Intelligent jacking system for bridge construction

Abstract

The invention discloses an intelligent jacking system for bridge construction. The intelligent jacking system comprises a data acquisition module, a control calculation module, an execution driving module and a self-learning optimization module. The data acquisition module acquires pressure value and displacement value data between the hydraulic oil cylinders and the bridge floor in real time to generate a pressure-displacement data matrix; the control calculation module calculates real-time stress and settlement state information of the bridge pile through an intelligent control algorithm library based on the matrix; and the execution driving module calls a preset control library according to the state information and generates a corresponding control instruction to drive the hydraulic system. And the self-learning optimization module dynamically optimizes control parameters by constructing a learning model of equipment working data.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge construction equipment, and particularly relates to an intelligent jacking system for bridge construction. Background Art

[0002] Bridge construction is a key link in modern transportation infrastructure construction, and its quality and safety are directly related to subsequent service performance and service life. During bridge construction, as the main load-bearing structure, the monitoring and control of the stress and settlement state of bridge piles are crucial. Traditional jacking systems mostly rely on manual experience for operation, making it difficult to achieve precise and real-time automatic control, and there are the following problems: mainly relying on manual observation and manual adjustment, it is impossible to accurately control the pressure and displacement of hydraulic cylinders, which easily leads to uneven stress or excessive settlement of bridge piles, affecting the stability and safety of the bridge structure; the bridge construction environment is complex and changeable, and traditional jacking systems lack the support of intelligent algorithms, making it difficult to automatically adjust control strategies according to different working conditions and unable to meet the requirements of modern bridge construction for intelligent and high-precision control; traditional systems cannot monitor the stress and settlement state of bridge piles in real time, making it difficult to detect potential safety hazards in a timely manner and unable to achieve preventive maintenance and risk warning; manual operation and manual adjustment are time-consuming and laborious, affecting the construction progress and making it difficult to meet the requirements of large bridge projects for efficient construction. Summary of the Invention

[0003] Based on this, it is necessary to provide an intelligent jacking system for bridge construction that can solve the above technical problems.

[0004] In a first aspect, the present application provides an intelligent jacking system for bridge construction, including: A data acquisition module for real-time obtaining pressure value data and displacement value data between a hydraulic cylinder and a bridge deck, and generating a pressure-displacement data matrix; A control calculation module for calculating the stress and settlement state of a bridge pile based on the pressure-displacement data matrix using an intelligent control algorithm library, and generating stress and settlement state information of the bridge pile; An execution drive module for generating corresponding control instructions based on the stress and settlement state information using a preset control library; A self-learning optimization module for constructing a learning model according to the working data of the jacking equipment, and dynamically updating the preset control library through the learning model and a parameter mapping mechanism.

[0005] In one embodiment, the control calculation module is further configured to: Based on the intelligent control algorithm library, combined with real-time working condition information, use the following formula to switch the intelligent control algorithm for calculating the stress and settlement state of the bridge pile: ; Wherein, The control algorithm selected at time t is denoted as, and i represents the index of the control algorithm. When i = 1, it corresponds to the fuzzy PID control algorithm; when i = 2, it corresponds to the neural network predictive control algorithm; when i = 3, it corresponds to the model predictive control MPC algorithm. represents the weight of the i-th control algorithm for the j-th operating condition parameter, and n represents the total number of operating condition parameters. represents the normalized value of the j-th operating condition parameter at time t: ; where, represents the real-time value of the j-th operating condition parameter. represents the minimum value of the j-th operating condition parameter. represents the maximum value of the j-th operating condition parameter.

[0006] In one embodiment, the control calculation module is further configured to: Calculate the force and settlement state of the bridge pile using the following model predictive control MPC algorithm: ; where, represents the optimal control input. represents the predicted output value. represents the desired output value. represents the weight of the control increment. and represents the output constraint. and represents the input constraint.

