Construction coordinate control method and system for pile driving operation

Abstract

The present invention relates to a construction coordinate control method and system based on a pile driving operation, and relates to the field of civil engineering pile driving control, including: providing a construction coordinate control method and system for pile driving operations, and enhancing the anti-interference ability and stability of the system by utilizing advanced signal processing technologies such as Kalman filters, ensuring that good monitoring effects can still be maintained in various complex environments, and enhancing the robustness of the system. By developing a complete set of pile position monitoring systems, automated management of the entire process from data collection and processing to final result output is achieved, manual intervention is reduced, work efficiency is improved, and automated monitoring is achieved. Through high-precision pile position information, reliable data support is provided for the planning, construction, and later maintenance of engineering projects, which assists in engineering quality management and cost control, provides decision-making support, and solves the technical problem that the existing pile position monitoring methods have low accuracy in actual applications, affecting engineering quality and construction efficiency.

Description

Technical Field

[0001] The present invention relates to the field of civil engineering piling control, and in particular to a construction coordinate control method and system for piling operations. Background Art

[0002] In modern civil engineering, piling operations are a crucial step in foundation construction, playing a key role in building stability and safety. Traditional methods for monitoring pile position rely primarily on single sensors (such as GPS) for positioning. However, due to the limitations of single sensors, such as signal interference and environmental noise, achieving the desired level of accuracy is often difficult. Furthermore, the changes in the hydraulic arm and vehicle body posture during excavator operation, as well as the high-intensity mechanical vibrations of the piling chuck (vibratory hammer), make it impractical to install sensors directly on the pile head, further increasing the monitoring difficulty.

[0003] Although existing multi-sensor fusion technology has been applied in some fields, its application in pile driving position monitoring is still immature. Common problems include:

[0004] 1. Sensor data synchronization problem: The data collection time and frequency of different sensors are inconsistent, resulting in time mismatch during data fusion.

[0005] 2. Data processing complexity: The processing and fusion of multi-sensor data require complex algorithm support, and existing data processing methods are inefficient when processing large amounts of data.

[0006] 3. Error accumulation problem: In the multi-step calculation process, the error of each step will accumulate, eventually leading to a large positioning error.

[0007] 4. Poor environmental adaptability: Existing technologies have poor adaptability in complex environments and are easily affected by factors such as weather and terrain.

[0008] The existence of these problems and shortcomings makes it difficult for existing pile position monitoring methods to meet high-precision requirements in practical applications, and there are technical problems that affect project quality and construction efficiency. Summary of the Invention

[0009] The present invention aims to solve the technical problem that the pile driving position monitoring method in the prior art has low accuracy in actual application, which affects the project quality and construction efficiency, and provides a construction coordinate control method and system for pile driving operations to solve the problem.

[0010] The technical solution of the present invention to solve the above technical problems is as follows:

[0011] In a first aspect, the present invention provides a construction coordinate control method for a pile driving operation, comprising:

[0012] Obtaining second constructed pile sensing information in response to first constructed pile sensing information uploaded from the sensor array using an extended Kalman filter rule;

[0013] performing error compensation on the second constructed pile driving perception information through a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile driving perception information;

[0014] Extracting a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, performing linear modeling, and obtaining a pile top coordinate prediction function;

[0015] The designed coordinates of the pile bottom are received, processed by the pile top coordinate prediction function, and the predicted coordinates of the pile top are output to execute construction control.

[0016] In a second aspect, the present invention provides a construction coordinate control system for a pile driving operation, comprising:

[0017] A second constructed pile sensing information module is configured to obtain second constructed pile sensing information in response to the first constructed pile sensing information uploaded from the sensor array using an extended Kalman filter rule;

[0018] a third constructed pile perception information module, configured to perform error compensation on the second constructed pile perception information by using a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile perception information;

[0019] A pile top coordinate prediction function module is used to extract a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, perform linear modeling, and obtain a pile top coordinate prediction function;

[0020] The pile top prediction coordinate module is used to receive the designed coordinates of the pile bottom, process them through the pile top coordinate prediction function, and output the predicted coordinates of the pile top to perform construction control.

[0021] The present invention provides a construction coordinate control method and system for pile driving operations. Compared to traditional pile position monitoring methods, this system utilizes multi-sensor data acquisition and fusion, integrating multiple sensors (such as GPS modules, IMUs, and laser rangefinders) to achieve comprehensive monitoring of pile positions and generate high-precision raw pile position data. Furthermore, by applying Kalman filtering technology, a Kalman filter model suitable for pile position monitoring is designed to dynamically filter sensor data and reduce errors caused by environmental noise and other interference factors. Furthermore, through the design of an error compensation mechanism, a neural network-based error compensation model based on an extreme learning machine (ELM) is established to address systematic deviations in pile position data. An adaptive learning rate adjustment strategy is introduced to enable online updating of the ELM model with real-time data, ensuring error compensation accuracy in dynamic scenarios. Furthermore, through system performance evaluation and optimization, an experimental platform comprising multiple sensors and a real-time monitoring module is constructed to comprehensively test the proposed monitoring system. Key system parameters (such as the filter window size and the ELM network structure) are adjusted based on the experimental data to ensure high accuracy and stability under various operating conditions.

