Intelligent early warning system for deep stirring differential settlement in reclamation area

Abstract

The invention belongs to the technical field of foundation treatment, and particularly relates to an intelligent early warning system for deep stirring differential settlement in a reclamation area, which comprises the following steps of: monitoring and acquiring data in a construction process in real time through a sensor mounted on a deep stirring pile, including key parameters such as torque, depth, pressure and the like; analyzing and identifying the collected data based on an intelligent algorithm, extracting stratum features and detecting abnormal conditions; according to an analysis result, construction parameters are regulated and optimized, and the stability of the construction process is ensured; an early warning report and a differential settlement prediction model are generated, and potential risks are recognized in advance; and performing quality evaluation on the processing area, and providing improvement suggestions. Through the application of the system in the scheme, the foundation treatment quality of the reclamation area can be effectively improved, differential settlement is reduced, the safety and economy of construction are ensured, and the system has wide application prospects.

Description

Technical Field

[0001] The present invention belongs to the field of foundation treatment technology, and in particular relates to an intelligent early warning system for uneven settlement during deep mixing pile construction in reclamation areas, which is used to improve the construction quality and safety of pile foundations under complex strata. Background Art

[0002] In recent years, the Deep Cement Mixing Method (DCM) has been widely used in offshore areas of Japan and South Korea for foundation reinforcement and improvement of ultra-thick soft soil layers. This method, with its many advantages, including simple equipment, zero vibration, noise, pollution, no lateral compression of the surrounding soil, short construction period, and low project cost, is suitable for improving and reinforcing nearly all soil types, including soft clay, sand, and organic soil. This technology has been successfully applied in several deep-sea soft soil foundation improvement projects in my country and was also used in the foundation reinforcement and improvement construction of the third runway reclamation project at Hong Kong International Airport.

[0003] However, the foundations of reclaimed areas are often composed of complex strata. Especially when a soft underlying layer is overlain by a relatively hard crust, traditional deep mixing pile construction methods can easily misjudge that the pile foundation has reached its design strength, ignoring the penetration of the pile body after penetrating the hard crust, leading to subsequent uneven settlement. This problem not only seriously affects the quality of the project but can also pose a major safety hazard. Therefore, there is an urgent need for an intelligent system that can monitor and determine the penetration status of the mixing pile after penetrating the hard crust in real time, accurately identify the presence of a weak underlying layer, and adjust construction parameters accordingly to ensure the quality and reliability of the foundation treatment. The introduction of such an intelligent system will effectively improve the effectiveness of foundation treatment, reduce the risk of uneven settlement, and thus ensure the overall stability and safety of the project. Summary of the Invention

[0004] The present invention aims to overcome at least one defect in the prior art and provide an intelligent early warning system for uneven settlement of deep mixing piles in reclamation areas to ensure the quality and reliability of foundation treatment.

[0005] The present invention provides an intelligent early warning system for uneven sedimentation caused by deep stirring in reclamation areas, which is characterized by comprising the following steps:

[0006] S101, performing data monitoring and data collection through sensors to obtain raw data;

[0007] S102, performing data analysis and identification on the original data based on an intelligent algorithm to obtain a data processing result;

[0008] S103, regulating and optimizing the system based on the data processing results;

[0009] S104, generating an early warning report and an uneven settlement prediction model;

[0010] S105. Conduct quality assessment of the processing area and make improvement recommendations.

[0011] In order to improve the monitoring and control accuracy of the mixing pile construction process and ensure construction safety, the sensors in step S101 include at least two or more of a torque sensor, a depth sensor, a pressure sensor, and an acceleration sensor. The data from the torque, pressure, acceleration, and depth sensors are comprehensively used to evaluate the formation information, and multi-source data is fused and analyzed collaboratively.

[0012] The torque sensor is installed on the main shaft of the mixing pile and is used to monitor the torque T (N·m) of the drill bit in real time during the construction process, identify the resistance conditions and hardness changes of different strata, and help identify the hardness changes of the strata and the mixing uniformity;

[0013] The depth sensor is installed on the drill bit to accurately measure the vertical depth of the drill bit to reflect the construction progress in real time, ensure that the mixing pile is constructed according to the designed depth, and identify potential sinking or deviation problems;

[0014] The pressure sensor is installed on the propulsion device of the drill bit and is used to monitor the drill bit pressure P (kPa) in real time to analyze the bearing capacity and density of the foundation and identify changes in the formation;

[0015] The acceleration sensor is installed on the main shaft of the mixing pile and is used to monitor the vibration spectrum a(f)(m / s 2 ), and identify the uniformity of the formation and whether there are obstacles through spectrum analysis (FFT algorithm).

[0016] Through a combination of multiple sensor technologies, the system can obtain key parameters of the construction process in real time and comprehensively, providing construction personnel with accurate construction feedback, thereby optimizing construction strategies and reducing construction risks.

[0017] The data acquisition includes data preprocessing and data transmission. The data preprocessing includes denoising, filtering and data compression operations, which are used to improve transmission efficiency and the accuracy of subsequent data analysis, and reduce data storage space and transmission time, thereby improving data transmission efficiency. The data transmission uses wireless transmission technology and wired transmission technology to achieve efficient and accurate transmission of data from the data source to the data terminal, meeting the data transmission requirements in different application scenarios.

