Construction control method for large-diameter bored pile in soil and rock mixture stratum of hillside

Abstract

This invention discloses a construction control method for large-diameter bored piles in hillside soil-rock mixture strata, including the construction of a data monitoring module, the deployment of an initial parameter module, the deployment of a real-time control module, preliminary control of the construction process, dynamic prediction of the construction process, dynamic control of the construction process, and continuous model optimization. The method involves real-time acquisition of drilling parameters, mud properties, vibration signals, and slurry return parameters; deployment of an initial parameter module to generate initial construction parameters based on geological feature vectors; deployment of a real-time control module employing a two-stage hybrid intelligent model architecture to sequentially perform real-time condition diagnosis and intelligent parameter control; implementation of a construction process encompassing preliminary control, dynamic prediction, and dynamic control; and finally, system self-learning through continuous model optimization. This method transforms the construction process from static experience-based dependence to dynamic intelligent control, effectively preventing problems such as borehole collapse and deviation, and significantly improving borehole quality, construction efficiency, and safety.

Description

Technical Field

[0001] This invention relates to the field of pile foundation engineering technology, and in particular to a construction control method for large-diameter bored cast-in-place piles in hillside soil-rock mixture strata. Background Technology

[0002] Soil-rock mixture strata, as a typical heterogeneous geological body, have a complex and varied material composition, typically consisting of a mixture of soil, gravel, boulders, and weak interlayers with distinct mechanical properties. This stratum has a loose structure, poor cementation, and is significantly affected by hillside topography and groundwater activity, resulting in extremely weak borehole wall self-stabilization. When constructing large-diameter bored piles in soil-rock mixture strata, drilling is highly susceptible to problems such as borehole collapse, grout leakage, and borehole deviation, severely restricting drilling efficiency and the final pile quality.

[0003] Currently, when constructing large-diameter bored piles in this type of stratum, the commonly used drilling techniques largely rely on the operator's experience in setting parameters and subjective judgment. This approach is ill-suited to the dynamic changes and heterogeneous characteristics of soil-rock mixture strata in both the longitudinal and transverse directions, often leading to a series of technical challenges such as low drilling efficiency, difficulty in ensuring hole quality, and damage to the integrity of the pile body.

[0004] For bored pile construction in complex geological conditions, existing technologies often employ several passive protection measures, such as pre-grouting to reinforce the soil around the borehole, using long steel casings for wall protection, or adjusting mud performance parameters. However, these traditional methods generally lack real-time sensing capabilities and dynamic response mechanisms for geological changes. Adjustments to construction parameters often lag behind changes in actual working conditions, resulting in limited effectiveness and high costs. Especially when encountering large rocks, weak interlayers, or sudden seepage paths during drilling, traditional methods struggle to quickly and accurately identify the working conditions and implement targeted measures, easily leading to construction interruptions or even safety accidents.

[0005] In recent years, with the advancement of intelligent construction technology, drilling process control methods based on data monitoring and model analysis have been gradually applied. Most existing research focuses on parameter optimization in homogeneous formations, adjusting parameters such as drill pressure, rotational speed, and torque by monitoring conventional drilling parameters and combining them with empirical models or simple control algorithms. However, these methods are clearly insufficient for highly heterogeneous formations like soil-rock mixtures. The key issue is the failure to effectively integrate and interpret critical characteristic signals that are highly sensitive to abrupt changes in formation and borehole stability, such as vibration spectrum characteristics and real-time changes in slurry return during drilling. Therefore, existing intelligent control methods have significant limitations in the accuracy of condition diagnosis and the precision of real-time control, failing to meet the urgent need for safe and efficient construction in soil-rock mixture formations.

[0006] In summary, existing technologies for large-diameter bored pile construction in hillside soil-rock mixture strata have significant shortcomings in real-time condition sensing, intelligent diagnosis, and dynamic control. There is an urgent need to develop an advanced control method capable of real-time sensing of strata changes, intelligent diagnosis of drilling conditions, and dynamic optimization of construction parameters. By introducing multi-source sensor data fusion and intelligent modeling technology, an intelligent system with real-time diagnosis and precise control capabilities can be constructed, thereby fundamentally improving the quality, efficiency, and safety of bored pile construction in such complex strata. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention proposes an intelligent control method for the construction of large-diameter bored piles in hillside soil-rock mixture strata. The core of this method lies in constructing and applying an intelligent control system based on multi-source data fusion and neural networks to achieve real-time diagnosis and dynamic optimization control of the drilling and hole-forming processes. This system not only integrates conventional construction parameters but also incorporates vibration spectrum and mud property signals sensitive to strata characteristics. Furthermore, it employs a phased model architecture of diagnosis and control to enhance the targeting and robustness of the control.