[0022] Furthermore, the present invention enhances the anti-interference ability and stability of the system by utilizing advanced signal processing technologies such as Kalman filters, ensuring that good monitoring effects can be maintained in various complex environments and enhancing the robustness of the system. By developing a complete set of pile position monitoring systems, the whole process from data acquisition and processing to final result output is automated, reducing manual intervention, improving work efficiency, and realizing automated monitoring. Through high-precision pile position information, reliable data support is provided for the planning, construction and later maintenance of engineering projects, assisting in project quality management and cost control, and providing decision-making support, thereby achieving the technical effect of improving the accuracy and stability of pile position monitoring while taking into account the realization of cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic flow chart of a construction coordinate control method for a pile driving operation provided by the present invention;

[0024] Figure 2 A schematic structural diagram of a construction coordinate control system for piling operations provided by the present invention;

[0025] Reference numerals: second constructed pile driving perception information module 11 , third constructed pile driving perception information module 12 , pile top coordinate prediction function module 13 , pile top prediction coordinate module 14 . DETAILED DESCRIPTION

[0026] 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 those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0027] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0028] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0029] Example 1:

[0030] like Figure 1 As shown, an embodiment of the present invention provides a construction coordinate control method for a pile driving operation, which specifically includes the following steps:

[0031] S10: Obtaining second constructed pile perception information in response to the first constructed pile perception information uploaded from the sensor array using an extended Kalman filter rule;

[0032] Furthermore, obtaining second constructed pile sensing information in response to the first constructed pile sensing information uploaded from the sensor array by using an extended Kalman filter, executing step S10 includes:

[0033] S11: The first constructed pile driving perception information includes constructed pile driving perception information at the first moment and constructed pile driving perception information at the Nth moment;

[0034] S12: Perform extended Kalman filtering on the piling perception information obtained at the first moment and the piling perception information obtained at the Nth moment to obtain the corrected piling perception information obtained at the first moment until the corrected piling perception information obtained at the Nth moment, and add the second piling perception information obtained at the Nth moment.

[0035] Furthermore, the first constructed piling perception information includes the top coordinates of the excavator's cab, the root coordinates of the robotic arm, the upper arm posture angle information, the middle arm acceleration information, the middle arm posture angle information, the middle arm acceleration information, the forearm posture angle information, and the forearm acceleration information.

[0036] Furthermore, the above coordinate information can be obtained by the following method:

[0037] Two high-precision RTK receivers are installed on the excavator, one on top of the cockpit and the other at the base of the robotic arm. RTK coordinate data is used to infer the coordinates of the boom tip. An inertial measurement unit (IMU) is installed on each arm. Comprising an accelerometer and gyroscope, it provides attitude, angular velocity, and acceleration data, enabling dynamic attitude angle measurement. The coordinates of the mid-arm connection point are calculated by combining the lengths and attitude angles of the boom and mid-arm. Similarly, the spatial coordinates of the pile tip are calculated based on the data from the mid-arm and forearm. After the pile is driven, the exact coordinates of the pile tip are recorded.

[0038] Furthermore, by traversing the perception information of the piles constructed at the first moment and the perception information of the piles constructed at the Nth moment and performing extended Kalman filtering processing, corrected perception information of the piles constructed at the first moment is obtained until corrected perception information of the piles constructed at the Nth moment, and executing step S12 includes:

[0039] S121: Obtaining preset perception information of the sensor array at the initial moment, and constructing an initial moment state estimation value and an initial moment state covariance matrix;

[0040] S122: performing time update according to the initial moment state estimation value and the initial moment state covariance matrix to obtain a first moment state prediction value and a first moment state covariance matrix;

[0041] S123: Based on the first-time state prediction value and the first-time state covariance matrix, combined with the first-time constructed pile driving perception information, measurement update is performed to obtain the first-time state estimation value and the first-time state covariance matrix, and add the first-time constructed pile driving correction perception information;

[0042] S124: until the extended Kalman filter processing is performed based on the state estimation value at the N-1th moment and the state covariance matrix at the N-1th moment, combined with the pile driving perception information at the Nth moment, to obtain the corrected pile driving perception information at the Nth moment.

[0043] Furthermore, the present invention first requires data collection through a high-precision sensor system to ensure that accurate location information is obtained. Specifically, the installation of a high-precision RTK receiver can obtain the geographic coordinates and attitude information of the entire excavator in real time. The centimeter-level positioning capability of RTK technology can calculate the three-dimensional position coordinates of the end of the excavator's boom, providing accurate reference data for subsequent steps. In addition, IMU sensors are installed on the excavator's boom, middle arm and forearm respectively, which can obtain the attitude angle (pitch angle, roll angle and yaw angle) and dynamic acceleration data of each arm in real time. Combined with the length parameters of the boom and middle arm, the spatial position of each connection point is dynamically calculated through the data collected by the IMU. If construction requires, a laser rangefinder or other displacement sensor can also be installed on the top of the pile to further improve the real-time monitoring accuracy of the spatial position of the pile head.

