etc.) to screen out the features that contribute most to the prediction ability of the settlement prediction model;
[0054] Step 3.4.7: Adjust the weights of numerical features, categorical features, polynomial features, and interactive features, such as using the feature importance score of a tree-based model (such as decision tree, random forest, etc.) to assign different weights to different features.
[0055] A further improvement of the present invention is that in step 3.9, during the process of preprocessing the data transmitted to the data processing center and storing and managing the data, a real-time monitoring and early warning system is constructed, and the real-time monitoring and early warning system includes:
[0056] Data acquisition module X: responsible for real-time collection of shield machine construction operating parameters: propulsion speed, earth pressure, shield thrust, providing basic data for analysis; shield machine operating parameters and tunnel settlement data
[0057] Data processing and analysis module Y: Use machine learning technology to process and analyze the collected shield machine construction data, identify abnormal situations through algorithms, and provide judgment basis for early warning;
[0058] Early warning module Z: According to the preset early warning mechanism, when the data processing and analysis module Y detects an abnormal situation, it will promptly issue an audible and visual alarm or SMS notification early warning information;
[0059] Control Center A: Responsible for the operation monitoring of the entire real-time monitoring and early warning system, including the control of each module, the distribution of early warning information and the feedback of processing results;
[0060] Communication interface B: realizes data transmission and communication between data acquisition module X, data processing and analysis module Y and early warning module Z to ensure the real-time and accuracy of information;
[0061] User interface C: provides operators with an intuitive monitoring interface that displays real-time data, warning information, and historical records, allowing construction personnel to take timely measures;
[0062] Among them, the data acquisition module X is connected to the shield machine through sensors to obtain construction data in real time; the data processing and analysis module Y uses machine learning technology to analyze the data in real time and connects with the early warning module Z; the early warning module Z is connected to the control center A through the communication interface B to achieve real-time transmission of early warning information; the control center A is responsible for monitoring the operating status of each module and interacting with the construction personnel through the user interface C. Through the above collaborative work, it is ensured that the early warning system plays a role in real-time monitoring and early warning during the construction of the shield machine, improving construction safety and efficiency.
[0063] A further improvement of the present invention is that in step 4, establishing a settlement prediction model using a machine learning algorithm specifically includes the following steps:
[0064] Step 4.1: Collect the data after feature engineering processing in step 3.4, establish a linear regression model, and use the linear regression model to analyze the relationship between the sedimentation value and numerical features, categorical features, polynomial features, and interactive features, expressed as:
[0065] Y=aX1+bX2+cX3+......zX n
[0066] Where Y represents the sedimentation value, X1~X nrepresents n features, where n is an integer greater than or equal to 1 and less than or equal to 4, and the least squares method is used to estimate the parameters a, b, c and z for prediction;
[0067] Step 4.2: The decision tree model predicts the sedimentation value in step 4.1 through feature learning. Assume that T is a decision tree, which divides the multidimensional space composed of one, two, three or four features of numerical features, categorical features, polynomial features and interactive features into M regions R1~R m , and there is a corresponding output value C in each area m ,The learning goal of the decision tree model is to minimize the loss function L(T), that is, to find an optimal decision tree model to predict the settlement value;
[0068] Step 4.3: Use the cross-validation method to evaluate the prediction performance of the decision tree model.
[0069] Step 4.4: Compare the performance indicators of different models, such as mean square error (MSE), coefficient of determination (R 2 ) etc., and select the model with the best performance;
[0070] Step 4.5: The model outputs settlement prediction values in real time based on input data, providing a basis for the construction and monitoring of tunnel projects.
[0071] A further improvement of the present invention is that the cross-validation method used in step 4.3 refers to a 5-fold cross-validation method, which specifically includes the following steps:
[0072] Step 4.3.1: Randomly divide the dataset processed by feature engineering in step 3.4 into 5 subsets of equal size, denoted as D1, D2, D3, D4, and D5, and perform 5 iterations;
[0073] Step 4.3.2, in the first iteration, use D1 as the test set, D2, D3, D4, and D5 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D1. In the second iteration, use D2 as the test set, D1, D3, D4, and D5 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D2. This is repeated until the fifth iteration, when D5 is used as the test set, D1, D2, D3, and D4 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D5.
[0074] Step 4.3.3. Calculate the average of the five evaluation results as the overall performance index of the decision tree model. Select the decision tree model with the best performance based on the overall performance index.
[0075] A further improvement of the present invention is that step 6 continuously optimizes and updates the settlement prediction model through a feedback mechanism, specifically comprising the following steps:
[0076] Step 6.1, retrain the settlement prediction model using the new shield mechanism parameters and settlement data, so that the settlement prediction model can capture the latest trends of geological changes and environmental factors, operating conditions, material properties, and equipment status;
[0077] Step 6.2, adjust the parameters of the settlement prediction model and find better parameter settings through the gradient descent algorithm optimization method;
[0078] Step 6.3: Apply ensemble learning methods or deep learning machine learning techniques to improve the performance of the settlement prediction model;
[0079] Step 6.4, after each data collection, the predicted results of the settlement prediction model are compared with the actual settlement data, and the prediction error is calculated;
[0080] Step 6.5: Regularly perform the 5-fold cross-validation method in step 4.3 to verify the quality of the settlement prediction model and ensure the fitness of the settlement prediction model.
[0081] The shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things of the present invention is applied in the field of tunnel construction.