[0008] Construction control methods for large-diameter bored piles in hillside soil-rock mixture strata include the following steps: S1. Construction of the data monitoring module: The intelligent control system includes a data monitoring module, an initial parameter module, a real-time control module, and an equipment execution interface. The data monitoring module is configured with a drilling parameter sensor group, a mud performance sensor group, a vibration signal sensor, and a mud return monitoring unit. The drilling parameter sensor group is installed at the drilling rig's power head and is used to collect the drilling pressure in real time during the drilling process. V 1 Rotation speed V 2 Torque V 3 With drilling speed V 4 ; The mud performance sensor array is deployed in the orifice mud circulation tank or main pipeline for continuous monitoring of the mud level in the wall-protecting mud. V 5 ,density V 6 With viscosity V 7 ; The vibration signal sensor is a triaxial accelerometer, installed on the upper part of the drill rod near the borehole or on the key load-bearing structure of the drilling rig. It is used to collect broadband vibration signals generated during drilling and extract the dominant vibration frequency through real-time spectrum analysis. V 8 and amplitude V9 ; The slurry return monitoring unit integrates a flow meter, a temperature sensor, and a sand content meter, and is installed on the slurry return outlet pipeline for real-time monitoring of the slurry return flow rate. V 10 , return temperature V 11 and sand content of slurry V 12 The data collected by the data monitoring module constitutes the monitoring parameter set. P =( V 1 , V 2 , V 3 , V 4 , V 5 , V 6 , V 7 , V 8 , V 9 , V 10 , V 11 , V 12 ); S2. Deployment of the initial parameter module: The initial parameter module, a key component of the intelligent control system, is designed to construct and apply a geological initial parameter mapping model. This model, trained using historical construction data and based on supervised machine learning, aims to establish a nonlinear mapping relationship from geological feature vectors to initial construction parameters, thereby providing scientifically sound initial parameter settings for drilling operations. The geological feature vectors include formation lithology complexity score, volumetric stone content, maximum boulder size, weak interlayer thickness, groundwater depth, and groundwater activity score. The initial construction parameters include initial drill pressure, lower limit of rotational speed range, upper limit of rotational speed range, torque safety threshold, initial drilling speed, initial mud density, and initial mud viscosity. The formation lithology complexity score is quantified on a scale of 1 to 10 based on the number of lithological types, hardness differences, and uniformity of distribution within the formation. The groundwater activity score is quantified on a scale of 1 to 10 based on the frequency and amplitude of historical water level fluctuations. S3, Deployment of the real-time control module: The real-time control module constitutes the core of the intelligent control system. It adopts a two-stage hybrid intelligent model architecture to realize the accurate identification of drilling conditions and the dynamic optimization of construction parameters in sequence. The real-time control module receives and integrates multi-source sensor signals output by the data monitoring module in real time. First, it performs a rapid diagnosis of the current drilling conditions. Then, based on the diagnosis results, it calls the corresponding expert strategy model to generate precise construction parameter adjustment instructions. The first stage is the real-time operational condition diagnostic model. The core function of this model is based on the monitoring parameter set. P The system intelligently identifies and classifies the working conditions during drilling; the real-time working condition diagnosis model is implemented using a multi-layer sensing mechanism classifier; the input feature vector of the real-time working condition diagnosis model is constructed by fusing multi-source real-time monitoring data, and this vector contains... V 1 to V 12 The mean and V 8 and V 9 The standard deviation is input into the fully connected hidden layer; the output of the real-time working condition diagnosis model is a 5-dimensional probability vector, whose components correspond to the probability of occurrence of 5 preset typical working conditions, including normal homogeneous drilling, encountering large rocks or obstacles, crossing weak and easily collapsible interlayers, leakage or loss of slurry, and local spalling of the borehole wall; when the output probability of any preset typical working condition exceeds the system's preset threshold, it is determined that the current situation is under the preset typical working condition, and the corresponding parameter intelligent control model in the second stage is triggered immediately. The second stage is the intelligent parameter control model, which adopts an integrated and scalable expert strategy model library. For each preset typical working condition that can be identified in the first stage, there is a dedicated expert strategy sub-model. Each expert strategy sub-model also adopts a multi-layer sensing mechanism structure. Each expert strategy sub-model is trained through supervised learning. The training data comes from the case data set of successful handling of various working conditions in the historical construction process. S4. Preliminary control of the construction process: After the drilling rig is in place and before formal drilling begins, preliminary control of the construction process is carried out. The specific implementation procedure is as follows: S401. Extraction of geological feature parameters: The operator extracts the geological feature vector from the geological survey report of the current project and inputs it into the initial parameter module. The module then calls its built-in geological initial parameter mapping model to perform forward calculation and outputs a set of initial setting values ​​for construction parameters for the current geological conditions of the drilled pile location. S402. Configuration of equipment control parameters: After obtaining the initial set values ​​of the construction parameters, they are sent to the drilling rig control system and the mud control system; the drilling rig control system automatically completes the initial setting and calibration of the initial drilling pressure, lower limit of the rotational speed range, upper limit of the rotational speed range, torque safety threshold, and initial drilling speed; the mud control system automatically adjusts the initial mud density and initial mud viscosity, and sets the initial liquid level of the mud circulation system. S403 Drilling Operation: After the initial configuration and two-way verification of all equipment parameters are completed, the drilling rig can start drilling operations with the optimized parameters. From the moment the drill bit contacts the ground, the data monitoring module immediately starts high-frequency data acquisition, and the entire system synchronously enters a closed-loop intelligent control cycle consisting of real-time working condition diagnosis and dynamic parameter adjustment. S5. Dynamic prediction of the construction process: During drilling, the data monitoring module continuously collects and updates the monitoring parameter set at a frequency of no less than 10Hz. P Meanwhile, the real-time diagnostic model for operating conditions in the real-time control module uses the set of monitoring parameters collected during the current control cycle. P Based on this, pattern recognition calculations are continuously performed; the real-time working condition diagnostic model performs a calculation every 500 milliseconds, outputting the classification result of the current drilling working condition and its corresponding probability value; The intelligent parameter control model in the real-time control module is based on the monitoring parameter set. P The output of the real-time diagnostic model of the working condition is used for forward calculation and reasoning to output a structured set of control instructions. This set of instructions clearly specifies the target set value or the adjustment range relative to the current value of each key construction parameter. S6. Dynamic control of the construction process: First, the control command set output by the real-time control module is checked for safety to ensure that all parameter adjustments are within the allowable operating range of the equipment. The verified control commands are then sent in parallel to the drilling rig control system and the mud control system in the form of digital signals through a standardized equipment execution interface. After the control command is executed, the data monitoring module immediately initiates a new round of high-frequency data acquisition to obtain the real-time monitoring parameter set following the command execution. P The system compares the preset target parameter values ​​with the actual parameter values ​​fed back by the equipment to analyze the execution deviation. At the same time, the system packages the control command issued this time, the corresponding working condition diagnostic label, and the monitoring data after execution together to form a complete record. S7. Continuous Model Optimization: After each complete construction control cycle, the system automatically integrates the key data generated in that cycle into a standardized training sample data package. This data package mainly contains three types of information: the first type is the control decision input characteristics, namely the standardized real-time monitoring parameter set at the moment before the parameter adjustment command is issued. P The first category is the working condition classification label output by the real-time working condition diagnostic model; the second category is the control decision output result, namely the construction parameter adjustment instruction vector generated and finally issued by the parameter intelligent control model; the third category is the control effect verification feedback, namely the monitoring parameter set collected after the instruction is executed and in the new control cycle. P The numerical value, and the working condition status obtained by re-diagnosis based on this; the system's model optimization process is automatically started by preset trigger conditions; the trigger conditions are set after a set number of boreholes are completed, or when the total number of stored training samples reaches a predetermined threshold; the optimization process uses multiple accumulated historical training sample data packets as training sets, and adopts supervised learning to update the parameters of the system's core model.