[0007] In one embodiment, the neural network predictive control algorithm in the control calculation module realizes optimized rolling control and online learning update through the following steps: Optimize the rolling control using the following formula: ; where, represents the two-channel prediction network with inputs of pressure P, displacement D, and control quantity u. represents the dynamic safety threshold. The term represents the prediction of the bridge pile state for the next 5 steps. The term represents the penalty control quantity and the deviation from the expected value . Update online using the following formula: ; where, represents the mean square error between the actual state and the predicted state . represents the neural network weight. Apply Activated normalization.

[0008] In one embodiment, the self-learning optimization module is further configured to: Obtain the working data of the jacking equipment in real time and fuse it to generate a working history dataset of the jacking equipment; Construct a learning model based on the state variables of the jacking equipment, and iteratively train it using the working history dataset to obtain a jacking equipment response prediction model; Based on the error backpropagation between the jacking equipment response prediction model and the preset control library, use a parameter mapping function to update the control algorithm weight coefficients in the preset control library.

[0009] In one embodiment, the system further includes a settlement trend prediction module for: Extract the time-frequency domain features of the pressure-displacement data matrix using the wavelet transform algorithm; Construct an LSTM-CNN branch deep learning model, analyze the time-frequency domain features, and generate time-frequency domain analysis information; Use the attention mechanism to fuse the time-frequency domain analysis information to generate a settlement trend prediction result; Dynamically set the pressure-displacement warning threshold using historical data, and combine it with the settlement trend prediction result to generate a hierarchical settlement warning.

[0010] In one embodiment, the system further includes a sensor fault diagnosis module for: Obtain the running data of the jacking equipment in real time, and use the Mahalanobis distance outlier detection algorithm to generate the deviation degree from the normal working condition; Judge whether the corresponding sensor fails according to the deviation degree from the normal working condition; When the corresponding sensor fails, activate the corresponding Kalman filter and generate a failure warning.

[0011] In one embodiment, the system further includes a remote monitoring module for: Based on the running data of the jacking equipment, dynamically estimate the actual displacement and pressure of the hydraulic cylinder in the jacking equipment through a state observer to obtain a dynamic estimation effect; When the sensor fault diagnosis module does not detect a failure, compare the error between the control instruction and the dynamic estimation result; When the sensor fault diagnosis module detects a failure, compare the error between the control instruction and the output data of the activated Kalman filter; When the error exceeds the preset threshold, trigger a downgrade control strategy and select a redundant control channel to execute.

[0012] In one embodiment, the remote monitoring module is further configured to: Obtain the construction video of the jacking equipment; The state of the jacking equipment is identified and detected in real time through a target detection model to generate a component state detection result; The SlowFast dual-path spatio-temporal network is used to extract the temporal correlation between the operation actions of construction workers and the vibration characteristics of the equipment to generate an operation compliance analysis result; The component state detection result and the operation compliance analysis result are fused, and a video analysis result is generated through a gated attention mechanism.

[0013] In one embodiment, the system is also used for: Taking the operating parameters and environmental data of the jacking equipment as inputs, and taking the hydraulic valve opening adjustment instruction and the motor speed control instruction of the jacking equipment as the action space, a multi-dimensional decision-making model is constructed; Based on the multi-dimensional decision-making model, combined with construction progress constraints and safety penalties, the DQN algorithm is used to generate an optimal energy supply strategy.

[0014] In a second aspect, the present application also provides a method for intelligent jacking of bridge construction, including: The pressure value data and displacement value data between the hydraulic cylinder and the bridge deck are obtained in real time, and a pressure-displacement data matrix is generated; Based on the pressure-displacement data matrix, an intelligent control algorithm library is used to calculate the force and settlement state of the bridge pile to generate force and settlement state information of the bridge pile; Based on the force and settlement state information, a corresponding control instruction is generated by using a preset control library; A learning model is constructed according to the working data of the jacking equipment, and the preset control library is dynamically updated through the learning model and the parameter mapping mechanism.

[0015] In a third aspect, the present application also provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for intelligent jacking of bridge construction are implemented.

[0016] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned method for intelligent jacking of bridge construction are implemented.