[0018] In order to realize automatic identification of stratum characteristics and real-time evaluation of the construction quality of the mixing pile, the raw data is analyzed and identified based on the intelligent algorithm in step S102 to obtain data processing results, including:

[0019] Analyzing the raw data based on an intelligent algorithm to obtain the data processing result, wherein the intelligent algorithm includes a machine learning algorithm and a deep learning algorithm;

[0020] By comparing historical data processing results with the data processing results, the formation characteristics are automatically identified and the uniformity of the mixing pile construction and the foundation bearing capacity are evaluated in real time. The historical data processing results are obtained by analyzing the historical data based on the intelligent algorithm, which improves the accuracy of data analysis, realizes the automatic identification of formation characteristics, and evaluates the construction quality and foundation bearing capacity in real time.

[0021] Further preferably, the analyzing the raw data based on an intelligent algorithm includes:

[0022] Processing noise and outliers in sensor data, performing standard processing on the raw data to obtain standard data, wherein the standardization processing includes standardization, filtering and noise removal;

[0023] The standard data is input into the intelligent algorithm to output key features, including the dynamic response mode of the stratum during the construction of the mixing pile and the fluctuation characteristics of the torque. and changing trends of pressure.

[0024] Among them, the dynamic response mode of the formation is: based on the torque gradient, pressure change rate and vibration spectrum energy distribution, the formation stiffness coefficient is constructed Real-time mapping of formation compressibility changes;

[0025] Since formation features vary greatly, after outputting the key features, cluster analysis and pattern recognition of formation data can be used to more quickly and accurately identify different formation features. The automatic identification of formation features includes:

[0026] The initial stratum determination is achieved through multi-sensor data fusion. First, the data is normalized and the raw data of the torque, pressure, acceleration and depth sensors are mapped to the interval [0,1]. Then, according to the initial stratum determination results, the stratum is divided into soft soil, sand layer and rock layer, and the contribution weight of each sensor is adjusted. According to the formula S = ω T S T +ω P S P +ω A S A +ω D S D Calculate formation scores;

[0027] Among them, ω T 、ω P 、ω A 、ω DThey are torque weight, pressure weight, acceleration weight and depth weight, is the normalized eigenvalue, i is the sensor name, x i,max x i,max and x i,min is x i,min Historical extreme values or theoretical range boundaries of sensor data;

[0028] Finally, the characteristics of the stratum under construction are determined based on the preset scoring range. For example, S < 0.3 is low risk, 0.3 ≤ S < 0.6 is medium risk, and S ≥ 0.6 is high risk. The weighting rules can be trained through historical data or calibrated through field tests and can be dynamically updated to adapt to complex geological conditions.

[0029] The Gaussian mixture model (GMM) is used to estimate the probability density of the feature space, divide the high-risk data into clusters, and output the stratum classification labels, including soft soil, sand layer, gravel layer, and rock layer;

[0030] Among them, the input feature vector is:

[0031]

[0032] Where S is the formation score, is the torque gradient, ΔP is the pressure change rate, E vib is the vibration energy ratio.

[0033] The GMM model parameters are:

[0034]

[0035] π i is the mixed weight, which represents the prior probability of each layer type in the overall data, satisfying The classification results are as follows: soft soil (π1), sand layer (π2), gravel layer (π3), rock layer (π4); μ i is the mean vector, describing the central position of the characteristic vector of each layer type; Σ i is the covariance matrix, which represents the dispersion and correlation of the characteristics within the same stratum class;

[0036] Cross-validate the above classification results with geological exploration data to ensure that the classification results are consistent with engineering knowledge;

[0037] Through the three-level pattern recognition technology of "initial judgment scoring → probability density classification → geological data cross-validation", risk patterns that may recur during the construction process can be identified, so that measures can be taken in advance to avoid or reduce the occurrence of construction accidents.

[0038] The features most relevant to the risk of differential settlement are mined from the key features, and an accurate prediction model is constructed based on these features. This model can learn the relationship between different stratum characteristics, construction parameters, and differential settlement, thereby accurately predicting the risk of differential settlement before or during construction. The prediction model is trained through the following steps:

[0039] Select key features that affect differential settlement risk, including formation score S, torque gradient Pressure change rate ΔP, vibration energy ratio E vib , GMM classification label L∈{1,2,3,4} is the input feature;

[0040] Based on historical data, a prediction model is constructed, and the relationship between stratum characteristics and uneven settlement is learned through a bidirectional LSTM network to obtain an output prediction model and output the settlement probability P. settlement ∈[0,1];

[0041] The output prediction model is evaluated using a test dataset. The system updates the model input based on real-time data, and the prediction results generated by the model are fed back to the construction parameter adjustment module to dynamically optimize the construction process.

[0042] Due to the complexity of the construction environment, various abnormal situations often occur during the construction process. Through real-time data processing and intelligent control methods, key parameters in the construction process can be accurately regulated and optimized, thereby improving construction quality and efficiency, ensuring the smooth progress and high-quality completion of the construction project. The regulation and optimization system includes:

[0043] Based on the data processing results, after detecting the risk of uneven settlement or construction anomalies, the control unit adjusts key construction parameters such as stirring speed, grouting volume, and drill bit pressure. For example, when the probability of stratum settlement is greater than 0.9, the construction parameters are adjusted to reduce the stirring speed, increase the grouting volume, and control the drill bit pressure to improve the foundation strength. When the probability of stratum settlement is greater than 0.7 and less than 0.9, the stirring intensity and grouting speed are adjusted to avoid excessive disturbance. When the probability of stratum settlement is less than 0.7 and greater than 0.5, the drill bit pressure and stirring intensity are adjusted first to prevent increased wear or drill bit jamming, and the feedback is fed back to the control unit through the database during construction.