[0009] Preferably, in step S2, the geological initial parameter mapping model is implemented using a feedforward neural network architecture; the network structure includes an input layer, two hidden layers, and an output layer; the number of nodes in the input layer corresponds to the number of geological feature parameters; the first hidden layer contains 64 neurons and uses the ReLU activation function; the second hidden layer contains 32 neurons and also uses the ReLU activation function; the number of nodes in the output layer corresponds to the number of construction parameters to be recommended and uses a linear activation function. During the model training phase, structured geological feature parameters extracted from geological survey reports of numerous completed projects were used as input features, and the corresponding optimal initial construction parameter combinations, verified through practice, were used as training labels. The training process employed the Adam optimizer, with an initial learning rate set to 0.001, a mean squared error loss function, a batch size of 32, and at least 100 training epochs. Early stopping was used to prevent overfitting. The input geological feature vectors underwent standardized preprocessing, specifically including stratigraphic lithological complexity scores, volumetric stone content, maximum boulder size, weak interlayer thickness, groundwater depth, and groundwater activity scores. The model outputs initial construction parameters, which are a 7-dimensional vector, corresponding to the initial settings of the following construction parameters: initial drilling pressure, lower limit of rotation speed range, upper limit of rotation speed range, torque safety threshold, initial drilling speed, initial mud density, and initial mud viscosity.

[0010] Preferably, in step S7, the parameter update method is as follows: First, the real-time control module is optimized, using the control decision input features in the training samples as model input, and the control decision output results, i.e., the actual issued parameter adjustment instructions that have been verified as effective on-site, as the training target label; by minimizing the error between the model's predicted instructions and the actual effective instructions, and using the error backpropagation algorithm, the network weight parameters inside the real-time diagnostic model and the intelligent parameter control model are updated, so that the model's decision logic continuously approaches the optimized strategy that has been tested in practice; Second, the geological initial parameter mapping model in the initial parameter module is optimized. This optimization is usually carried out after completing a complete engineering project or accumulating a sufficient number of successful cases covering different geological conditions. The optimization training uses standardized geological feature parameters of various geological boreholes as input, and the initial combination of construction parameters that has been verified in field practice and ultimately achieved good pile formation results as the training target label. Through training, the parameters of this mapping model are continuously corrected, enhancing its ability to recommend scientific and reasonable initial construction parameters according to specific geological conditions.