[0082] The beneficial effects of the present invention are:
[0083] 1. Improve construction efficiency and quality
[0084] First, through real-time monitoring and Internet of Things technology, real-time transmission and analysis of construction data are realized, which improves the grasp of the construction status of the shield machine, enables construction personnel to obtain immediate guidance, and thus improves construction efficiency. Secondly, artificial intelligence algorithms are used for geological data analysis and intelligent path planning, optimize construction paths, reduce geological challenges, and further increase construction efficiency. Furthermore, an artificial intelligence control system is introduced to realize automatic adjustment of key construction parameters, reduce manual intervention, and reduce operational risks. In addition, by constructing a settlement prediction model, the risk of ground settlement can be predicted and warned in advance, so that the construction team can take timely measures to control and alleviate settlement and ensure construction quality. Finally, decision support is provided based on the prediction results to help construction personnel optimize construction plans and improve the quality and safety of construction. In general, the present invention greatly improves construction efficiency and quality and reduces risks through data-driven intelligent means.
[0085] 2. Reduce construction risks
[0086] The present invention effectively reduces the risk of shield machine construction by integrating advanced data analysis and intelligent control technology. Specifically, the following measures improve construction safety:
[0087] First, real-time monitoring and data analysis are achieved. Using the Internet of Things technology, the shield machine operation status and construction environment data are collected in real time, and analyzed through artificial intelligence technologies such as data mining, time series analysis, prediction models and natural language processing to accurately identify anomalies and predict potential risks.
[0088] Secondly, the present invention can predict subsidence and provide risk warnings. Combining geological conditions, construction parameters and monitoring data, the artificial intelligence model is used to predict surface subsidence, and automatically warn in high-risk situations to guide construction adjustments.
[0089] Furthermore, intelligent control is achieved. According to real-time data and prediction results, tunneling parameters are automatically adjusted to ensure ground stability and construction efficiency, and reduce human operational errors.
[0090] Finally, remote monitoring and collaboration functions were introduced. AIoT technology supports remote monitoring and expert collaboration, improves construction transparency, ensures rapid response to complex issues, optimizes resource allocation, and improves project efficiency and safety.
[0091] 3. Accurate data analysis and prediction
[0092] The core of the present invention is to achieve accurate analysis and prediction of engineering data through the deep integration of Internet of Things technology and advanced artificial intelligence algorithms. Specifically, the invention is first committed to the collection of real-time data. Relying on the advanced architecture of the Internet of Things, the system continuously monitors the key operating parameters of the shield machine, such as excavation rate, earth pressure balance status, tool wear degree, etc., and simultaneously captures environmental information such as the geological conditions of the construction site and groundwater distribution. These key data are quickly and securely transmitted to the central integrated control system via a high-density sensor network, laying a solid data foundation for subsequent analysis.
[0093] In the data preprocessing stage, the present invention strictly cleans the collected raw data to remove noise and erroneous information. Subsequently, through the standardization and normalization process, the comparability of various types of data is ensured, thereby providing a reliable premise for data feature extraction and selection. This step aims to further optimize the accuracy of the prediction model.
[0094] Furthermore, the present invention constructs an artificial intelligence model based on deep learning and neural network technology to achieve the prediction function of the shield machine operation status and surface settlement. These models use pre-processed data as input, and gradually improve the prediction ability by continuously learning the hidden laws and patterns behind the data. In terms of model selection, the present invention is eclectic and comprehensively compares multiple algorithms including BP neural network and RBF neural network to ensure that the most suitable model for the actual application scenario is selected.
[0095] In the model training and optimization phase, the present invention uses historical data to conduct detailed model training, and iteratively adjusts model parameters and structures to pursue higher prediction accuracy. At the same time, cross-validation, regularization and other technical means are used to effectively avoid overfitting problems and enhance the generalization ability of the model. In addition, the model can achieve online learning and updating based on real-time data to adapt to changes in construction environment and conditions.
[0096] In terms of real-time prediction and analysis, the present invention realizes the rapid prediction of shield machine operation and settlement by inputting real-time data into a mature prediction model. The prediction results can be presented in real time on the monitoring interface, providing instant decision support for construction site personnel and management personnel. Through in-depth analysis of the prediction results, potential risks and hidden dangers can be discovered in a timely manner, providing a scientific basis for engineering decision-making.
[0097] In terms of visualization and decision support, the present invention uses visualization tools such as graphics and charts to make real-time data and forecast results clear at a glance, greatly improving the readability and understanding of information. At the same time, the system also has a decision support function, which can automatically recommend optimization plans or suggestions for adjusting parameters based on the forecast results. In addition, it supports remote monitoring and collaborative operations, which facilitates resource optimization configuration and scheduling.
[0098] Finally, the present invention also provides comprehensive protection in terms of security and reliability. Encryption and protection measures are used to ensure the security of data collection and transmission, avoiding the risk of data leakage and tampering. Robustness tests are conducted on the model to ensure its stability and reliability under extreme conditions. At the same time, a complete data backup and recovery mechanism is established to effectively prevent the impact of data loss on the accuracy of prediction and analysis.
[0099] 4. Intelligent management and optimized resource utilization
[0100] The present invention uses AIoT technology to collect shield machine operating status and construction environment data in real time, build an intelligent early warning system, realize remote monitoring and collaboration, and ensure safe and efficient construction. The prediction model is used to perform predictive maintenance of key components and dynamic analysis of construction site resource requirements, and intelligently dispatch resources to avoid waste. At the same time, based on the collected data and prediction results, the construction plan is optimized, and the decision support system is used to provide accurate decision-making suggestions for managers to significantly improve construction efficiency, reduce risks and optimize costs.