[0044] Risk index R = α1ΔS + α2ΔP + α3ΔT + α4ΔA, with weights dynamically adapted according to formation type;

[0045] Among them, ΔS, ΔP, ΔT, and ΔA represent the deviations of depth, pressure, torque, and acceleration, respectively. α1, α2, α3, and α4 are weight coefficients. The settlement risk level is evaluated according to the size of the R value.

[0046] Based on the above results, the construction is guided. For example, the early warning trigger mechanism is:

[0047] Level 1 warning: When R>0.6 or P settlement When the rate is >70%, there will be sound and light alarm, and the speed and injection volume will be automatically fine-tuned;

[0048] Level 2 warning: When R>0.75 or P settlement When the rate is >80%, stop the machine for inspection and perform local high-pressure grouting;

[0049] Level 3 warning: When R>0.9 and P settlement When the load is >90%, shut down for inspection and emergency support;

[0050] Parameters can be set according to specific projects, and the optimized parameters are transmitted back to the construction equipment in real time to ensure continuous monitoring and dynamic adjustment of the construction process. Based on cross-validation, performance indicators are set: accuracy > 90%, recall rate > 85%. If the model does not meet the performance requirements, closed-loop optimization is performed.

[0051] In order to enable the construction team and relevant management personnel to quickly and comprehensively understand the abnormal conditions detected by the system and make timely and accurate decisions, the early warning report in step S104 includes a detailed description of the detected abnormal conditions, risk assessment levels, and relevant data charts. This is intended to help the construction team fully understand the nature, scale, and potential impact of the risk of uneven settlement or construction abnormalities, so that targeted measures can be taken for prevention and control.

[0052] Combined with the above data charts, the system conducts a comprehensive and in-depth assessment of the foundation quality of the treatment area and provides a detailed quality assessment report. The step S105 includes:

[0053] Conduct a comprehensive assessment of the foundation quality of the treatment area and provide a detailed quality assessment report, which includes foundation bearing capacity, uniformity index and construction efficiency;

[0054] The foundation bearing capacity is calculated as follows:

[0055] q u =1.2cN c +γD f N q η 搅拌

[0056] Where, c is soil cohesion (kPa), N c is the foundation bearing capacity coefficient, γ is the soil density (kN / m 3 ), D f is the foundation burial depth (m), N q is the foundation bearing capacity coefficient, η 搅拌is the formation stiffness correction factor (0.8-1.2);

[0057] The uniformity index is:

[0058]

[0059] Among them, σ T , σ P are torque and pressure standard deviation, μ T 、μ P are the mean values of torque and pressure, respectively.

[0060] The construction efficiency is:

[0061]

[0062] Based on the evaluation results, further improvement suggestions were put forward to ensure the high quality and safety of future construction, providing strong guarantees for the smooth progress and high-quality completion of the project.

[0063] Compared with existing technologies, this invention has at least the following advantages: It solves the problem of difficulty in determining whether mixing piles are adequately mixed when crossing weak underlying layers in reclamation areas, as well as the lack of quantitative standards for foundation bearing capacity. This invention improves the quality of foundation treatment, effectively reducing or even preventing uneven settlement, significantly improving construction quality while saving costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0065] Figure 1 This is a flow chart of an intelligent early warning system for deep-layer mixing and uneven settlement in reclamation areas provided by the present invention;

[0066] Figure 2 This is a schematic diagram of the overall structure of an intelligent early warning system for deep-layer mixing and uneven settlement in reclamation areas provided by the present invention;

[0067] Figure 3 This is a schematic diagram of the layout of the mixing pile sensor;

[0068] Figure 4 Schematic diagram of data processing;

[0069] Figure 5 This is a schematic diagram of the system early warning report. DETAILED DESCRIPTION

[0070] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0071] Example: Figure 1 As shown, the present invention provides an intelligent early warning system for uneven settlement of deep-layer mixing in reclamation areas, comprising:

[0072] S101, performing data monitoring and data collection through sensors to obtain raw data;

[0073] S102, performing data analysis and identification on the original data based on an intelligent algorithm to obtain a data processing result;

[0074] S103, regulating and optimizing the system based on the data processing results;

[0075] S104, generating an early warning report and an uneven settlement prediction model;

[0076] S105. Conduct quality assessment of the processing area and make improvement recommendations.

[0077] Specifically, each step is implemented as follows:

[0078] Step S101: By placing sensors on the deep mixing piles, various parameters during the construction of the mixing piles are monitored in real time, and the data collected by the sensors are transmitted to the data processing center through wireless transmission technology combined with wired transmission technology to achieve real-time data monitoring and collection;

[0079] Step S102: Analyze the collected raw data based on intelligent algorithms (including machine learning models, neural networks, etc.), identify possible abnormalities in the construction process, and automatically identify stratum characteristics by comparing historical data with current data, and evaluate the uniformity of the mixing pile construction and the bearing capacity of the foundation in real time;

[0080] Step S103: Based on the data analysis results, the system automatically adjusts the parameters of the pile construction, including mixing speed, grouting volume, drill pressure, and rotation speed, to ensure the uniformity and quality of the foundation treatment. If potential uneven settlement risks are detected, the system immediately adjusts the construction plan and, if necessary, suspends construction to prevent further quality issues.