[0011] In summary, compared with the prior art, the construction control method for large-diameter bored piles in hillside soil-rock mixture strata provided by the present invention has the following significant advantages: 1) Real-time perception and accurate diagnosis capability: By integrating multi-source sensors such as drilling parameters, vibration spectrum, and mud properties, the system can capture sudden changes in formation such as encountering large rocks, weak interlayers, or mud leakage in real time, and quickly classify working conditions such as normal drilling and borehole instability using a lightweight neural network model. This real-time diagnosis capability overcomes the lag of traditional judgment based on human experience and significantly improves the accuracy and response speed of working condition identification. 2) Dynamically optimize construction parameters: A two-stage hybrid intelligent model is adopted, which automatically calls expert strategies based on real-time diagnostic results and dynamically adjusts key parameters such as drilling pressure, rotation speed, and mud performance. For example, when encountering large rocks, the impact function is automatically activated, or the mud viscosity is adjusted when there is leakage, thereby effectively preventing problems such as hole collapse and hole deviation, and improving drilling efficiency and pile quality. 3) A fundamental shift from static experience-based control to dynamic intelligent control has been achieved: The closed-loop control circuit of "perception-diagnosis-decision-execution-verification" formed by the real-time control module has completely changed the outdated mode where construction parameters were static or could only be passively fine-tuned; the system can actively and dynamically adjust all key parameters such as drilling pressure, rotation speed, and mud properties according to the ever-changing actual conditions inside the hole, so that the construction process is always kept in the optimal state, effectively curbing the occurrence of common quality problems such as hole collapse and hole deviation, and significantly improving the quality of hole formation and construction efficiency; 4) A new intelligent collaborative control mechanism has been established that runs through the entire construction process: the initial parameter module intelligently recommends the optimal starting parameters based on geological characteristics, laying a high starting point for high-quality construction; the real-time control module ensures precise control during the construction process; and the model continuous optimization mechanism enables the system to have self-learning capabilities, and can use the data accumulated from each construction to iteratively upgrade the model, achieving a virtuous cycle of "getting smarter with use"; this full-process, self-evolving intelligent collaborative control mechanism greatly reduces the reliance on the personal experience of operators and improves the standardization level and overall reliability of construction. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating the construction control method for large-diameter bored piles in hillside soil-rock mixture strata, as shown in an embodiment of the present invention. Detailed Implementation

[0013] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments given herein are for illustration and explanation only and are not intended to limit the present invention.

[0014] This invention discloses as follows Figure 1 The construction control method for large-diameter bored piles in hillside soil-rock mixture strata, as shown, includes the following steps: S1. Construction of the Data Monitoring Module: The intelligent control system includes a data monitoring module, an initial parameter module, a real-time control module, and an equipment execution interface. The data monitoring module is configured with a drilling parameter sensor group, a mud performance sensor group, a vibration signal sensor, and a mud return monitoring unit. The drilling parameter sensor group is installed at the drilling rig's power head and is used to collect the drilling pressure in real time during the drilling process. V 1 Rotation speed V 2 Torque V 3 With drilling speed V 4 ; The mud performance sensor array is deployed in the orifice mud circulation tank or main pipeline for continuous monitoring of the mud level in the wall-protecting mud. V 5 ,density V 6 With viscosity V 7 ; The vibration signal sensor is a triaxial accelerometer, installed on the upper part of the drill rod near the borehole or on the key load-bearing structure of the drilling rig. It is used to collect broadband vibration signals generated during drilling and extract the dominant vibration frequency through real-time spectrum analysis. V 8 and amplitudeV 9 This is used to characterize the interaction between the drill bit and the formation, reflecting working conditions such as changes in formation lithology, encountering large rocks, or local instability of the borehole wall. In specific implementation, the sampling frequency of the triaxial accelerometer is not less than 5kHz, and the range is not less than ±50g to meet the requirement of complete capture of drilling impact and vibration signals. The real-time spectrum analysis can be achieved using fast Fourier transform, specifically using a Hanning window. The analysis window length is usually 1024 points, and the frequency resolution can be set according to the sampling frequency. The slurry return monitoring unit integrates a flow meter, a temperature sensor, and a sand content meter, and is installed on the slurry return outlet pipeline for real-time monitoring of the slurry return flow rate. V 10 , return temperature V 11 and sand content of slurry V 12 Used to comprehensively determine whether mud leakage has occurred, whether there is frictional heat generation within the borehole, whether there is groundwater activity, or whether the borehole cleaning effect is satisfactory; the data collected by the data monitoring module constitutes a monitoring parameter set. P =( V 1 , V 2 , V 3 , V 4 , V 5 , V 6 , V 7 , V 8 , V 9 , V 10 , V 11 , V 12 ).

[0015] S2. Deployment of the Initial Parameter Module: As a key component of the intelligent control system, the initial parameter module's core function is to construct and apply a geological initial parameter mapping model. This model, based on supervised machine learning and trained using historical construction data, aims to establish a nonlinear mapping relationship from geological feature vectors to initial construction parameters, thereby providing scientifically reasonable initial parameter settings for drilling operations. The geological feature vectors include stratum lithology complexity score, volumetric stone content, maximum boulder size, weak interlayer thickness, groundwater depth, and groundwater activity score. The initial construction parameters include initial drilling pressure, lower limit of rotational speed range, upper limit of rotational speed range, torque safety threshold, initial drilling speed, initial mud density, and initial mud viscosity. The stratum lithology complexity score is quantified on a scale of 1 to 10 based on the number of lithological types, hardness differences, and uniformity of distribution within the strata. The groundwater activity score is quantified on a scale of 1 to 10 based on the frequency and amplitude of historical water level fluctuations. In specific implementation, the geological initial parameter mapping model is implemented using a feedforward neural network architecture; the network structure includes an input layer, two hidden layers, and an output layer; the number of nodes in the input layer corresponds to the number of geological feature parameters; the first hidden layer contains 64 neurons and uses the ReLU activation function; the second hidden layer contains 32 neurons and also uses the ReLU activation function; the number of nodes in the output layer corresponds to the number of construction parameters to be recommended and uses a linear activation function. During the model training phase, structured geological feature parameters extracted from geological survey reports of numerous completed projects were used as input features, and the corresponding optimal initial construction parameter combinations, verified through practice, were used as training labels. The training process employed the Adam optimizer, with an initial learning rate set to 0.001, a mean squared error loss function, a batch size of 32, and at least 100 training epochs. Early stopping was used to prevent overfitting. The input geological feature vectors underwent standardized preprocessing, specifically including stratigraphic lithological complexity scores, volumetric stone content, maximum boulder size, weak interlayer thickness, groundwater depth, and groundwater activity scores. The model outputs initial construction parameters, which are a 7-dimensional vector, corresponding to the initial settings of the following construction parameters: initial drilling pressure, lower limit of rotation speed range, upper limit of rotation speed range, torque safety threshold, initial drilling speed, initial mud density, and initial mud viscosity. Through this model, the system can intelligently recommend the initial settings of construction parameters based on the real-time geological conditions, laying an optimized foundation for the subsequent drilling process. During implementation, operators need to extract the aforementioned geological feature vectors from the geological survey report of the current project and then input them into the initial parameter module. The module automatically calls the geological initial parameter mapping model to perform calculations, outputs a complete combination of initial construction parameters, and sends them to the drilling rig and mud control system through the equipment execution interface to complete the initial configuration.