[0044] To ensure data is used in the same time reference frame, a timestamp mechanism and synchronization protocol are used to ensure synchronization of RTK and IMU data. Coordinate system conversion is used to unify the data into the excavator's central coordinate system to ensure consistency.

[0045] To improve data reliability, raw data must be verified and preprocessed, including outlier detection and noise reduction. Statistical methods can be used to detect outliers in sensor output (such as sudden changes in RTK coordinates or IMU angles), thereby eliminating invalid data. RTK data can be clipped or deviations detected based on historical trajectories. IMU data can be analyzed for angular velocity and acceleration continuity to detect anomalous changes. RTK and IMU data must also be low-pass filtered to remove high-frequency noise caused by environmental noise and electromagnetic interference. A sliding window method or simple mean filter can be used to smooth sensor data and reduce fluctuations. IMU and RTK data are cross-validated to ensure consistency, and re-collected if they exceed the tolerance range.

[0046] After data acquisition and verification, the system outputs real-time attitude angles and position information for the excavator's boom, mid-arm, and forearm, along with preliminary positioning coordinates for the pile head, serving as input for subsequent steps. It also outputs data reliability metrics (such as noise level and data consistency score) for data quality assessment. Based on this data, the system performs pile position positioning and control in subsequent steps, ensuring accurate and compliant pile construction.

[0047] Furthermore, data is distributed and processed during piling operations. Before piling operations begin, target point data is sent to the piling equipment to guide the operation. Simultaneously, relevant sensor data (such as coordinates, posture, and pile head position) is transmitted to the equipment's control system in real time, ensuring that the equipment executes the task according to the planned path.

[0048] To ensure effective fusion of data collected by different sensors, the timestamps of all sensors must be synchronized to avoid data inconsistencies caused by time differences. Based on this, a central controller or server synchronously collects data from each sensor and uses data fusion algorithms such as weighted averaging, Bayesian fusion, or neural networks to combine the data from different sensors, improving overall operation accuracy and reliability.

[0049] During the piling process, the equipment monitors the precise position, depth, inclination angle, and other data of the pile head in real time and compares them with the target position. If the deviation exceeds the allowable range, the system automatically adjusts the robotic arm's posture through a control algorithm to ensure that the pile head remains in the correct position. Furthermore, a real-time data feedback mechanism transmits deviation information to the central control system, enabling dynamic adjustment of operating parameters to ensure construction accuracy. After the piling data processing is completed, the system outputs real-time monitoring data, including information such as the pile head position, depth, and inclination angle, and generates a data quality assessment report. Based on the assessment results, the system adjusts control parameters to optimize operating precision and ensure accuracy and stability during the construction process.

[0050] This invention designs and integrates multiple sensors (such as GPS modules, IMUs, laser rangefinders, etc.) to achieve all-round monitoring of pile driving positions; studies the synchronous acquisition technology of sensor data to ensure data consistency and accuracy; and develops a data fusion algorithm to integrate data from different sources into a unified coordinate system to form high-precision original pile position data.

[0051] Furthermore, based on the first-time state prediction value and the first-time state covariance matrix, combined with the first-time constructed pile driving perception information, measurement update is performed to obtain the first-time state estimation value and the first-time state covariance matrix, and the first-time constructed pile driving correction perception information is added to the first-time constructed pile driving correction perception information. Executing step S123 includes:

[0052] S1231: The first moment state estimation value includes the cockpit top coordinate estimation value, the manipulator arm root coordinate estimation value, the upper arm attitude angle estimation value, the middle arm acceleration estimation value, the middle arm attitude angle estimation value, the middle arm acceleration estimation value, the forearm attitude angle estimation value, and the forearm acceleration estimation value;

[0053] S1232: Based on the cockpit top coordinate estimate, the manipulator arm root coordinate estimate, the boom attitude angle estimate, the middle arm acceleration estimate, the middle arm attitude angle estimate, the middle arm acceleration estimate, the forearm attitude angle estimate, and the forearm acceleration estimate, the boom length, the middle arm length, and the forearm length are combined to perform spatial coordinate unification to obtain the boom end coordinate, the middle arm connection point coordinate, the pile bottom spatial coordinate, and the pile top spatial coordinate, and set them as the quaternion attitude estimate at the first moment;

[0054] S1233: Add the first-moment state estimation value, the first-moment state covariance matrix, and the first-moment four-element attitude estimation value to the first-moment constructed pile driving correction perception information.