[0101] 5. Intelligent construction monitoring and settlement deformation prediction
[0102] The present invention cleans, standardizes and normalizes the raw data, applies artificial intelligence algorithms for intelligent analysis, identifies abnormal conditions and potential risks in shield machine construction, and displays the analysis results in real time through a visual interface. At the same time, the system integrates real-time and historical data to build a settlement prediction model, improves prediction accuracy, and realizes real-time early warning and automatic adjustment of operating parameters. The prediction results are displayed through the monitoring interface to provide decision support for managers, helping them to make quick and accurate decisions and ensure construction safety and quality.
[0103] 6. Improve the intelligence level of shield machines
[0104] The present invention realizes intelligent monitoring, prediction and control of the shield machine by integrating real-time data collection, intelligent data analysis, settlement prediction model and optimized decision support. Artificial intelligence technology is used to monitor and analyze the status of the shield machine in real time, predict the life of key components and potential failures, and issue maintenance reminders in advance or automatically execute maintenance procedures to ensure stable operation of the equipment. At the same time, the accessories and consumables are intelligently managed through the Internet of Things technology, and procurement plans are automatically generated to achieve real-time tracking of inventory and consumption and the timeliness of material supply, greatly improving the intelligent maintenance and maintenance level of the shield machine.
[0105] In summary, combining artificial intelligence and Internet of Things technology, developing a shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things is of great significance for improving the automation level of tunnel construction, ensuring construction safety, and improving construction efficiency. This method can not only realize real-time monitoring of shield machine operating parameters and tunnel settlement, but also predict and analyze through deep learning algorithms, discover potential safety hazards in advance, and provide a scientific basis for construction decisions. BRIEF DESCRIPTION OF THE DRAWINGS
[0106] Figure 1 It is a flow chart of the present invention.
[0107] Figure 2 It is a data processing flow chart of the present invention.
[0108] Figure 3It is a flow chart of data transmission to data storage of the present invention.
[0109] Figure 4 It is a flow chart of testing and operation of the real-time settlement prediction model of the present invention.
[0110] Figure 5 It is a flow chart of updating the real-time settlement model of the present invention. DETAILED DESCRIPTION
[0111] The following will disclose the embodiments of the present invention with drawings. For the purpose of clear description, many practical details will be described together in the following description. However, it should be understood that these practical details should not be used to limit the present invention. That is to say, in some embodiments of the present invention, these practical details are not necessary.
[0112] like Figure 1-5 As shown, the present invention is a shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things, which specifically includes the following steps:
[0113] Step 1: Collect shield machine operating parameters and tunnel settlement data in real time, and conduct all-round real-time monitoring.
[0114] The operating parameters generated by the shield machine during the excavation process, such as thrust, cutter torque, excavation volume, and tunnel settlement monitoring data, are huge and need to be collected and transmitted in real time to ensure timely grasp of the construction status and decision support; in step 1, the operating parameters of the shield machine include temperature, pressure, speed, torque, and thrust; tunnel settlement data include distance readings of each monitoring point at different times (laser rangefinder), tilt angle and direction (inclinometer), three-dimensional spatial coordinate changes of monitoring points (GPS), settlement rate and cumulative settlement, settlement distribution map and trend analysis.
[0115] Real-time collection of shield machine operating parameters and tunnel settlement data specifically includes the following steps:
[0116] Step 1.1: Ensure that the sensor installation location has a wide field of view and is unobstructed;
[0117] Step 1.2: Install and calibrate the equipment to ensure data accuracy;
[0118] Step 1.3, connect the sensor to the data acquisition system and configure the appropriate sampling frequency and range;
[0119] Step 1.4: Set up an automatic data collection system to record data regularly.
[0120] Step 2: Use AIoT technology to transmit the shield machine operating parameters and tunnel settlement data collected in step 1 to the data processing center in real time. Figure 3 The data transmission method in step 2 is:
[0121] a) Wired communication interface, such as Ethernet, serial communication interface;
[0122] b) Wireless communication interfaces, such as Wi-Fi, Bluetooth, LoRa, and NB-IoT.
[0123] The real-time transmission of data to the data processing center through AIoT technology includes the following steps:
[0124] Step 2.1: Data perception: Collect data using signal sampling:
[0125] x[g]=x(t)*δ(t-gT)
[0126] Where x[g] represents the signal value at the g-th sampling point, x(t) represents the continuous time signal, δ(t) represents the unit impulse function, and T represents the sampling period;
[0127] Step 2.2, signal processing: Use y[g]=H(x[g]) to perform signal processing on the data in step 2.1, where y[g] represents the processed signal and H represents the signal processing function, such as digital filter, nonlinear transformation, etc.
[0128] Step 2.3, data transmission: The data processed in step 2.2 is compressed and encrypted and then transmitted to the data processing center. The data transmission specifically includes the following steps:
[0129] Step 2.3.1, data compression: using a coding formula such as Huffman coding, assigning shorter coding lengths to characters with high frequency of occurrence and longer coding lengths to characters with low frequency of occurrence, in order to reduce the amount of data transmission;
[0130] Step 2.3.2, data encryption: Use the following encryption formula to encrypt the data:
[0131] E(m)=m^e mod N
[0132] Where E(m) represents the encrypted information, m represents the original information, and e and N represent the two components of the public key respectively;
[0133] Step 2.3.3, transmit the encrypted data to the data processing center.