[0016] S3, Deployment of the real-time control module: The real-time control module constitutes the core of the intelligent control system. It adopts a two-stage hybrid intelligent model architecture to realize the accurate identification of drilling conditions and the dynamic optimization of construction parameters in sequence. The real-time control module receives and integrates multi-source sensor signals output by the data monitoring module in real time. First, it performs a rapid diagnosis of the current drilling conditions. Then, based on the diagnosis results, it calls the corresponding expert strategy model to generate precise construction parameter adjustment instructions. The first stage is the real-time operational condition diagnostic model. The core function of this model is based on the monitoring parameter set. P The system intelligently identifies and classifies the working conditions during drilling. The real-time working condition diagnosis model is implemented using a multilayer perceptron classifier, with the following network structure: the number of nodes in the input layer is determined by the dimension of the feature vector; the subsequent layers include three fully connected hidden layers, with 128 neurons in the first hidden layer, 64 neurons in the second hidden layer, and 32 neurons in the third hidden layer, all using the ReLU activation function; the number of neurons in the output layer is consistent with the number of working condition categories preset by the system, such as five typical working conditions, and the Softmax activation function is used to output the probability of belonging to each working condition category. The input feature vector of the real-time diagnostic model for operating conditions is constructed by fusing multi-source real-time monitoring data; firstly, the continuously collected monitoring parameter set... P The 12 parameters were standardized and preprocessed, and their average value within the most recent 5-second time window was calculated as a representation of the instantaneous state of the parameters; simultaneously, the vibration dominant frequency collected by the vibration signal sensor was analyzed. V 8 and amplitude V 9 The data is further analyzed by calculating its standard deviation within the same five-second time window to capture the wave characteristics of the vibration signal; finally, a fused feature vector is constructed, which contains... V 1 to V 12 The mean and V 8 and V 9 The standard deviation of the equation is then input into the fully connected hidden layer. The output of the real-time working condition diagnosis model is a 5-dimensional probability vector, whose components correspond to the probability of occurrence of 5 preset typical working conditions, including normal homogeneous drilling, encountering large rocks or obstacles, traversing weak and easily collapsible interlayers, leakage or loss of grout, and local spalling of the borehole wall. The system sets a confidence threshold of 0.7. When the probability of any working condition in the output exceeds this threshold, it is determined that the current working condition is in that specific condition, and the corresponding intelligent parameter control model in the second stage is triggered. The second stage is the intelligent parameter control model, which adopts an integrated and scalable expert strategy model library. For each typical working condition that can be identified in the first stage, a dedicated expert strategy sub-model is preset. Each expert strategy sub-model also adopts a multi-layer sensing mechanism structure, and its network architecture is unified as follows: input layer, two hidden layers and output layer; the first hidden layer contains 64 neurons and the second hidden layer contains 32 neurons. The hidden layers all use the ReLU activation function; the number of neurons in the output layer corresponds to the number of construction parameters that need to be adjusted. Each expert strategy sub-model is trained using supervised learning; the training data comes from a collection of case studies documenting successful handling of various working conditions during historical construction processes; the input features of each training sample include: a standardized set of real-time monitoring parameters. P The training parameters include the numerical values ​​of the working condition, the vector of the working condition label output by the real-time working condition diagnostic model after one-hot encoding, the duration of the currently diagnosed working condition, and the parameter adjustment instructions issued in the previous control cycle. The training objective is to minimize the mean square error between the construction parameter adjustment amount predicted by the model and the actual optimal adjustment amount recorded in the case. The Adam optimizer is used in the training process, and the learning rate is set to 0.001.

[0017] Taking the expert strategy sub-model for encountering large rocks or obstacles as an example, its output is a specific parameter adjustment command vector. This vector clearly indicates the adjustment target or amount for each key construction parameter, such as adjusting the drill pressure to 60% of the current value, increasing the rotational speed to 120% of the current value, increasing the torque safety threshold by 15%, outputting a command to enable the impact function, indicating that the mud density should not be adjusted for the time being, instructing the mud viscosity to increase by 10%, and suggesting lifting the drill and letting it stand for 120 seconds. The outputs of the expert strategy sub-models for other working conditions are similar, all being a set of construction parameter adjustment amounts optimized for that specific working condition. Through the collaborative work of the above two-stage models, the real-time control module achieves rapid response to complex drilling conditions and precise dynamic optimization of construction parameters.