[0055] S20: performing error compensation on the second constructed pile driving perception information through a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile driving perception information;

[0056] Furthermore, error compensation is performed on the second constructed pile driving perception information through a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile driving perception information, and the execution steps include:

[0057] S21: extracting a historical filtered verified log processed by the extended Kalman filter rule for the sensor array, wherein the historical filtered verified log includes a first sensor system residual, a first sensor state covariance matrix estimate, and a first sensor filter gain, up to a Q-th sensor system residual, a Q-th sensor state covariance matrix estimate, and a Q-th sensor filter gain, and a label identifying a compensation value for a deviation between filtered perception data and true perception data;

[0058] S22: Using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, to train the filter error fitting network.

[0059] Furthermore, the simplified adaptive extended Kalman filter processing method is as follows:

[0060] During the piling process, the equipment vibrates, and the received information contains a large amount of noise. This makes it impossible to determine the final construction point. Using a Kalman filter, we remove noise from the sensor data, improve data quality and stability, and ultimately locate the top coordinates of the construction pile.

[0061] The extended Kalman filter is a generalized model of the Kalman filter theory in nonlinear systems, which enables the filter estimation algorithm to be applicable to application scenarios under Gaussian nonlinear systems. The main process is divided into two steps: time update and measurement update.

[0062] Time update: During the time update process, the state estimate value at the previous moment is required and the state covariance matrix , calculate the state prediction value at the current moment and the state covariance matrix .

[0063]

[0064]

[0065] in, , They are called the state transfer matrix and the state covariance noise matrix respectively.

[0066] Measurement update: During the measurement update process, the predicted value obtained in the time update is required , the state covariance matrix and measured values , determine the current state estimate and the state covariance matrix ( is the identity matrix).

[0067]

[0068]

[0069] in, is the measurement transfer matrix, and It is called the Kalman gain matrix and its specific form is as follows:

[0070]

[0071] in, is the covariance matrix of the measurement noise.

[0072] Once the current state estimate is obtained , the current time can be calculated according to the following formula The attitude quaternion and deviation values :

[0073]

[0074]

[0075] As mentioned above, the extended Kalman filter can adapt to the noise characteristics of weak nonlinear systems and output higher attitude data. However, due to the state covariance matrix in the filtering process The amount of calculation is very large, which will reduce the overall algorithm efficiency. Therefore, the present invention adopts a simplified calculation scheme. Its basic idea is to use Cholesky decomposition to simplify the matrix operations in the extended Kalman filter update process, reduce the computational complexity within the algorithm, and also use Cholesky decomposition to ensure the semi-positive definiteness of the covariance matrix, thereby avoiding the problems of non-convergence and biased estimation that may occur in nonlinear systems, improving the stability of the linear filtering framework, and increasing the estimation accuracy.

[0076] The specific simplified scheme is as follows: During the time update, for the state covariance matrix Perform specific operations. Define the covariance matrix The decomposition form is as follows:

[0077]

[0078] Then, suppose ,therefore:

[0079]

[0080] Further,

[0081]

[0082] Make this place , , so we have:

[0083]

[0084] In order to meet the conditions , the following equation is obtained:

[0085]

[0086] Then get Substituting the parameters in the above formula into the original filtering process can simplify the matrix calculation in the update process.

[0087] The present invention enhances the anti-interference ability and stability of the system by utilizing advanced signal processing technologies such as Kalman filter, ensuring that good monitoring effects can be maintained in various complex environments and enhancing the robustness of the system.

[0088] Further, using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, the filter error fitting network is trained, and executing step S22 includes:

[0089] S221: Initialize the number of hidden layer neurons of the extreme learning machine to be the same as the dimension of the input data, and obtain a first-level extreme learning machine network architecture;

[0090] S222: Constructing a filtering error fitting loss function, wherein the filtering error fitting loss function represents a mean square error between the perception data adjusted by the compensation value predicted by the filtering error fitting and the actual perception data;

[0091] S223: Using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, to train the first extreme learning machine network architecture to obtain a first-level filter error fitting network;

[0092] S224: When the loss value of the first-level filter error fitting network is greater than or equal to the loss threshold, increase the number of hidden layer neurons until a Y-level filter error fitting network with a loss value less than the loss threshold is obtained, and set it as the filter error fitting network;

[0093] S225: When the loss value of the first-level filter error fitting network is less than a loss threshold, setting the first-level filter error fitting network as the filter error fitting network.

[0094] Furthermore, the present invention uses an artificial neural network method to solve the problem of nonlinear mapping, and adopts an error compensation network based on an extreme learning machine. The network can learn the linearization loss in the filtering process from the training data and predict the initial value of the filter. and the true pose value The difference between the two is defined as the compensation value .

[0095] The inputs of the error compensation network based on the extreme learning machine are: system residual , state covariance matrix estimation , and the filter gain , and the output of the network is .

[0096] like Figure 2 As shown, assuming that a set of filtering parameters in the data set is ,in is the input vector, is the output vector, the specific form is:

[0097]

[0098]

[0099] The present invention adopts a specific training strategy to screen important parameters, and the strategy is divided into two parts: a dedicated quantization strategy and a matrix principal component selection.