[0081] In step S104, the system generates an early warning report on uneven settlement based on real-time monitoring data and analysis results, including potential settlement areas, risk levels, and recommended countermeasures. Simultaneously, the system establishes and updates an uneven settlement prediction model, continuously optimizing it based on historical data and construction experience to provide more accurate risk predictions.

[0082] In step S105, after construction is complete, the system conducts a comprehensive assessment of the foundation quality in the treatment area and provides a detailed quality assessment report, including foundation bearing capacity, differential settlement risk, and construction efficiency. Based on the assessment results, further improvement suggestions are made to ensure high quality and safety in future construction.

[0083] In step S101, the specific operations are as follows:

[0084] S11: On the deep mixing pile construction equipment, the sensors are arranged as follows: a torque sensor is installed on the main shaft to monitor the torque of the drill bit in real time during the construction process, reflecting the resistance of different strata, thereby helping to identify the soft and hard changes in the strata and the mixing uniformity; a depth sensor is installed on the drill bit to accurately measure the vertical depth of the drill bit to reflect the construction progress in real time and ensure that the mixing piles are constructed according to the designed depth, while identifying potential sinking or offset problems; a pressure sensor is installed on the propulsion device of the drill bit to monitor the pressure of the drill bit on the stratum in real time, which is used to analyze the bearing capacity of the foundation and identify changes in the stratum; in addition, an acceleration sensor is arranged on the main shaft of the mixing pile to detect the vibration and stability of the equipment to identify possible obstacles such as stratum anomalies or hard layers.

[0085] S12: These sensors continuously and in real time collect various data during the construction process and store this data in on-site edge computing devices. The collection frequency can be adjusted based on construction needs to ensure the real-time and accurate data.

[0086] S13: Before data transmission, the edge computing device will perform preliminary preprocessing on the collected raw data, such as denoising, filtering, and data compression, to improve transmission efficiency and the accuracy of subsequent data analysis.

[0087] S14: The collected data is transmitted to the data processing center in real time via wireless transmission technology (such as 5G, LoRa, or Wi-Fi). To ensure the reliability of data transmission, the system can also be equipped with wired transmission technology to avoid data loss due to network interruption or other factors.

[0088] Through the above steps, step S101 can ensure accurate monitoring and collection of various key data during the construction of deep mixing piles in the reclamation area, laying a solid foundation for subsequent intelligent analysis and regulation.

[0089] In step S102, data analysis and identification specifically include the following steps:

[0090] S21: Data processing involves cleaning the raw data collected by the sensors, removing noise, erroneous data points, or outliers, and improving data quality by removing data that exceeds a reasonable range and smoothing signal fluctuations. At the same time, different types of sensor data are normalized and converted into a standardized numerical range [0,1] to facilitate subsequent unified analysis.

[0091] S22: During the feature extraction process, the system extracts characteristic parameters such as stirring torque fluctuations, drill bit pressure changes, and vibration frequency changes based on construction requirements and key factors affecting construction quality. It also processes real-time data through time series analysis methods such as sliding average and weighted average to identify trends, periodic fluctuations, or sudden anomalies, thereby tracking changes in key parameters during the construction process in real time and determining whether there are abnormal fluctuations.

[0092] S23: During the intelligent algorithm application phase, the system trains machine learning models (such as support vector machines, random forests, and K-nearest neighbor algorithms) based on historical construction data to identify normal and abnormal patterns during construction. Model training fully considers stratum characteristics, construction equipment parameters, and known construction quality issues. During actual construction, the system inputs current monitoring data into the trained model in real time for immediate prediction and identification, and outputs an assessment of the current construction status, including key issues such as whether mixing is sufficient and whether there is a weak underlying layer.

[0093] S24: The anomaly identification process includes two steps: threshold judgment and pattern recognition. First, the system judges the real-time monitoring data based on a pre-set threshold or a dynamic threshold automatically adjusted by historical data. If certain key parameters (such as torque or vibration frequency) exceed or fall below the set threshold, the system will mark the situation as a potential anomaly. At the same time, the system also uses pattern recognition algorithms to match real-time data with historical anomaly data patterns. If the current data pattern is found to be similar to the historical anomaly pattern, the system will immediately identify the situation as an anomaly.

[0094] S25: Once the system detects an anomaly or identifies a potential construction risk (such as the possibility of uneven settlement), it will immediately send an alarm message to the construction control center and generate control suggestions, automatically or prompting the operator to adjust the construction parameters (such as reducing the mixing speed, increasing the grouting volume, etc.) to correct potential problems in construction.

[0095] S26: The system records all real-time monitoring data, analysis results, and treatment measures in a database for subsequent analysis and model optimization. Periodically or after construction is completed, the system uses the latest data to optimize the intelligent algorithm and update the model to improve prediction accuracy and recognition accuracy.

[0096] Through the above implementation method, step S102 can effectively identify abnormal situations that may occur during the construction of deep mixing piles in the reclamation area, providing a scientific basis for subsequent construction control and early warning.