[0018] S4. Preliminary Control of the Construction Process: After the drilling rig is in place and before formal drilling begins, preliminary control of the construction process is carried out. The specific implementation procedure is as follows: S401. Extraction of geological feature parameters: The operator extracts the geological feature vector from the geological survey report of the current project and inputs it into the initial parameter module. The module then calls its built-in geological initial parameter mapping model to perform forward calculation and outputs a set of initial setting values ​​for construction parameters for the current geological conditions of the drilled pile location. S402. Configuration of equipment control parameters: After obtaining the initial set values ​​of the construction parameters, they are sent to the drilling rig control system and the mud control system; the drilling rig control system automatically completes the initial setting and calibration of the initial drilling pressure, lower limit of the rotational speed range, upper limit of the rotational speed range, torque safety threshold, and initial drilling speed; the mud control system automatically adjusts the initial mud density and initial mud viscosity, and sets the initial liquid level of the mud circulation system. S403 Drilling Operation: After the initial configuration and two-way verification of all equipment parameters are completed, the drilling rig can start drilling operations with the optimized parameters. From the moment the drill bit contacts the ground, the data monitoring module immediately starts high-frequency data acquisition, and the entire system synchronously enters a closed-loop intelligent control cycle consisting of real-time working condition diagnosis and dynamic parameter adjustment.

[0019] S5. Dynamic prediction of the construction process: During drilling, the data monitoring module continuously collects and updates the monitoring parameter set at a frequency of no less than 10Hz. P Meanwhile, the real-time diagnostic model for operating conditions in the real-time control module uses the monitoring parameter set collected in the previous control cycle. P Based on this, pattern recognition calculations are continuously performed; the real-time working condition diagnostic model performs a calculation every 500 milliseconds, outputting the classification result of the current drilling working condition and its corresponding probability value; The intelligent parameter control model in the real-time control module is based on the monitoring parameter set. P The output of the real-time diagnostic model is used to perform forward calculations and inferences, and output a structured set of control instructions. This set of instructions clearly specifies the target set values ​​or adjustment ranges relative to the current values ​​of various key construction parameters.

[0020] S6. Dynamic control of the construction process: First, the control instruction set output by the real-time control module is checked for safety to ensure that all parameter adjustments are within the allowable working range of the equipment. The verified control commands are sent in parallel to the drilling rig control system and mud control system in the form of digital signals through a standardized equipment execution interface. The drilling rig control system drives key actuators in real time based on received commands. Specific control actions include: precisely adjusting hydraulic pressure to achieve drilling pressure. V 1 Target setpoint; adjust the speed of the power head motor to control the rotation speed. V 2Dynamically set the safety threshold of the torque limiter V 3 When the working condition diagnostic model identifies a situation where a large rock or obstacle is encountered, the system will automatically send a command to the drilling rig to activate or adjust the impact function. The drilling rig control system must immediately execute the received command and send the actual set values ​​of each parameter back to the intelligent control system as feedback signals. The mud control system precisely controls the density of the mud by adjusting the automatic addition devices for weighting agents and thickeners, based on received instructions. V 6 With viscosity V 7 To reach the target value; maintain the orifice mud level by adjusting the speed of the circulating pump or the opening of the pipeline valves. V 5 At the set height. If the instruction set contains special process instructions, such as an instruction to prepare high-viscosity plugging slurry when a leakage condition is diagnosed, the mud system will automatically start the preset special slurry preparation process; After the control command is executed, the data monitoring module immediately initiates a new round of high-frequency data acquisition to obtain the real-time monitoring parameter set following the command execution. P The system compares the preset target parameter values ​​with the actual parameter values ​​fed back by the equipment to analyze the execution deviation. At the same time, the system packages the control command issued this time, the corresponding working condition diagnostic label, and the monitoring data after execution together to form a complete record. This record is not only used to evaluate the effectiveness of this control action in real time and determine whether the construction status is improving in the expected direction, but also stored in the database as a historical sample to provide training data for the continuous optimization of the model in the subsequent S7 steps.

[0021] S7. Continuous Model Optimization: After each complete construction control cycle, the system automatically integrates the key data generated in that cycle into a standardized training sample data package. This data package mainly contains three types of information: the first type is the control decision input characteristics, namely the standardized real-time monitoring parameter set at the moment before the parameter adjustment command is issued. P The first category is the working condition classification label output by the real-time working condition diagnostic model; the second category is the control decision output result, namely the construction parameter adjustment instruction vector generated and finally issued by the parameter intelligent control model; the third category is the control effect verification feedback, namely the monitoring parameter set collected after the instruction is executed and in the new control cycle. P The numerical value, and the operating condition status obtained based on this re-diagnosis.

[0022] The system's model optimization process is automatically initiated by preset trigger conditions. These trigger conditions can be set after a set number of boreholes are completed, or when the total number of stored training samples reaches a predetermined threshold. The optimization process uses multiple accumulated historical training sample data packages as the training set and employs supervised learning to update the parameters of the system's core model.