[0100] The dedicated quantization strategy refers to normalizing each input matrix according to its order of magnitude, which is beneficial to the training of the neural network. In addition, this network selects the Sigmoid function as the activation function, and all inputs are mapped to a fixed interval [0,1], thereby enhancing the network's ability to express complex nonlinear systems. In addition, the matrix principal component selection refers to selecting only the main diagonal elements as network inputs based on the characteristics of the Kalman gain matrix and the state covariance matrix itself. Therefore, when training the error compensation network, the present invention only retains the system residuals. , state covariance matrix estimation , and the filter gain The main diagonal elements of , the final input of the network for:

[0101]

[0102] in, Represents the estimated value of the state Dimensions, Indicates measured values dimension.

[0103] The network training phase is calculated through the following steps: First, determine the number of hidden layer neurons , whose initial value is compatible with the input data dimension, and then randomly initialize the network input layer weights , output layer weight and the hidden layer bias value , construct the output matrix of the hidden layer and the weight matrix of the output layer The final calculation network output is the attitude compensation value , and calculate the network loss at the same time. If the loss value meets the requirements, the training is stopped and the network hyperparameters are , and If the loss value does not meet the requirements, the number of hidden layer neurons is increased and the above steps are repeated.

[0104] The network testing phase uses the network hyperparameters obtained through training , and Build a network model and use test data Predict the output of each network layer and determine the posture compensation , calculate the network loss and evaluate the network performance. The loss function of the error compensation network based on the extreme learning machine is defined as the mean square error (MSE) between the network prediction value and the true posture error. Since the posture values ​​are all represented by quaternions, the difference is calculated using the quaternion operation rules, specifically:

[0105]

[0106] Specifically, based on quaternion multiplication, the loss function can be expressed as:

[0107]

[0108] In order to improve the accuracy of the final posture determination, the present invention adopts a weighted posture smoothing module, which includes a weighted fusion method and a posture smoothing filter. A learnable weight factor is added in the smooth fusion process. , which can correct the influence of the error compensation network prediction value on the final attitude output value. The specific process of the weighted fusion method can be described by the following quaternion calculation process:

[0109]

[0110] in, Represents the initial value of attitude prediction output by the simplified adaptive extended Kalman filter module; is the attitude compensation value from the error compensation network.

[0111] The weighted fusion attitude value will be input into the attitude smoothing filter proposed in the present invention, and the simplified adaptive extended Kalman filter will be used to determine the initial state of the smoothing process, that is, to calculate the initial state smoothing value and the initial covariance matrix smoothed values Following the first-order Markov assumption, the state estimated smooth value and covariance matrix of the previous moment can be inferred from the state smooth value of the previous moment and the smooth result of the current moment, and then executed in reverse chronological order. The detailed calculation formula is as follows:

[0112]

[0113]

[0114]

[0115] Based on the Extreme Learning Machine (ELM), a neural network-based error compensation model is established to address systematic deviations in pile position data. The ELM model is trained using historical pile position data to optimize network weights and activation functions, ensuring the generalization of the error compensation model. An adaptive learning rate adjustment strategy is introduced to enable online updates of the ELM model to real-time data, ensuring error compensation accuracy in dynamic scenarios.

[0116] S30: extracting a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, performing linear modeling, and obtaining a pile top coordinate prediction function;

[0117] Furthermore, a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence are extracted from the third constructed pile driving posture information, and linear modeling is performed to obtain a pile top coordinate prediction function. Execution of step S30 includes:

[0118] S31: extracting a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence according to the third constructed pile driving posture information, performing linear modeling by a least square method or an extension line method, and obtaining the pile top coordinate prediction function.

[0119] Furthermore, based on the initial filtering values, an algorithm fitting is performed on the pile position data. By analyzing the coordinates of the constructed pile tops within a group of piles, the optimal processing is typically performed using multiple algorithms such as the extension line method and the least squares method. The construction positions of the piles within the same group are adjusted to ensure that their pile tops are aligned in a straight line and are evenly spaced. The following are two specific examples.

[0120] Extension Line Method: If a pile has already been driven, the first pile location is connected to the pile with the longest distance from it. Otherwise, the line connecting the two closest piles is used. A two-dimensional line equation is constructed using the coordinates of the two located points. The x-coordinates of each subsequent pile are used to predict its y-coordinate and determine the top coordinate of the pile. A single pile prediction is performed for each subsequent pile on the line, and the predicted data is returned.

[0121] Least Squares Method: Summarize the coordinate information of driven and undriven piles, put the data set into the least squares algorithm for calculation, and virtualize a two-dimensional straight line equation. Using the x-coordinate of each next pile to be driven, its y-coordinate is predicted to obtain the pile top coordinate, and then a single pile judgment is performed.

[0122] Assume that the pile position data point is , the equation of the fitted straight line is .