[0134] Step 2.4: The machine learning algorithm predicts the data transmitted to the data processing center in step 2.3. The linear regression formula for machine learning is as follows: y=wx+b, where y represents the predicted value, x represents the input feature, w represents the weight, and b represents the bias, which is used to determine the optimal weight and bias to accurately predict unknown data.
[0135] Step 2.5: Use a convolutional neural network to perform deep learning processing on the data processed in step 2.4. The formula of the convolutional neural network is: y = f(Gx+b), where G represents weight, b represents bias, and f represents activation function, such as Sigmoid, ReLU, etc. Input x and get y, that is, input the data processed in step 2.4, and after deep learning processing, get the result of data processing;
[0136] Step 2.6: Use the PID controller to perform negative feedback on the deep learning processed data obtained in step 2.5 to obtain the adjusted control output data: adjust the control strategy according to the control output u(t) and error e(t), as well as the proportional Kp, integral Ki, and differential Kd coefficients. The PID controller formula is as follows:
[0137] u(t)=Kp*e(t)+Ki*∫e(t)dt+Kd*de(t) / dt
[0138] By comparing the difference between the actual output of the system and the expected output, i.e. the error e(t), and using the three control functions of proportional p, integral Ki and differential Kd to adjust the control output u(t), the purpose of reducing the error and stabilizing the system is achieved;
[0139] Step 2.7: Use the decision tree model for the adjusted control output obtained in step 2.6, and obtain the decision result D(x) through the feature probability P(x|Ck) and category probability P(Ck) in the adjusted control output data in step 2.6:
[0140] D(x)=argmax(P(x|Ck)*P(Ck))
[0141] Where D(x) represents the decision result, P(x|Ck) represents the probability of feature x given category Ck, and P(Ck) represents the probability of category Ck. Categories are different groups into which data can be included after analysis by the decision tree algorithm. Intelligent decision making is added to improve the automation and intelligence level of the system. Through the decision tree algorithm, that is, the decision tree model, data is analyzed and processed to make more accurate and adaptable decisions. Based on the real-time collected data and the preset model, it can independently judge and select the best action plan, and can adapt to the complex and changing actual environment.
[0142] AIoT technology realizes real-time data collection and transmission through the above steps, ensuring that the data processing center can receive the latest shield machine operating parameters and tunnel settlement data in real time, so as to make quick responses and decisions.
[0143] Step 3: Preprocess the data transmitted to the data processing center to provide data support for subsequent settlement prediction.
[0144] The data preprocessing process is mainly dedicated to removing noise, outliers and incomplete data to improve data quality. Next, the data is normalized, such as normalization or standardization, to unify its format and scale. In addition, in the feature engineering step, key features are extracted from the original data to enhance the model's prediction capabilities. In the data integration link, the present invention combines data from different sources, such as integrating shield machine parameters with tunnel settlement data, to provide comprehensive prediction information. In order to simplify the complexity of the model, the present invention adopts data dimensionality reduction methods, such as principal component analysis. At the same time, the data is divided into training sets, validation sets and test sets to facilitate model training and evaluation. The preprocessing stage also includes anomaly detection to identify potentially important information. For time series data, the present invention performs smoothing, differencing and other analyses to reveal temporal dynamic characteristics. Data visualization helps understand the characteristics of the data, and effective data storage and management ensure the convenience of subsequent data processing. Specifically includes the following steps:
[0145] Step 3.1, data cleaning: denoising, outlier processing and incomplete data supplementation;
[0146] Step 3.2: Integrate the shield machine parameters with the tunnel settlement data and fuse data from different sources to provide comprehensive prediction information for establishing a settlement prediction model in step 4.
[0147] The present invention uses the Kalman filter algorithm to integrate the data collected by different sensors, which specifically includes the following contents:
[0148] 3.2.1. Prediction:
[0149] Predict the state at the next moment: Prediction covariance: in, is the state estimate, F is the state transfer matrix, B is the control matrix, u is the control vector, P is the estimated covariance, and Q is the process noise covariance.
[0150] 3.2.2 Update:
[0151] Calculate the Kalman gain: K k =P k∣k-1 H k T(H k P k∣k-1 H k T +R k ) -1
[0152] Update the state estimate: Update covariance: P k∣k =(IK k H k ) k∣k-1
[0153] Where K is the Kalman gain, H is the observation matrix, z is the observation vector, R is the observation noise covariance, and I is the identity matrix.
[0154] The specific mechanism is as follows:
[0155] The raw sensor data is preprocessed using the Karl Walk filter algorithm to remove noise and outliers, and then multi-sensor data fusion technology is used to integrate different types of sensor data to obtain more comprehensive and accurate information. The machine learning algorithm is then used to analyze the data in real time to monitor the status of the shield machine and predict potential problems that may arise.
[0156] Compared with the existing technology, the present invention has the advantage of using data fusion technology, which significantly improves the monitoring accuracy of the shield machine, ensures the real-time and predictive operation, and can be flexibly adjusted according to specific needs and environment to adapt to different construction scenarios, thereby improving safety and reliability and reducing the failure rate.