[0097] In step S103, system control and optimization specifically include the following steps:

[0098] S31: Regarding dynamic parameter adjustment, the system monitors mixing torque and depth data in real time to identify the hardness and uniformity of the current formation and automatically adjusts the mixing speed. If the mixing torque is high or the drilling depth changes slowly, the system reduces the mixing speed to increase mixing time and ensure uniform mixing. Conversely, if the torque is low and the depth changes rapidly, the system increases the mixing speed to improve construction efficiency. Furthermore, the system automatically adjusts the grouting volume based on pressure sensor data and mixing torque information. When a weak underlying layer is detected, the grouting volume is increased to strengthen the foundation, while in hard formations, the grouting volume is reduced to avoid material waste and excessive hardening. Furthermore, the system monitors the drill bit's downforce and speed in real time and optimizes the adjustments based on the formation information. In soft formations, the system increases the downforce and slows the speed to ensure adequate mixing and avoid uneven settlement. In hard formations, the pressure and speed are reduced accordingly to avoid excessive wear on the equipment.

[0099] S32: In terms of automatic response to abnormal situations, the system automatically identifies insufficient mixing by comparing real-time torque data with preset standard values and promptly adjusts mixing parameters, such as increasing mixing time or reducing mixing speed. It also issues an alarm to prompt the operator to check. If the system detects rapid penetration of the drill bit through a depth sensor and determines that it may have penetrated a weak underlying layer, it automatically slows the drilling speed, increases the grouting volume, and adjusts the drill bit pressure and speed as needed to ensure construction quality.

[0100] S33: The feedback and optimization loop includes a real-time feedback mechanism and historical data optimization. The system continuously monitors the impact of adjusted construction parameters on the foundation treatment results and feeds the results back to the data analysis module. Through this real-time feedback and data accumulation, intelligent algorithms continuously optimize the construction parameter adjustment strategy, gradually improving foundation treatment results and construction efficiency. Simultaneously, the system records each construction parameter adjustment and results, creating a historical data archive. This historical data is analyzed and used to update and optimize the control strategy, making it more adaptable to different geological conditions and construction environments.

[0101] S34: In the intelligent control system, manual control is combined with intelligent optimization. When dealing with special situations or complex geological conditions, although the system can automatically adjust construction parameters, it still allows the operator to manually intervene and make adjustments. The system provides auxiliary decision support during this process, helping the operator to develop the optimal adjustment plan. After the adjustment is completed, it continues to automatically monitor and optimize to ensure the best construction results.

[0102] Step S103 ensures the uniformity and quality of the foundation during deep mixing pile construction through intelligent parameter control and optimization. Upon identifying anomalies, the system responds quickly and automatically adjusts construction parameters, reducing the risk of uneven settlement. Furthermore, through real-time feedback and continuous optimization, the system can gradually improve construction efficiency and foundation treatment results.

[0103] In step S104, the following steps are further specifically included:

[0104] S41: The data integration and preprocessing stage includes two parts: first, integrating the data collected by all sensors, such as stirring torque, depth, pressure, and vibration information, ensuring the consistency of data format and removing duplicate or erroneous data points; second, filtering and standardizing the raw data to remove noise and fill missing values, thereby ensuring data accuracy and consistency.

[0105] S42: The process of generating an early warning report includes the following steps: First, identify abnormal conditions, including uneven mixing, weak formations, or equipment failures, by setting thresholds and applying anomaly detection algorithms (such as statistical control charts or machine learning models). Next, perform a risk assessment on the identified abnormal conditions and calculate the potential risk level of uneven settlement, which combines historical data, model prediction results, and the current construction environment and formation characteristics. Finally, automatically generate an early warning report based on the results of anomaly detection and risk assessment. The report includes a detailed description of the abnormal condition, the assessed risk level and its scope of influence, recommended treatment measures for the abnormal condition (such as adjusting construction parameters, increasing monitoring frequency, etc.), and graphical risk assessment results and key parameter change trends during the construction process to help construction personnel and decision makers quickly understand.

[0106] S43: The steps of generating a prediction model include model selection and training, model validation and optimization, and prediction generation. First, a suitable prediction model is selected, such as linear regression, support vector machine, or deep learning model, and trained based on historical construction data and experimental data. During the training process, factors such as stratum characteristics, construction parameters, and historical settlement data need to be considered. Next, the model is verified through cross-validation and test data sets to evaluate its prediction accuracy. Based on the verification results, the model parameters are adjusted to optimize the model performance. Finally, based on the trained model, the current construction data is predicted to generate prediction results for uneven settlement, while taking into account future construction conditions, stratum changes, and possible risk factors.

[0107] S44: Model updating and maintenance includes dynamic updating and maintenance mechanisms. The system should regularly update the prediction model to incorporate new construction data and feedback to maintain model accuracy and applicability. These updates can be performed periodically or in real time based on actual needs. Simultaneously, a model maintenance mechanism should be established to track model performance and address issues such as overfitting and data drift to ensure long-term stable operation of the system.

[0108] Through the above steps, S104 can generate detailed early warning reports and accurate prediction models, providing scientific basis and reliable support for the prevention and control of uneven settlement during construction.

[0109] In step S105, the following steps are further specifically included:

[0110] S51: Post-construction Data Collection and Processing. After deep mixing pile construction is complete, the system initiates the quality assessment phase. This phase begins by collecting data on foundation settlement, pile strength, and foundation bearing capacity using additional sensors deployed in the construction area (such as pressure sensors, displacement sensors, and foundation settlement meters). This data is then transmitted to the system's central data processing center, where it undergoes pre-processing steps such as filtering and noise removal to ensure accuracy and reliability.