[0023] In practice, the parameter update method is as follows: First, the two-stage model in the real-time control module is optimized. The control decision input features in the training samples are used as the model input, and the control decision output, i.e., the actual issued parameter adjustment commands that have been verified as effective on-site, are used as the training target labels. By minimizing the error between the model's predicted commands and the actual effective commands, and using the error backpropagation algorithm, the network weight parameters inside the real-time diagnostic model and the intelligent parameter control model are updated, so that the model's decision logic continuously approaches the optimized strategy that has been tested in practice.

[0024] Secondly, the geological initial parameter mapping model in the initial parameter module is optimized. This optimization is typically performed after completing a full engineering project or accumulating a sufficient number of successful cases covering different geological conditions. The optimization training uses standardized geological characteristic parameters from various geological boreholes as input, and the initial combination of construction parameters that has been verified in field practice and ultimately achieved good pile-forming results is used as the training target label. Through training, the parameters of this mapping model are continuously corrected, enhancing its ability to recommend scientific and reasonable initial construction parameters based on specific geological conditions.

[0025] Through the aforementioned continuous model optimization mechanism, the system forms a virtuous cycle of "data accumulation, model update, and performance improvement," ensuring that the intelligent control method can adapt to complex and ever-changing geological conditions and become increasingly accurate and reliable over time.

[0026] The above describes one or more embodiments of the present invention in a relatively specific and detailed manner, but it should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A construction control method for large-diameter bored piles in hillside soil-rock mixture strata, characterized in that, Includes the following steps: S1. Construction of the Data Monitoring Module: The intelligent control system includes a data monitoring module, an initial parameter module, a real-time control module, and an equipment execution interface. The data monitoring module is configured with a drilling parameter sensor group, a mud performance sensor group, a vibration signal sensor, and a mud return monitoring unit. The drilling parameter sensor group is installed at the drilling rig's power head and is used to collect the drilling pressure in real time during the drilling process. V 1 Rotation speed V 2 Torque V 3 With drilling speed V 4 ; The mud performance sensor array is deployed in the orifice mud circulation tank or main pipeline for continuous monitoring of the mud level in the wall-protecting mud. V 5 ,density V 6 With viscosity V 7 ; The vibration signal sensor is a triaxial accelerometer, installed on the upper part of the drill rod near the borehole or on the key load-bearing structure of the drilling rig. It is used to collect broadband vibration signals generated during drilling and extract the dominant vibration frequency through real-time spectrum analysis. V 8 and amplitude V 9 ; The slurry return monitoring unit integrates a flow meter, a temperature sensor, and a sand content meter, and is installed on the slurry return outlet pipeline for real-time monitoring of the slurry return flow rate. V 10 , return temperature V 11 and sand content of slurry V 12 The data collected by the data monitoring module constitutes the monitoring parameter set. P =( V 1 , V 2 , V 3 , V 4 , V 5 , V 6 , V 7 , V 8 , V 9 , V 10 , V 11 , V 12 ); S2. Deployment of the Initial Parameter Module: The initial parameter module adopts a geological initial parameter mapping model to establish a nonlinear mapping relationship from geological feature vectors to initial construction parameters. The geological feature vectors include stratum lithology complexity score, volumetric stone content, maximum boulder size, weak interlayer thickness, groundwater depth, and groundwater activity score. The initial construction parameters include initial drill pressure, lower limit of rotation speed range, upper limit of rotation speed range, torque safety threshold, initial drilling speed, initial mud density, and initial mud viscosity. The stratum lithology complexity score is quantified on a scale of 1 to 10 based on the number of lithological types, hardness differences, and uniformity of distribution within the strata. The groundwater activity score is quantified on a scale of 1 to 10 based on the frequency and amplitude of historical water level fluctuations. S3, Deployment of the real-time control module: The real-time control module adopts a two-stage hybrid intelligent model architecture, based on the monitoring parameter set. P The current drilling conditions are diagnosed, and then the corresponding expert strategy model is invoked based on the diagnosis results to generate construction parameter adjustment instructions. The first stage is the real-time working condition diagnosis model, whose core function is based on the monitoring parameter set. P Intelligent identification and classification of working conditions during drilling; the input feature vector of the real-time working condition diagnosis model is constructed by fusing multi-source real-time monitoring data, and this vector contains... V 1 to V 12 The mean and V 8 and V 9 The standard deviation is input into the fully connected hidden layer; the output of the real-time working condition diagnosis model is a 5-dimensional probability vector, whose components correspond to the probability of occurrence of 5 preset typical working conditions, including normal homogeneous drilling, encountering large rocks or obstacles, crossing weak and easily collapsible interlayers, leakage or loss of slurry, and local spalling of the borehole wall; when the output probability of any preset typical working condition exceeds the system's preset threshold, it is determined that the current situation is under the preset typical working condition, and the corresponding parameter intelligent control model in the second stage is triggered immediately. The second stage is the intelligent parameter control model, which adopts an integrated and scalable expert strategy model library. For each preset typical working condition that can be identified in the first stage, there is a dedicated expert strategy sub-model. Each expert strategy sub-model also adopts a multi-layer sensing mechanism structure. Each expert strategy sub-model is trained through supervised learning. The training data comes from the case data set of successful handling of various working conditions in the historical construction process. S4. Preliminary Control of the Construction Process: After the drilling rig is in place and before formal drilling begins, preliminary control of the construction process is carried out, including the following steps: S401. Extraction of geological feature parameters: The operator extracts the geological feature vector from the geological survey report of the current project and inputs it into the initial parameter module. The module then calls its built-in geological initial parameter mapping model to perform forward calculation and outputs a set of initial setting values ​​for construction parameters for the current geological conditions of the drilled pile location. S402. Configuration of equipment control parameters: After obtaining the initial set values ​​of the construction parameters, they are sent to the drilling rig control system and the mud control system; the drilling rig control system automatically completes the initial setting and calibration of the initial drilling pressure, lower limit of the rotational speed range, upper limit of the rotational speed range, torque safety threshold, and initial drilling speed; the mud control system automatically adjusts the initial mud density and initial mud viscosity, and sets the initial liquid level of the mud circulation system. S403 Drilling Operation: After the initial configuration and two-way verification of all equipment parameters are completed, the drilling rig can start drilling operations with the optimized parameters. From the moment the drill bit contacts the ground, the data monitoring module immediately starts high-frequency data acquisition, and the entire system synchronously enters a closed-loop intelligent control cycle consisting of real-time working condition diagnosis and dynamic parameter adjustment. S5. Dynamic prediction of the construction process: During drilling, the data monitoring module continuously collects and updates the monitoring parameter set at a frequency of no less than 10Hz. P Meanwhile, the real-time diagnostic model for operating conditions in the real-time control module uses the set of monitoring parameters collected during the current control cycle. P Based on this, pattern recognition calculations are continuously performed; the real-time working condition diagnostic model outputs the classification results of the current drilling working condition and its corresponding probability values; the intelligent parameter control model in the real-time control module adjusts the parameters according to the monitoring parameter set. P The output of the real-time diagnostic model of the working condition is used for forward calculation and reasoning to output a structured set of control instructions. This set of instructions clearly specifies the target set value or the adjustment range relative to the current value of each key construction parameter. S6. Dynamic control of the construction process: First, the control command set output by the real-time control module is checked for safety to ensure that all parameter adjustments are within the allowable working range of the equipment; the verified control commands are sent to the drilling rig control system and mud control system in parallel in the form of digital signals through the standardized equipment execution interface. After the control command is executed, the data monitoring module immediately initiates a new round of high-frequency data acquisition to obtain the real-time monitoring parameter set following the command execution. P The system compares the preset target parameter values ​​with the actual parameter values ​​fed back by the equipment to analyze the execution deviation. At the same time, the system packages the control command issued this time, the corresponding working condition diagnostic label, and the monitoring data after execution together to form a complete record. S7. Continuous Model Optimization: After each complete construction control cycle, the intelligent control system automatically integrates the data generated in this cycle into a standardized training sample data package; the model optimization process of the intelligent control system is automatically started by preset trigger conditions; the trigger conditions are set to complete a set number of borehole constructions, or when the total number of stored training samples reaches a predetermined threshold; the optimization process uses multiple accumulated historical training sample data packages as training sets and adopts supervised learning to update the parameters of the system's core model.