[0017] The above intelligent jacking system for bridge construction realizes intelligent control through the coordination of the following four core modules: The data acquisition module collects the pressure value and displacement value data between the hydraulic cylinder and the bridge deck in real time, generates a pressure-displacement data matrix, and provides basic monitoring data for the system; The control calculation module, based on the data matrix, dynamically calculates the real-time stress state and settlement information of the bridge pile through the built-in intelligent control algorithm library; The execution drive module calls the preset control instruction library according to the calculation result, generates specific control signals to drive the hydraulic system to execute actions; The self-learning optimization module dynamically optimizes the parameters of the preset control library by constructing a learning model of the equipment working data and combining the parameter mapping mechanism, improving the system adaptability. Through the full-process closed-loop control of real-time data acquisition, intelligent algorithm drive, and dynamic parameter optimization, it solves the problems of low accuracy of manual operation and inability to respond to working condition changes in real time in traditional construction, realizes the precise control of the stress and settlement state of the bridge pile, and ensures the construction safety and efficiency. Brief Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0019] Figure 1 It is a structural diagram of an intelligent jacking system for bridge construction of the present invention; Figure 2 It is a structural diagram of an embodiment of an intelligent jacking system for bridge construction of the present invention; Figure 3 It is a flowchart of an intelligent jacking method for bridge construction of the present invention. Detailed Embodiments

[0020] In order to make the purpose, technical solutions and advantages of the present application clearer, the following further details the present application in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0021] An intelligent jacking system for bridge construction in the present application includes a data acquisition module 101, a control and calculation module 102, an execution and drive module 103, and a self-learning and optimization module 104. Each module is connected through a communication network. The application scenarios cover the real-time monitoring, intelligent control, and dynamic optimization of the force and settlement states of bridge piles during the bridge construction process. When precise control of the force and settlement states of bridge piles is required, the data acquisition module 101 obtains the pressure value and displacement value data between the hydraulic cylinder and the bridge deck in real time through sensors, and generates a pressure-displacement data matrix. The data is transmitted to the control and calculation module 102 through the communication network. The control and calculation module 102 calculates the real-time force and settlement state information of the bridge pile using the intelligent control algorithm library. The execution and drive module 103 generates control instructions according to the calculation results and drives the hydraulic system to execute actions. The self-learning and optimization module 104 dynamically optimizes the control parameters by constructing a learning model of the equipment working data. The entire system realizes precise regulation of the force and settlement states of bridge piles through a full-process closed-loop control of real-time data acquisition, intelligent algorithm drive, and dynamic parameter optimization, ensuring construction safety and efficiency.

[0022] In one embodiment, as Figure 1 shown, an intelligent jacking system for bridge construction is provided. In this embodiment, taking the deployment of the system on a terminal as an example, it can be understood that the system can also be deployed on a server, and can also adopt an architecture including a terminal and a server, and be realized through the interaction between the terminal and the server. In this embodiment, the system includes the following modules: The data acquisition module 101 is used for: S1. Obtain the pressure value data and displacement value data between the hydraulic cylinder and the bridge deck in real time, and generate a pressure-displacement data matrix.

[0023] Among them, the pressure between the hydraulic cylinder and the bridge deck can be obtained in real time through a pressure sensor such as a strain gauge pressure sensor or a capacitive pressure sensor, and the displacement change of the hydraulic cylinder can be obtained in real time using a displacement sensor such as an LVDT linear variable differential transformer or an optical encoder. The sensor converts the pressure signal and displacement signal into electrical signals, and the analog signal is converted into a digital signal through an analog-to-digital converter. By constructing a pressure-displacement data matrix, the digital signal can be converted into structured data convenient for subsequent analysis and processing, providing key support for subsequent intelligent control and dynamic optimization, and ensuring that the system can monitor and regulate the force and settlement states of bridge piles in real time and accurately.

[0024] The control and calculation module 102 is used for: S2. Based on the pressure-displacement data matrix, calculate the force and settlement state of the bridge pile using the intelligent control algorithm library, and generate the force and settlement state information of the bridge pile.