[0123] Define the error function ,in is the number of data points.

[0124] By solving the error function The minimum value of the best fitting straight line is obtained and .

[0125] Solve using the method of partial derivatives: and Solving the equations yields

[0126]

[0127]

[0128] S40: receiving the designed coordinates of the pile bottom, processing them through the pile top coordinate prediction function, and outputting the predicted coordinates of the pile top to execute construction control.

[0129] Furthermore, the designed coordinates of the pile bottom are received, processed by the pile top coordinate prediction function, and the predicted coordinates of the pile top are output to perform construction control. The execution of step S40 further includes:

[0130] S41: Obtain the piling construction constraint rules:

[0131] Rule 1: The deviation between the pile bottom and the designed pile bottom distance cannot exceed 30 mm;

[0132] Rule 2: The verticality of the pile itself cannot exceed 1%, where the verticality of the pile itself is equal to the length of the shadow from the top to the bottom of the pile / the height of the pile * 100%;

[0133] S42: Verify the constructed pile positions and the predicted pile positions according to the piling construction constraint rules. If an abnormality is found, mark the associated pile positions as abnormal and issue an early warning.

[0134] Furthermore, both the piles that have been constructed and the piles to be driven need to comply with the design standards of the site, so it is necessary to judge the construction specifications of all piles in the same group of piles.

[0135] For pile positions that have been driven:

[0136] The distance between the bottom of the driven pile and the designed pile bottom cannot exceed 30 mm.

[0137] The distance between the top of the driven pile and the designed shadow of the pile bottom cannot exceed 30 mm.

[0138] The verticality of the pile itself cannot exceed one percent (the calculation formula is the shadow length from the top of the pile to the bottom of the pile / pile height*100%).

[0139] For the predicted pile:

[0140] The distance between the predicted pile top and the designed pile bottom shadow cannot exceed 30 mm.

[0141] The predicted verticality of the pile itself cannot exceed one percent (calculated as the shadow length from the top of the pile to the bottom of the pile / pile height*100%).

[0142] The present invention provides reliable data support for the planning, construction and post-maintenance of engineering projects through high-precision pile position information, thereby facilitating project quality management and cost control.

[0143] Example 2, as Figure 2 As shown, based on the same inventive concept as the construction coordinate control method for piling operation provided in the first embodiment, the embodiment of the present invention further provides a construction coordinate control system for piling operation, comprising:

[0144] A second constructed pile sensing information module is configured to obtain second constructed pile sensing information in response to the first constructed pile sensing information uploaded from the sensor array using an extended Kalman filter rule;

[0145] a third constructed pile perception information module, configured to perform error compensation on the second constructed pile perception information by using a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile perception information;

[0146] A pile top coordinate prediction function module is used to extract a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, perform linear modeling, and obtain a pile top coordinate prediction function;

[0147] The pile top prediction coordinate module is used to receive the designed coordinates of the pile bottom, process them through the pile top coordinate prediction function, and output the predicted coordinates of the pile top to perform construction control.

[0148] Furthermore, the second constructed piling perception information module includes:

[0149] The first constructed pile driving perception information includes constructed pile driving perception information at the first moment and constructed pile driving perception information at the Nth moment;

[0150] By traversing the perception information of the pile driving constructed at the first moment and the perception information of the pile driving constructed at the Nth moment and performing extended Kalman filtering processing, the corrected perception information of the pile driving constructed at the first moment is obtained until the corrected perception information of the pile driving constructed at the Nth moment, and the second perception information of the pile driving constructed is added.

[0151] Furthermore, the second constructed piling perception information module further includes:

[0152] Obtaining the preset perception information of the sensor array at the initial moment, and constructing the initial moment state estimation value and the initial moment state covariance matrix;

[0153] Performing time updating according to the initial moment state estimation value and the initial moment state covariance matrix to obtain a first moment state prediction value and a first moment state covariance matrix;

[0154] Based on the first moment state prediction value and the first moment state covariance matrix, combined with the first moment constructed pile driving perception information, measurement update is performed to obtain the first moment state estimation value and the first moment state covariance matrix, and the first moment state correction perception information is added to the first moment constructed pile driving;

[0155] Until the extended Kalman filter processing is performed based on the state estimation value at the N-1th moment and the state covariance matrix at the N-1th moment, combined with the perception information of the pile driving constructed at the Nth moment, the corrected perception information of the pile driving constructed at the Nth moment is obtained.