2. The construction control method for large-diameter bored piles in hillside soil-rock mixture strata according to claim 1, characterized in that, In step S2, the geological initial parameter mapping model is implemented using a feedforward neural network architecture; the network structure includes an input layer, two hidden layers, and an output layer; the number of nodes in the input layer corresponds to the number of geological feature parameters; the first hidden layer contains 64 neurons and uses the ReLU activation function; the second hidden layer contains 32 neurons and also uses the ReLU activation function; the number of nodes in the output layer corresponds to the number of construction parameters to be recommended and uses a linear activation function. During the model training phase, structured geological feature parameters extracted from geological survey reports of a large number of completed projects are used as input features, and the corresponding optimal initial construction parameter combinations verified in practice are used as training labels. The training process uses the Adam optimizer with an initial learning rate of 0.001, a mean squared error loss function, a batch size of 32, and at least 100 training epochs. Early stopping is used to prevent overfitting. The input geological feature vectors need to be standardized and preprocessed, and their specific components include stratigraphic lithology complexity score, volumetric stone content, maximum boulder size, weak interlayer thickness, groundwater depth, and groundwater activity score. The model outputs initial construction parameters, which are a 7-dimensional vector, corresponding to the initial settings of the following construction parameters: initial drilling pressure, lower limit of rotation speed range, upper limit of rotation speed range, torque safety threshold, initial drilling speed, initial mud density, and initial mud viscosity.

3. The construction control method for large-diameter bored piles in hillside soil-rock mixture strata according to claim 1, characterized in that, In step S7, the parameter update method is as follows: First, the real-time control module is optimized, using the control decision input features in the training samples as model input, and the control decision output results, i.e., the actual issued parameter adjustment instructions that have been verified as effective on-site, as the training target label; by minimizing the error between the model's predicted instructions and the actual effective instructions, and using the error backpropagation algorithm, the network weight parameters inside the real-time diagnostic model and the intelligent parameter control model are updated; Second, the geological initial parameter mapping model in the initial parameter module is optimized, and the optimized training uses the standardized geological feature parameters of various geological boreholes as input, and the initial combination of construction parameters verified in field practice as the training target label.

Images