[0157] Step 3.3: Data normalization: normalize the cleaned data in step 3.1 to unify the format and scale;
[0158] Step 3.4: Feature engineering processes the data processed in step 3.1. The specific feature engineering processes include the following steps:
[0159] Step 3.4.1, select the features related to the target variable in the settlement prediction model, i.e., the settlement value, and eliminate the features irrelevant to the settlement value to reduce the complexity of the settlement prediction model;
[0160] Step 3.4.2, normalize or standardize the numerical features of the data input into the data processing center to eliminate the dimensional effect and improve the convergence speed of the settlement prediction model;
[0161] Step 3.4.3, encode the categorical features obtained by encoding the non-numeric data input to the data processing center, such as one-hot encoding, label encoding, etc., so that the settlement prediction model can process non-numeric data;
[0162] Step 3.4.4: Use correlation analysis, principal component analysis (PCA) and other methods to reduce the dimension of the numerical features in step 3.4.2 and the categorical features in step 3.4.3 to reduce the redundant information between the features;
[0163] Step 3.4.5, create new features, namely polynomial features and interactive features, to enhance the expression ability of the settlement prediction model;
[0164] Step 3.4.6, use feature selection methods such as recursive feature elimination, model-based feature selection and other selection methods to screen out the features that contribute most to the prediction ability of the settlement prediction model;
[0165] Step 3.4.7: Adjust the weights of numerical features, categorical features, polynomial features, and interactive features. For example, use tree-based models such as decision trees and random forests to score feature importance and assign different weights to different features.
[0166] Step 3.5: Segment the data processed by feature engineering in step 3.4 into a training set, a validation set, and a test set to facilitate the training and evaluation of the settlement prediction model;
[0167] Step 3.6: Perform anomaly detection on the training set, validation set, and test set in step 3.5, and identify potentially important information in the preprocessing stage;
[0168] Step 3.7, perform time series analysis on the data transmitted to the data processing center, and perform smoothing, difference and other processing on the time series data to reveal the time dynamic characteristics;
[0169] Step 3.8: Visualize the data features processed by time series analysis in step 3.7 to help understand the data characteristics;
[0170] Step 3.9: During the pre-processing of the data transmitted to the data processing center, data is stored and managed to ensure the convenience of subsequent data processing.
[0171] In the process of data storage and management during the pre-processing of the data transmitted to the data processing center, the present invention also constructs a real-time monitoring and early warning system, which includes:
[0172] Data acquisition module X: responsible for real-time collection of shield machine construction operating parameters: propulsion speed, earth pressure, shield thrust, providing basic data for analysis; shield machine operating parameters and tunnel settlement data
[0173] Data processing and analysis module Y: Use machine learning algorithms to process and analyze the collected shield machine construction data, identify abnormal situations through machine learning algorithms, and provide judgment basis for early warning;
[0174] Early warning module Z: According to the preset early warning mechanism, when the data processing and analysis module Y detects an abnormal situation, it will promptly issue an audible and visual alarm or SMS notification early warning information;
[0175] Control Center A: Responsible for the operation monitoring of the entire real-time monitoring and early warning system, including the control of each module, the distribution of early warning information and the feedback of processing results;
[0176] Communication interface B: realizes data transmission and communication between data acquisition module X, data processing and analysis module Y and early warning module Z to ensure the real-time and accuracy of information;
[0177] User interface C: provides operators with an intuitive monitoring interface that displays real-time data, warning information, and historical records, allowing construction personnel to take timely measures;
[0178] Among them, the data acquisition module X is connected to the shield machine through sensors to obtain construction data in real time; the data processing and analysis module Y uses machine learning technology to analyze the data in real time and connects with the early warning module Z; the early warning module Z is connected to the control center A through the communication interface B to achieve real-time transmission of early warning information; the control center A is responsible for monitoring the operating status of each module and interacting with the construction personnel through the user interface C. Through the above collaborative work, it is ensured that the early warning system plays a role in real-time monitoring and early warning during the construction of the shield machine, improving construction safety and efficiency.
[0179] Step 4: Use a variety of machine learning algorithms, including linear regression and decision trees, to establish a settlement prediction model to predict tunnel settlement in real time. Machine learning algorithms can automatically learn features and patterns from historical data without the need to manually set complex rules. By analyzing large amounts of data, these models can capture nonlinear relationships and complex patterns in the data and achieve highly accurate settlement predictions. The specific steps include the following:
[0180] Step 4.1: Collect the data after feature engineering processing in step 3.4, establish a linear regression model, and use the linear regression model to analyze the relationship between the sedimentation value and numerical features, categorical features, polynomial features, and interactive features, expressed as:
[0181] Y=aX1+bX2+cX3+......zX n
[0182] Where Y represents the sedimentation value, X1~X n represents n features, where n is an integer greater than or equal to 1 and less than or equal to 4, and the least squares method is used to estimate the parameters a, b, c and z for prediction;
[0183] Step 4.2: The decision tree model predicts the sedimentation value in step 4.1 through feature learning. Assume that T is a decision tree, which divides the multidimensional space composed of one, two, three or four features of numerical features, categorical features, polynomial features and interactive features into M regions R1~R m , and there is a corresponding output value C in each area m , the learning goal of the decision tree model is to minimize the loss function L(T), that is, to find an optimal decision tree model to predict the settlement value; minimizing the loss function L(T) is expressed as:
[0184]
[0185] Among them, |T| is the number of leaf nodes of decision tree T, t is the leaf node of decision tree T, and there are N leaf nodes. t There are N sample points in the m class. tm , m=1,2,...M, H t (T) is the information entropy of the leaf node, as a parameter.
[0186] Step 4.3: Use the cross-validation method to evaluate the prediction performance of the decision tree model.