[0025] Among them, based on the pressure-displacement data matrix, the real-time force and settlement state characteristics of the bridge pile can be extracted. The intelligent control algorithm library can include various algorithms such as fuzzy PID control algorithm, neural network predictive control algorithm, and model predictive control algorithm, which are used to dynamically calculate the force and settlement state of the bridge pile. According to different actual working conditions, the most suitable algorithm can be dynamically selected. Using the intelligent control algorithm library, the calculation results of the pressure-displacement data matrix are sorted into the force and settlement state information of the bridge pile, including the force state information of the bridge pile such as the current force magnitude and distribution of the bridge pile; the settlement state information of the bridge pile such as the settlement speed and settlement amount of the bridge pile. The force and settlement state information of the bridge pile provides a decision-making basis to ensure that the system can regulate the force and settlement state of the bridge pile in real time and accurately, and guarantee the safety and efficiency of the construction.

[0026] The execution drive module 103 is used for: S3, based on the force and settlement state information, using the preset control library, to generate corresponding control instructions.

[0027] Among them, the preset control library contains a variety of control strategies and parameters for generating specific control instructions. It can be to calculate appropriate control strategies through sliding mode control algorithm, adaptive control algorithm, robust control algorithm, etc., and convert the calculated control strategies into specific control instructions to drive the jacking equipment to perform corresponding actions. The calculation and application of these control strategies depend on the specific working conditions and system requirements to ensure that the force and settlement state of the bridge pile are always within the safe range.

[0028] The self-learning optimization module 104 is used for: S4, constructing a learning model according to the working data of the jacking equipment, and dynamically updating the preset control library through the learning model and parameter mapping mechanism.

[0029] Among them, working data such as the pressure, displacement, and settlement state of the hydraulic cylinder are obtained from the operation process of the jacking equipment to form a working history data set. The working history data set is used to iteratively train the learning model so that it can accurately predict the response of the equipment. It can be to compare the prediction results of the learning model with the control strategies in the preset control library and calculate the error. Through the parameter mapping mechanism of error backpropagation, the weight coefficients of the control algorithms in the preset control library are dynamically adjusted using the parameter mapping function, and the optimized parameters are fed back into the preset control library to achieve the dynamic update of the control strategy, enabling the intelligent jacking system for bridge construction to adapt to complex construction environments, improve the accuracy and safety of construction, reduce manual intervention at the same time, and improve construction efficiency.

[0030] In one of the embodiments, the control calculation module 102 is further used for: S5, based on the intelligent control algorithm library, combined with the real-time working condition information, use the following formula to switch the intelligent control algorithm for calculating the force and settlement state of the bridge pile: ; wherein, represents the control algorithm selected at time t, i represents the index of the control algorithm, i = 1 corresponds to the fuzzy PID control algorithm, i = 2 corresponds to the neural network predictive control algorithm, and i = 3 corresponds to the model predictive control MPC algorithm. represents the weight of the i-th control algorithm for the j-th operating condition parameter, and n represents the total number of operating condition parameters. represents the normalized value of the j-th operating condition parameter at time t: ; wherein, represents the real-time value of the j-th operating condition parameter, represents the minimum value of the j-th operating condition parameter, represents the maximum value of the j-th operating condition parameter.

[0031] Specifically, dynamically selecting the most suitable intelligent control algorithm according to the real-time operating condition information can improve the accuracy of calculating the force and settlement state of the bridge pile and the adaptability of the system. By combining the real-time operating condition parameters, the system can select the optimal solution among the three algorithms of fuzzy PID control, neural network predictive control, and model predictive control MPC to ensure precise regulation during the construction process. For example, at a certain moment t, the real-time operating condition parameters corresponding to different j values are obtained, including temperature, humidity, and load pressure. According to the real-time values of these parameters and their preset minimum values , and through normalization processing, the normalized value of the j-th operating condition parameter at time t is obtained. By combining the weights of each control algorithm for each operating condition parameter, the comprehensive score of each algorithm is calculated, and the algorithm with the highest score is selected as the current control algorithm. Flexibly adjusting the control strategy according to the real-time operating conditions can significantly improve the accuracy of calculating the force and settlement state of the bridge pile and the overall adaptability of the system.