[0111] S52: Quality Assessment Model Application. The system applies a calibrated and trained quality assessment model to comprehensively analyze the foundation quality of the treated area. This model combines multiple data analysis techniques, including regression analysis, statistical analysis, and foundation bearing capacity calculation, taking into account both construction and post-processing data. The assessment covers foundation settlement, pile strength distribution, and foundation bearing capacity uniformity. The system generates a detailed assessment report, identifying strengths of the foundation treatment and potential problem areas.

[0112] S53: Risk Analysis and Prediction. Based on the analysis results of the quality assessment model, the system conducts a risk analysis of the construction area, identifying possible risks such as uneven settlement and insufficient foundation bearing capacity, and predicting the potential impact of these risks on future use. The system generates a risk prediction model, combining historical and current data to predict future settlement trends and foundation stability.

[0113] S54: Based on the results of quality assessment and risk analysis, the system generates targeted improvement suggestions. These suggestions include adjusting construction parameters (such as mixing speed, grouting volume, and pressure) to optimize construction results, improving construction techniques (such as increasing the number of mixing times and optimizing mixing depth), increasing the density of monitoring equipment in potential risk areas to improve real-time understanding of foundation conditions, and proposing follow-up treatment measures for identified weak areas (such as local reinforcement and re-grouting).

[0114] S55: The system generates a detailed quality assessment report and improvement suggestion report containing all analysis results, risk predictions, recommended actions, and relevant data charts. These reports are electronically delivered to the project leader and relevant engineers for further review and implementation. The system also provides an archiving function for reports, facilitating reference and quality tracking for subsequent projects.

[0115] S56: The system will continuously monitor the long-term performance of the construction area and automatically adjust improvement recommendations based on actual conditions. If necessary, the system will update the prediction model to reflect the latest data and construction conditions to ensure continuous optimization of the foundation treatment results.

[0116] Through the above specific implementation methods, S105 ensures a comprehensive assessment and scientific improvement of the foundation quality after deep mixing pile construction, further improving construction quality and safety.

[0117] The present invention provides an intelligent early warning system for uneven settlement of deep mixing in reclamation areas, which can accurately identify stratum information and adjust construction parameters in real time, providing a scientific theoretical basis and practical support for foundation treatment in reclamation areas.

[0118] like Figure 2 As shown, the present invention also provides an intelligent early warning system for deep-layer stirring and uneven settlement in reclamation areas, including connections and interactions among an input layer, a processing layer, and an output layer.

[0119] In particular, it includes step S101, the input layer includes construction equipment and sensor modules, the construction equipment is connected to the sensor, and the sensor unit collects various key parameters in the construction process in real time through torque sensors, depth sensors, pressure sensors and acceleration sensors.

[0120] The processing layer includes a data transmission unit, a central data processing center, and a database. The sensor modules are connected to the data transmission unit and transmit the collected data to the central data processing center via a combination of wireless and wired transmission technologies. Specifically, step S102 includes the central data processing center preprocessing and analyzing the received raw data, performing data predictions using an intelligent algorithm module, and storing the results in the database.

[0121] The output layer includes a control and regulation unit, a user interface, an early warning and report generation unit, and a continuous monitoring and feedback unit. Specifically, it includes step S103, where data processed by the central data processing center is transmitted to the control and regulation unit, automatically adjusting construction parameters to optimize the construction process. These parameters are then presented directly to the user through the user interface. Specifically, it includes steps S104 and S105. Finally, when uneven settlement is detected, the alarm and report generation unit generates an early warning signal, a quality assessment report, and a forecast report, and updates the uneven settlement prediction model, providing construction personnel with real-time risk assessments and improvement recommendations.

[0122] In particular, the sensor arrangement mentioned in step S101, such as Figure 3 The figure below shows the specific placement of various sensors on deep mixing pile construction equipment. The figure indicates the installation locations of the torque sensor, depth sensor, pressure sensor, and accelerometer. These sensors are used to monitor the torque changes, pile depth, downforce, and vibration of the mixing pile in real time. By working together, these sensors enable the system to comprehensively collect key data during the construction process, providing accurate information support for subsequent analysis and control.

[0123] like Figure 4 The data processing flow chart, shown in Figure 1, illustrates the entire process from sensor data collection to generating differential settlement prediction results. Specifically, it includes the data collection and monitoring portion in step S101. First, real-time data monitoring and collection is achieved using the installed torque sensor, pressure sensor, depth sensor, and acceleration sensor. The data is then transmitted using a combination of wireless and wired transmission technologies.

[0124] In particular, Figure 4It also includes data processing and analysis in step S102. In the data processing layer, the noise and outliers in the sensor data are processed, and the data is normalized or standardized before data synchronization. In the intelligent analysis layer, the data that has been standardized, filtered and noise-removed is input into the intelligent algorithm. The system will process different types of data (such as torque, depth, pressure and acceleration) separately to ensure that each parameter can be accurately analyzed. The system first extracts key features from the input data. These features may include the pattern of stratum changes during the construction of the mixing pile, the fluctuation characteristics of the torque, the trend of pressure changes, etc.