[0187] The cross-validation method used in step 4.3 refers to a 5-fold cross-validation method, which specifically includes the following steps:
[0188] Step 4.3.1: Randomly divide the dataset processed by feature engineering in step 3.4 into 5 subsets of equal size, denoted as D1, D2, D3, D4, and D5, and perform 5 iterations;
[0189] Step 4.3.2, in the first iteration, use D1 as the test set, D2, D3, D4, and D5 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D1. In the second iteration, use D2 as the test set, D1, D3, D4, and D5 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D2. This is repeated until the fifth iteration, when D5 is used as the test set, D1, D2, D3, and D4 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D5.
[0190] Step 4.3.3. Calculate the average of the five evaluation results as the overall performance index of the decision tree model. Select the decision tree model with the best performance based on the overall performance index.
[0191] Step 4.4: Compare the performance indicators of different models, such as mean square error (MSE), coefficient of determination (R 2 ) etc., and select the model with the best performance;
[0192] Step 4.5: The model outputs settlement prediction values in real time based on input data, providing a basis for the construction and monitoring of tunnel projects.
[0193] Step 5: According to the prediction results of step 4, the operating parameters of the shield machine are adjusted in real time to control the ground response.
[0194] Preventive maintenance plays a core role in the strategy of proactively preventing ground subsidence. By proactively predicting possible ground subsidence problems, the present invention can take corresponding preventive measures in a timely manner, thereby avoiding or significantly reducing the adverse effects on the tunnel structure and its surrounding environment. At the same time, the real-time settlement data provided by the prediction model can be quickly fed back to the shield machine operators, allowing them to accurately adjust the operating parameters in a timely manner based on these prediction results.
[0195] In terms of improving safety, by precisely controlling the ground response, potential damage to underground pipelines, buildings and roads can be effectively reduced, thereby improving the safety of the entire construction process. The prediction results provide operators with solid decision support, enabling them to optimize the construction process based on data-driven decisions.
[0196] Step 5: adjusting the operating parameters of the shield machine in real time according to the prediction results of step 4, specifically including the following steps:
[0197] Step 5.1: Preliminarily set the operating parameters based on geological conditions, equipment performance and settlement monitoring data;
[0198] Step 5.2: During the advancement of the shield machine, continuously collect data such as the cutter head speed, torque and advancement speed, and compare them with the predicted results;
[0199] Step 5.3: When the predicted result deviates from the actual settlement, timely adjust the operating parameters. Specifically, if the actual settlement is less than expected, appropriately increase the propulsion speed and cutter head speed, and pay attention to controlling the torque within a safe range;
[0200] If the actual settlement is greater than expected, reduce the propulsion speed and cutterhead speed, and adjust the torque as needed to reduce disturbance to the formation;
[0201] Step 5.4: During the adjustment process, closely observe the impact of parameter changes on settlement, and gradually correct the parameters to make the prediction consistent with the actual settlement.
[0202] Step 6: Continuously optimize and update the settlement prediction model through the feedback mechanism.
[0203] The feedback mechanism enables the prediction model to continuously self-learn and improve, adapting to geological changes and other influencing factors during the construction process. This closed-loop system updates the model in real time, reduces prediction bias, enhances robustness, ensures that the model is based on the latest data prediction, and optimizes shield machine operations by comparing the predicted and actual settlement differences. This dynamic adjustment and data-driven optimization helps to continuously monitor the construction process, improve efficiency, and reduce delays and costs caused by settlement problems.
[0204] The continuous optimization and updating of the settlement prediction model through the feedback mechanism includes the following steps:
[0205] Step 6.1, retrain the settlement prediction model using the new shield mechanism parameters and settlement data, so that the settlement prediction model can capture the latest trends of geological changes and environmental factors, operating conditions, material properties, and equipment status;
[0206] Step 6.2, adjust the parameters of the settlement prediction model and find better parameter settings through the gradient descent algorithm optimization method;
[0207] Step 6.3: Apply ensemble learning methods or deep learning machine learning techniques to improve the performance of the settlement prediction model;
[0208] Step 6.4, after each data collection, the predicted results of the settlement prediction model are compared with the actual settlement data, and the prediction error is calculated;
[0209] Step 6.5: Regularly perform the 5-fold cross-validation method in step 4.3 to verify the quality of the settlement prediction model and ensure the fitness of the settlement prediction model.
[0210] According to the size of the gap between each prediction result and the actual result, adjust the model structure to better capture the characteristics of the data. Specifically, increase or decrease the parameters of the model to reduce overfitting or underfitting; use more or more accurate data for training; preprocess the data, such as removing outliers and normalizing; try different algorithms or models; use cross-validation to adjust model parameters; introduce regularization terms to reduce the complexity of the model; optimize the model's hyperparameters; and use ensemble learning methods to improve model performance;
[0211] The shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things of the present invention realizes accurate prediction and optimization of surface settlement during shield tunnel construction by collecting and analyzing construction data in real time, applying artificial intelligence models to predict settlement, optimizing models and adjusting construction parameters in real time. This method not only improves the safety and efficiency of shield tunnel construction, but also provides strong technical support for urban rail transit construction.