[0156] Furthermore, the second constructed piling perception information module further includes:

[0157] The first moment state estimation value includes the cockpit top coordinate estimation value, the manipulator arm root coordinate estimation value, the upper arm attitude angle estimation value, the middle arm acceleration estimation value, the middle arm attitude angle estimation value, the middle arm acceleration estimation value, the forearm attitude angle estimation value, and the forearm acceleration estimation value;

[0158] Based on the cockpit top coordinate estimate, the manipulator arm root coordinate estimate, the boom attitude angle estimate, the middle arm acceleration estimate, the middle arm attitude angle estimate, the middle arm acceleration estimate, the forearm attitude angle estimate, and the forearm acceleration estimate, the boom length, the middle arm length, and the forearm length are combined to perform spatial coordinate unification to obtain the boom end coordinate, the middle arm connection point coordinate, the pile bottom spatial coordinate, and the pile top spatial coordinate, which are set as the quaternion attitude estimate at the first moment;

[0159] The first moment state estimation value, the first moment state covariance matrix and the first moment four-element attitude estimation value are added to the first moment constructed pile driving correction perception information.

[0160] Furthermore, the third constructed piling perception information module includes:

[0161] Extracting a historical filtered verified log of the sensor array processed by an extended Kalman filter rule, wherein the historical filtered verified log includes a first sensor system residual, a first sensor state covariance matrix estimate, and a first sensor filter gain, up to a Qth sensor system residual, a Qth sensor state covariance matrix estimate, and a Qth sensor filter gain, and a label identifying a compensation value for a deviation between filtered perception data and true perception data;

[0162] The filter error fitting network is trained using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain.

[0163] Furthermore, the third constructed piling perception information module also includes:

[0164] Initialize the number of hidden layer neurons of the extreme learning machine to be the same as the dimension of the input data, and obtain the first-level extreme learning machine network architecture;

[0165] Constructing a filter error fitting loss function, wherein the filter error fitting loss function represents the mean square error between the perception data adjusted by the compensation value predicted by the filter error fitting and the actual perception data;

[0166] Using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, the first extreme learning machine network architecture is trained to obtain a first-level filter error fitting network;

[0167] When the loss value of the first-level filter error fitting network is greater than or equal to the loss threshold, the number of hidden layer neurons is increased until a Y-level filter error fitting network with a loss value less than the loss threshold is obtained, which is set as the filter error fitting network;

[0168] When the loss value of the first-level filter error fitting network is less than a loss threshold, the first-level filter error fitting network is set as the filter error fitting network.

[0169] Furthermore, the pile top coordinate prediction function module includes:

[0170] A constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence are extracted according to the third constructed pile driving posture information, and linear modeling is performed using the least squares method or the extension line method to obtain the pile top coordinate prediction function.

[0171] Furthermore, the pile top prediction coordinate module includes:

[0172] Obtain the piling construction constraint rules:

[0173] Rule 1: The deviation between the pile bottom and the designed pile bottom distance cannot exceed 30 mm;

[0174] Rule 2: The verticality of the pile itself cannot exceed 1%, where the verticality of the pile itself is equal to the verticality from pile top to pile;

[0175] Bottom shadow length / pile height*100%;

[0176] The constructed pile positions and predicted pile positions are checked according to the piling construction constraint rules. If any anomalies are found, the associated pile positions are marked as abnormal and an early warning is issued.

[0177] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0178] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0179] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0180] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0181] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0182] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0183] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A construction coordinate control method for pile driving operation, characterized in that: include: Obtaining second constructed pile sensing information in response to first constructed pile sensing information uploaded from the sensor array using an extended Kalman filter rule; performing error compensation on the second constructed pile driving perception information through a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile driving perception information; Extracting a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, performing linear modeling, and obtaining a pile top coordinate prediction function; receiving the designed coordinates of the pile bottom, processing them through the pile top coordinate prediction function, and outputting the predicted coordinates of the pile top to execute construction control; The method of performing error compensation on the second constructed pile driving perception information by using a filter error fitting network bound to an extended Kalman filter rule to obtain the third constructed pile driving perception information includes: Extracting a historical filtered verified log of the sensor array processed by an extended Kalman filter rule, wherein the historical filtered verified log includes a first sensor system residual, a first sensor state covariance matrix estimate, and a first sensor filter gain, up to a Qth sensor system residual, a Qth sensor state covariance matrix estimate, and a Qth sensor filter gain, and a label identifying a compensation value for a deviation between filtered perception data and true perception data; Using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, to train the filter error fitting network; The method uses the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and uses the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, to train the filter error fitting network, including: Initialize the number of hidden layer neurons of the extreme learning machine to be the same as the dimension of the input data, and obtain the first-level extreme learning machine network architecture; Constructing a filter error fitting loss function, wherein the filter error fitting loss function represents the mean square error between the perception data adjusted by the compensation value predicted by the filter error fitting and the actual perception data; Using the label of the compensation value that identifies the deviation between the filtered perception data and the true perception data as supervision, and using the first sensor system residual, the first sensor state covariance matrix estimate, and the first sensor filter gain, up to the Qth sensor system residual, the Qth sensor state covariance matrix estimate, and the Qth sensor filter gain, the first extreme learning machine network architecture is trained to obtain a first-level filter error fitting network; When the loss value of the first-level filter error fitting network is greater than or equal to the loss threshold, the number of hidden layer neurons is increased until a Y-level filter error fitting network with a loss value less than the loss threshold is obtained, which is set as the filter error fitting network; When the loss value of the first-level filter error fitting network is less than a loss threshold, the first-level filter error fitting network is set as the filter error fitting network.