[0125] By applying anomaly detection algorithms to the extracted features, the system is able to identify possible problems during the construction process. For example, a sudden increase in torque may mean that there is a hard layer in the stratum, and an abnormal drop in pressure may indicate the presence of cavities or weak layers in the underlying stratum. While identifying anomalies, through comprehensive analysis of data from different sensors, characteristics such as the hardness, layer thickness, and uniformity of the stratum can be identified. For example, by analyzing the relationship between torque and depth, it is possible to determine whether the mixing pile has entered different strata, thereby identifying the location and characteristics of the hard or weak layers. The system uses clustering algorithms (such as K-means) to classify the data into different stratum feature groups, which can better predict the risk of uneven settlement during construction. Each cluster group represents a type of stratum (such as hard, soft, or medium hardness layers), and these categories help the system perform further predictive analysis.

[0126] Next, the system uses pattern recognition technology to identify recurring risk patterns during construction. These patterns are used as input variables in a predictive model, helping the system proactively identify potential differential settlement risks. For example, specific torque and pressure change patterns may be identified as precursors to settlement risk, and the system generates predictive variables based on these patterns.

[0127] In particular, Figure 4Also comprise the prediction model training in step S102, extract the data set of above-mentioned processing and select the feature that has the greatest influence on the prediction risk of uneven settlement according to the above-mentioned feature selection extracted, select appropriate prediction model according to the nature of the problem.Train on the selected model, use historical construction data to train the model, make it learn the relationship between different features and uneven settlement.During the training process, the model can minimize the prediction error by adjusting its internal parameters (such as weight, bias, etc.).In order to avoid model overfitting and ensure that it performs well on unseen data, cross validation (such as k-fold cross validation) is used to validate the model, and according to the result of cross validation, the model is optimized.Test data set is used to evaluate on the finalized model.Test data set does not overlap with training data set and validation data set, ensures that the test results can reflect the performance of the model in actual application.The trained and validated model will be deployed to the system for real-time prediction of the risk of uneven settlement under current construction conditions.The system will continuously update the model input according to real-time data, and the prediction results generated by the model will be fed back to the adjustment module of the construction parameters to dynamically optimize the construction process.

[0128] Through a detailed model training and validation process, the system builds a reliable prediction model that accurately identifies risk factors that may lead to uneven settlement during construction. Cross-validation and model optimization ensure the model's robustness and accuracy, enabling efficient prediction in practical applications.

[0129] In particular, Figure 4 It also includes automatic control in step S103. According to the analysis results, after detecting the risk of uneven settlement or construction abnormality, the system adjusts key construction parameters such as stirring speed, grouting volume, and drill bit pressure through the control unit to optimize the uniformity and quality of foundation treatment. The system also feeds back to the control unit through the database during construction, and transmits the optimized parameters back to the construction equipment in real time to ensure continuous monitoring and dynamic adjustment of the construction process, thereby effectively reducing the occurrence of uneven settlement.

[0130] In particular, Figure 5 This section illustrates the key content and structure of the early warning report generated by the system in step S104, including a detailed description of the detected anomaly, the risk assessment level, recommended countermeasures, and relevant data charts. The report visually displays the risk level through color or charts, and details the nature and potential impact of the anomaly. The recommended measures section provides methods for addressing the detected potential risks, while the data charts graphically present the changing trends of key parameters during the construction process and the results of the prediction model, providing comprehensive reference information for construction personnel and managers.

Claims

1. An intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas, characterized by: include: S101, performing data monitoring and data collection through sensors to obtain raw data; S102, performing data analysis and identification on the original data based on an intelligent algorithm to obtain a data processing result; S103, regulating and optimizing the system based on the data processing results; S104, generating an early warning report and an uneven settlement prediction model; S105. Conduct quality assessment of the processing area and make improvement recommendations.

2. The intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas according to claim 1 is characterized in that: The data collection includes data preprocessing and data transmission; The data preprocessing includes denoising, filtering and data compression operations to improve transmission efficiency and the accuracy of subsequent data analysis; The data transmission uses wireless transmission technology and wired transmission technology.

3. The intelligent early warning system for uneven settlement of deep-layer mixing in reclamation areas according to claim 1 is characterized in that: The sensors in step S101 include at least two or more of a torque sensor, a depth sensor, a pressure sensor, and an acceleration sensor; The torque sensor is installed on the main shaft of the mixing pile to monitor the torque of the drill bit in real time during the construction process, reflecting the resistance of different strata and helping to identify the soft and hard changes of the strata and the mixing uniformity; The depth sensor is installed on the drill bit to accurately measure the vertical depth of the drill bit to reflect the construction progress in real time, ensure that the mixing pile is constructed according to the designed depth, and identify potential sinking or deviation problems; The pressure sensor is installed on the propulsion device of the drill bit and is used to monitor the pressure of the drill bit on the formation in real time to analyze the bearing capacity of the foundation and identify changes in the formation; The acceleration sensor is installed on the main shaft of the mixing pile and is used to detect the vibration and stability of the equipment to identify abnormal formations or obstacles in hard layers.

4. The intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas according to claim 1 is characterized in that: The data processing results obtained by performing data analysis and identification on the raw data based on the intelligent algorithm in step S102 include: Analyzing the raw data based on an intelligent algorithm to obtain the data processing result, wherein the intelligent algorithm includes a machine learning algorithm and a deep learning algorithm; By comparing historical data processing results with the data processing results, the formation characteristics are automatically identified and the uniformity of the mixing pile construction and the foundation bearing capacity are evaluated in real time. The historical data processing results are obtained by analyzing the historical data through the intelligent algorithm.