[0212] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims
1. A method for real-time monitoring and settlement prediction of shield machines based on artificial intelligence Internet of Things, characterized by: The shield machine real-time monitoring and settlement prediction method specifically includes the following steps: Step 1: Collect shield machine operating parameters and tunnel settlement data in real time, and conduct all-round real-time monitoring; Step 2: Transmit the operating parameters of the shield machine and the tunnel settlement data collected in step 1 to the data processing center in real time through AIoT technology; Step 3: pre-process the data transmitted to the data processing center to provide data support for subsequent settlement prediction; Step 4: Use machine learning algorithms to establish a settlement prediction model to predict tunnel settlement in real time; Step 5: According to the prediction result of step 4, the operating parameters of the shield machine are adjusted in real time to control the ground response; Step 6: Continuously optimize and update the settlement prediction model through the feedback mechanism.
2. According to claim 1, a method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things is characterized by: In step 2, the real-time transmission of data to the data processing center through AIoT technology includes the following steps: Step 2.1: Data perception: Collect data using signal sampling: x[g]=x(t)*δ(t-gT) Where x[g] represents the signal value at the g-th sampling point, x(t) represents the continuous time signal, δ(t) represents the unit impulse function, and T represents the sampling period; Step 2.2, signal processing: use y[g]=H(x[g]) to perform signal processing on the data in step 2.1, where y[g] represents the processed signal and H represents the signal processing function; Step 2.3, data transmission: compress and encrypt the data processed in step 2.2 and transmit it to the data processing center; Step 2.4: The machine learning algorithm predicts the data transmitted to the data processing center in step 2.
3. The linear regression formula for machine learning is as follows: y=wx+b, where y represents the predicted value, x represents the input feature, w represents the weight, and b represents the bias, which is used to determine the optimal weight and bias to accurately predict unknown data. Step 2.5: Use a convolutional neural network to perform deep learning processing on the data processed in step 2.
4. The formula of the convolutional neural network is: y = f(Gx+b), where G represents weight, b represents bias, and f represents activation function. Input x to get y, that is, input the data processed in step 2.4, and after deep learning processing, get the result of data processing; Step 2.6: Use the PID controller to perform negative feedback on the deep learning processed data obtained in step 2.5 to obtain the adjusted control output data: adjust the control strategy according to the control output u(t) and error e(t), as well as the proportional Kp, integral Ki, and differential Kd coefficients. The PID controller formula is as follows: u(t)=Kp*e(t)+Ki*∫e(t)dt+Kd*de(t) / dt By comparing the difference between the actual output of the system and the expected output, i.e. the error e(t), and using the three control functions of proportional p, integral Ki and differential Kd to adjust the control output u(t), the purpose of reducing the error and stabilizing the system is achieved; Step 2.7: Use the decision tree model for the adjusted control output obtained in step 2.6, and obtain the decision result D(x) through the feature probability P(x|Ck) and category probability P(Ck) in the adjusted control output data in step 2.6: D(x)=argmax(P(x|Ck)*P(Ck)) Where D(x) represents the decision result, P(x|Ck) represents the probability of feature x given category Ck, and P(Ck) represents the probability of category Ck.
3. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 2 is characterized in that: The data transmission of step 2.3 specifically includes the following steps: Step 2.3.1, data compression: using a coding formula to assign short coding lengths to characters with high frequency of occurrence and long coding lengths to characters with low frequency of occurrence, so as to reduce the amount of data transmission; Step 2.3.2, data encryption: Use the following encryption formula to encrypt the data: E(m)=m^e mod N Where E(m) represents the encrypted information, m represents the original information, and e and N represent the two components of the public key respectively; Step 2.3.3, transmit the encrypted data to the data processing center.
4. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 1 is characterized in that: Step 3 pre-processes the data transmitted to the data processing center, specifically including the following steps: Step 3.1, data cleaning: denoising, outlier processing and incomplete data supplementation; Step 3.2, integrate the shield machine parameters and tunnel settlement data, and fuse data from different sources to provide comprehensive prediction information for establishing a settlement prediction model in step 4; Step 3.3: Data normalization: normalize the cleaned data in step 3.1 to unify the format and scale; Step 3.4: Feature engineering the data processed in step 3.1; Step 3.5: Segment the data processed by feature engineering in step 3.4 into a training set, a validation set, and a test set to facilitate the training and evaluation of the settlement prediction model; Step 3.6: Perform anomaly detection on the training set, validation set, and test set in step 3.5, and identify potentially important information in the preprocessing stage; Step 3.7, perform time series analysis on the data transmitted to the data processing center, smooth and differentiate the time series data to reveal the time dynamic characteristics; Step 3.8: Visualize the data features processed by time series analysis in step 3.7 to help understand the data characteristics; Step 3.9: During the pre-processing of the data transmitted to the data processing center, data is stored and managed to ensure the convenience of subsequent data processing.
5. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 4 is characterized in that: The feature engineering of step 3.4 specifically includes the following steps: Step 3.4.1, select the features related to the target variable in the settlement prediction model, i.e., the settlement value, and eliminate the features irrelevant to the settlement value to reduce the complexity of the settlement prediction model; Step 3.4.2, normalize the numerical features of the data input into the data processing center to eliminate the dimensional effect and improve the convergence speed of the settlement prediction model; Step 3.4.3, encode the categorical features obtained by encoding and converting the non-numeric data input to the data processing center, so that the settlement prediction model can process non-numeric data; Step 3.4.4: Perform feature dimensionality reduction on the numerical features of step 3.4.2 and the categorical features of step 3.4.3 to reduce redundant information between features; Step 3.4.5, create new features, namely polynomial features and interactive features, to enhance the expression ability of the settlement prediction model; Step 3.4.6: Use feature selection methods to select the features that contribute most to the prediction ability of the settlement prediction model; Step 3.4.7: Adjust the weights of numerical features, categorical features, polynomial features, and interactive features, and assign different weights to different features.
6. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 4 is characterized in that: In step 3.9, during the process of preprocessing the data transmitted to the data processing center, during the process of data storage and management, a real-time monitoring and early warning system is constructed, and the real-time monitoring and early warning system includes: Data acquisition module X: responsible for real-time collection of shield machine construction operating parameters: propulsion speed, earth pressure, shield thrust, providing basic data for analysis; shield machine operating parameters and tunnel settlement data Data processing and analysis module Y: Use machine learning algorithms to process and analyze the collected shield machine construction data, identify abnormal situations through machine learning algorithms, and provide judgment basis for early warning; Early warning module Z: According to the preset early warning mechanism, when the data processing and analysis module Y detects an abnormal situation, it will promptly issue an audible and visual alarm or SMS notification early warning information; Control Center A: Responsible for the operation monitoring of the entire real-time monitoring and early warning system, including the control of each module, the distribution of early warning information and the feedback of processing results; Communication interface B: realizes data transmission and communication between data acquisition module X, data processing and analysis module Y and early warning module Z to ensure the real-time and accuracy of information; User interface C: provides operators with an intuitive monitoring interface that displays real-time data, warning information, and historical records, allowing construction personnel to take timely measures; Among them, the data acquisition module X is connected to the shield machine through sensors to obtain construction data in real time; the data processing and analysis module Y uses machine learning technology to analyze the data in real time and is connected to the early warning module Z; the early warning module Z is connected to the control center A through the communication interface B to realize the real-time transmission of early warning information; the control center A is responsible for monitoring the operating status of each module and interacting with the construction personnel through the user interface C.
7. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 1 is characterized in that: In step 4, using a machine learning algorithm to establish a settlement prediction model specifically includes the following steps: Step 4.1: Collect the data after feature engineering processing in step 3.4, establish a linear regression model, and use the linear regression model to analyze the relationship between the sedimentation value and numerical features, categorical features, polynomial features, and interactive features, expressed as: Y=aX1+bX2+cX3+......zX n Where Y represents the sedimentation value, X1~X n represents n features, where n is an integer greater than or equal to 1 and less than or equal to 4, and the least squares method is used to estimate the parameters a, b, c and z for prediction; Step 4.2: The decision tree model predicts the sedimentation value in step 4.1 through feature learning. Assume that T is a decision tree, which divides the multidimensional space composed of one, two, three or four features of numerical features, categorical features, polynomial features and interactive features into M regions R1~R m , and there is a corresponding output value C in each area m ,The learning goal of the decision tree model is to minimize the loss function L(T), that is, to find an optimal decision tree model to predict the settlement value; Step 4.3: Use the cross-validation method to evaluate the prediction performance of the decision tree model. Step 4.4: Compare the performance indicators of different models and select the model with the best performance; Step 4.5: The model outputs settlement prediction values in real time based on input data, providing a basis for the construction and monitoring of tunnel projects.
8. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 7 is characterized in that: The cross-validation method used in step 4.3 refers to a 5-fold cross-validation method, which specifically includes the following steps: Step 4.3.1: Divide the dataset processed by feature engineering in step 3.4 into 5 subsets of equal size, denoted as D1, D2, D3, D4, and D5, and perform 5 iterations; Step 4.3.2, in the first iteration, use D1 as the test set, D2, D3, D4, and D5 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D1. In the second iteration, use D2 as the test set, D1, D3, D4, and D5 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D2. This is repeated until the fifth iteration, when D5 is used as the test set, D1, D2, D3, and D4 are combined as the training set, the decision tree model is trained, and the performance of the decision tree model is evaluated on D5. Step 4.3.
3. Calculate the average of the five evaluation results as the overall performance index of the decision tree model. Select the decision tree model with the best performance based on the overall performance index.
9. The method for real-time monitoring and settlement prediction of a shield machine based on artificial intelligence Internet of Things according to claim 1 is characterized in that: The step 6 continuously optimizes and updates the settlement prediction model through a feedback mechanism, and specifically includes the following steps: Step 6.1, retrain the settlement prediction model using the new shield mechanism parameters and settlement data, so that the settlement prediction model can capture the latest trends of geological changes and environmental factors, operating conditions, material properties, and equipment status; Step 6.2, adjust the parameters of the settlement prediction model and find better parameter settings through the gradient descent algorithm optimization method; Step 6.3: Apply ensemble learning methods or deep learning machine learning techniques to improve the performance of the settlement prediction model; Step 6.4, after each data collection, the predicted results of the settlement prediction model are compared with the actual settlement data, and the prediction error is calculated; Step 6.5: Regularly perform the 5-fold cross-validation method in step 4.3 to verify the quality of the settlement prediction model and ensure the fitness of the settlement prediction model.
10. Application of the shield machine real-time monitoring and settlement prediction method based on artificial intelligence Internet of Things according to any one of claims 1 to 9 in the field of tunnel construction.
Citation Information
Cited By
Tunnel construction safety risk management intelligent decision support platform
CN119204703A
Intelligent monitoring method for large-diameter shield tunnel in karst water-rich stratum
CN119538163A
Shield tunneling real-time control method based on random forest and particle swarm optimization algorithm
CN121296135A
Deep reinforcement learning-based intelligent optimization decision-making method for operation parameters of heading machine
CN122197652A