2. The method according to claim 1, wherein Obtaining second constructed pile perception information in response to first constructed pile perception information uploaded from a sensor array by using an extended Kalman filter, including: The first constructed pile driving perception information includes constructed pile driving perception information at the first moment and constructed pile driving perception information at the Nth moment; By traversing the perception information of the pile driving constructed at the first moment and the perception information of the pile driving constructed at the Nth moment and performing extended Kalman filtering processing, the corrected perception information of the pile driving constructed at the first moment is obtained until the corrected perception information of the pile driving constructed at the Nth moment, and the second perception information of the pile driving constructed is added.

3. The method according to claim 2, wherein The method performs extended Kalman filtering on the piling perception information at the first moment and the piling perception information at the Nth moment to obtain corrected piling perception information at the first moment until corrected piling perception information at the Nth moment, including: Obtaining the preset perception information of the sensor array at the initial moment, and constructing the initial moment state estimation value and the initial moment state covariance matrix; Performing time updating according to the initial moment state estimation value and the initial moment state covariance matrix to obtain a first moment state prediction value and a first moment state covariance matrix; Based on the first moment state prediction value and the first moment state covariance matrix, combined with the first moment constructed pile driving perception information, measurement update is performed to obtain the first moment state estimation value and the first moment state covariance matrix, and the first moment state correction perception information is added to the first moment constructed pile driving; Until the extended Kalman filter processing is performed based on the state estimation value at the N-1th moment and the state covariance matrix at the N-1th moment, combined with the perception information of the pile driving constructed at the Nth moment, the corrected perception information of the pile driving constructed at the Nth moment is obtained.

4. The method according to claim 3, wherein Based on the first-time state prediction value and the first-time state covariance matrix, combined with the first-time constructed pile driving perception information, measurement update is performed to obtain the first-time state estimation value and the first-time state covariance matrix, and the correction perception information of the first-time constructed pile driving is added, including: The first moment state estimation value includes the cockpit top coordinate estimation value, the manipulator arm root coordinate estimation value, the upper arm attitude angle estimation value, the middle arm acceleration estimation value, the middle arm attitude angle estimation value, the middle arm acceleration estimation value, the forearm attitude angle estimation value, and the forearm acceleration estimation value; Based on the cockpit top coordinate estimate, the manipulator arm root coordinate estimate, the boom attitude angle estimate, the middle arm acceleration estimate, the middle arm attitude angle estimate, the middle arm acceleration estimate, the forearm attitude angle estimate, and the forearm acceleration estimate, the boom length, the middle arm length, and the forearm length are combined to perform spatial coordinate unification to obtain the boom end coordinate, the middle arm connection point coordinate, the pile bottom spatial coordinate, and the pile top spatial coordinate, which are set as the quaternion attitude estimate at the first moment; The first moment state estimation value, the first moment state covariance matrix and the first moment four-element attitude estimation value are added to the first moment constructed pile driving correction perception information.

5. The method according to claim 1, wherein Extracting a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, performing linear modeling, and obtaining a pile top coordinate prediction function, including: A constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence are extracted according to the third constructed pile driving posture information, and linear modeling is performed using the least squares method or the extension line method to obtain the pile top coordinate prediction function.

6. The method according to claim 1, wherein The method further includes: receiving the designed coordinates of the pile bottom, processing them through the pile top coordinate prediction function, and outputting the predicted coordinates of the pile top to execute construction control; Obtain the piling construction constraint rules: Rule 1: The deviation between the pile bottom and the designed pile bottom distance cannot exceed 30 mm; Rule 2: The verticality of the pile itself cannot exceed 1%, where the verticality of the pile itself is equal to the verticality from pile top to pile; Bottom shadow length / pile height*100%; The constructed pile positions and predicted pile positions are checked according to the piling construction constraint rules. If any anomalies are found, the associated pile positions are marked as abnormal and an early warning is issued.

7. A construction coordinate control system for pile driving operation, characterized in that: Used to implement the method according to any one of claims 1 to 6, comprising: A second constructed pile perception information module is configured to obtain second constructed pile perception information in response to the first constructed pile perception information uploaded from the sensor array using an extended Kalman filter rule; a third constructed pile perception information module, configured to perform error compensation on the second constructed pile perception information by using a filter error fitting network bound to an extended Kalman filter rule to obtain third constructed pile perception information; A pile top coordinate prediction function module is used to extract a constructed pile bottom coordinate sequence and a constructed pile top coordinate sequence from the third constructed pile driving posture information, perform linear modeling, and obtain a pile top coordinate prediction function; The pile top prediction coordinate module is used to receive the designed coordinates of the pile bottom, process them through the pile top coordinate prediction function, and output the predicted coordinates of the pile top to perform construction control.

Images