5. The intelligent early warning system for uneven settlement of deep-layer mixing in reclamation areas according to claim 4 is characterized in that: The analyzing the raw data based on the intelligent algorithm includes: Processing noise and outliers in sensor data, performing standard processing on the raw data to obtain standard data, wherein the standardization processing includes standardization, filtering and noise removal; The standard data is input into the intelligent algorithm to output key features, which include the pattern of stratum changes during the construction of the mixing pile, the fluctuation characteristics of the torque, and the change trend of the pressure.

6. The intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas according to claim 3 is characterized in that: The control and optimization system in step S103 includes the following steps: S301. Identify recurring risk patterns during construction using a three-stage pattern recognition process: initial scoring → probability density classification → geological data cross-validation. Initial strata determination is achieved through multi-sensor data fusion. A Gaussian mixture model (GMM) estimates the probability density of the feature space, classifies high-risk data into clusters, and outputs strata classification labels, including soft soil, sand, gravel, and rock. These classification results are cross-validated with geological exploration data to ensure consistency with engineering knowledge. S302. Select key features that affect the risk of differential settlement, including formation score S, torque gradient Pressure change rate ΔP, vibration energy ratio E vib , GMM classification label L∈{1,2,3,4} is used as input feature, and a prediction model is built based on historical data. The relationship between formation characteristics and uneven settlement is learned through a bidirectional LSTM network to obtain an output prediction model, which outputs the settlement probability P settlement ∈[0,1]; S303. Based on the data processing results, after detecting the risk of uneven settlement or construction anomaly, the control unit adjusts key construction parameters such as stirring speed, grouting volume, and drill bit pressure. When the probability of stratum settlement is greater than 90%, the construction parameters are adjusted to reduce the stirring speed, increase the grouting volume, and control the drill bit pressure to improve the foundation strength. When the probability of stratum settlement is greater than 70% and less than 90%, the stirring intensity and grouting speed are adjusted to avoid excessive disturbance. When the probability of stratum settlement is less than 70% and greater than 50%, the drill bit pressure and stirring intensity are preferentially adjusted to prevent increased wear or drill bit jamming, and the information is fed back to the control unit through the database during construction. S304. The system records the parameter adjustments and results of each construction operation to form a historical data archive. The results are fed back to the data analysis module. Based on cross-validation, performance indicators are set: accuracy > 90%, recall > 85%. If the model does not meet the performance requirements, closed-loop optimization is performed to optimize the construction parameter adjustment strategy. S306. Transmit the optimized parameters back to the construction equipment in real time to ensure continuous monitoring and dynamic adjustment of the construction process.

7. The intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas according to claim 1 is characterized in that: The early warning report in step S104 includes the following steps: S411, identifying abnormal situations by setting thresholds and applying anomaly detection algorithms; S412. Conduct risk assessment on the identified abnormal conditions and calculate the potential settlement probability and uneven settlement risk level; Based on historical data, a prediction model is constructed, and the relationship between stratum characteristics and uneven settlement is learned through a bidirectional LSTM network to obtain an output prediction model and output the settlement probability P. settlement ∈[0,1]; The differential settlement risk index is calculated according to the following formula: R=α1ΔS+α2ΔP+α3ΔT+α4ΔA Where ΔS, ΔP, ΔT, and ΔA represent the deviations of depth, pressure, torque, and acceleration, respectively, and α1, α2, α3, and α4 are weighting coefficients. The subsidence risk level is assessed based on the R value. According to the above calculations, there are three levels of warning thresholds, the parameters of which can be adjusted according to the actual project. Level 1 warning: When R>0.6 or P settlement When the rate is >70%, an audible and visual alarm will be given and the speed and injection volume will be automatically fine-tuned; Level 2 warning: When R>0.75 or P settlement When the rate is >80%, the machine should be shut down for inspection and local high-pressure grouting should be performed; Level 3 warning: When R>0.9 and P settlement When the load is >90%, the machine will be shut down for inspection and emergency support; S413. The system generates an early warning report based on the anomaly detection and risk assessment results.

8. The intelligent early warning system for uneven settlement of deep-layer stirring in reclamation areas according to claim 7 is characterized in that: The content of the early warning report includes a description of the abnormal situation, the uneven settlement level and its impact range, recommended treatment measures for the abnormal situation, the trend of changes in key parameters and graphical risk assessment results.

9. The intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas according to claim 3, characterized in that: The automatic identification of formation characteristics includes: The system uses a Gaussian mixture model to classify the raw data into different formation feature groups to predict the risk of uneven settlement during construction; The stratum characteristic group includes soft soil, sand layer, gravel layer and rock layer.

10. The intelligent early warning system for uneven sedimentation in deep-layer mixing in reclamation areas according to claim 1, characterized in that: The step S105 further includes: The system uses quality assessment models and risk prediction models to conduct quality assessment and risk analysis of the foundation in the treatment area, generates detailed assessment reports and improvement suggestions, and predicts future settlement trends and foundation stability; The quality assessment model includes three parameters: foundation bearing capacity, uniformity index, and construction efficiency; The foundation bearing capacity is calculated as follows: q u =1.2cN c +γD f N q or 搅拌 The uniformity index is: Among them, σ T , σ P are torque and pressure standard deviation, μ T 、μ P are the mean values of torque and pressure, respectively. The construction efficiency is: The assessment report includes foundation settlement, pile strength distribution, foundation bearing capacity, construction efficiency and uneven settlement risk, as well as recommended measures and relevant data charts.

Images