[0032] In one embodiment, the control calculation module 102 is further configured to: S6, use the following model predictive control MPC algorithm to calculate the force and settlement state of the bridge pile: ; subject to: ; wherein, represents the optimal control input, represents the predicted output value, represents the desired output value, Indicates the weight of the control increment, and Indicates the constraint of the output, and Indicates the constraint of the input.

[0033] Exemplarily, MPC (Model Predictive Control) is a model-based optimal control strategy that realizes precise regulation in a dynamic environment by predicting the future behavior of the system and optimizing the control sequence. The term is used to measure the deviation between the predicted output and the desired output within the next H steps. Is used to penalize the drastic change of the control input to balance the control accuracy and smoothness. The weight is used to adjust the priority between the tracking error and the control input energy. The constraint conditions and are used to ensure that the forces on the bridge piles (such as stress and displacement) and the settlement amount are within the safe range. The constraint conditions and are used to limit the physical limits of the control inputs such as the pressure of the hydraulic cylinder and the rotational speed of the motor in the jacking equipment.

[0034] In one embodiment, the neural network predictive control algorithm in the control calculation module 102 realizes the optimized rolling control and online learning update through the following steps: S7. Optimize the rolling control using the following formula: ; Where, represents a dual-channel prediction network with the input of pressure P, displacement D, and control quantity u, represents the dynamic safety threshold, the term represents the prediction of the bridge pile state in the next 5 steps, the term represents the penalty control quantity and the deviation from the expected value ; S8. Update online using the following formula: ; Where, represents the mean square error between the actual state and the predicted state , represents the regularization applied to the neural network weight by activation.

[0035] Specifically, the dual-channel prediction network of the optimized rolling control captures the long-term temporal dependency of the pressure-displacement sequence, such as the settlement trend, in the LSTM layer according to the input pressure P, displacement D and control amount u, and uses the attention mechanism to weight the fusion of the features of the key time steps to enhance the sensitivity to abnormal working conditions and output the predicted values ​​of the bridge pile state, such as stress and settlement, for the next 5 steps. Dynamic safety threshold It can be set in real time according to the bridge design specifications and actual conditions, such as the maximum allowable stress Material strength, maximum sedimentation rate .pass Constrain the bridge pile status not to exceed the safety boundary and pass Suppress the drastic fluctuation of control instructions. Online learning updates dynamically modify neural network parameters based on real-time data to improve the model's adaptability to complex working conditions. Through a learning rate of 0.001, control the parameter update step size to avoid oscillation and calculate the gradient descent update. The loss function is based on the mean square error , quantifying the deviation between the actual state and the predicted value, Item passed The activation function imposes nonlinear constraints on the weights, such as: hour, , can suppress excessive weights, when hour, , which can avoid invalid parameter interference. Accurate state prediction is achieved through a dual-channel prediction network, combined with dynamic safety constraints and control smoothness optimization to ensure that the stress and settlement of bridge piles are always within a controllable range. The online learning mechanism further enhances the model's adaptability to complex working conditions, forming an intelligent control link of perception-decision-learning, significantly improving the safety and efficiency of bridge construction.

[0036] In one embodiment, the self-learning optimization module 104 is further configured to: S9, real-time acquisition of the working data of the lifting equipment, and fusion generation of a historical data set of the lifting equipment work; S10, constructing a learning model based on the state variables of the lifting equipment, and iteratively training the working history data set to obtain a lifting equipment response prediction model; S11, based on the error back propagation between the jacking equipment response prediction model and the preset control library, a parameter mapping function is used to update the control algorithm weight coefficient in the preset control library.

[0037] Exemplarily, the working data of the jacking equipment may include: hydraulic system parameters, cylinder pressure, flow and oil temperature; mechanical state parameters, bridge pile displacement, settlement rate and hydraulic cylinder stroke; environmental parameters, temperature, humidity and soil resistance; control instructions, hydraulic valve opening, motor speed and PWM duty cycle, etc. The data fusion technology of sliding time window and feature engineering can be used to fuse heterogeneous data: align multi-source sensor data by timestamp, calculate derived features such as pressure change rate (dP / dt) and displacement acceleration (d²D / dt²), remove outliers and fill missing values, and convert the fused data into a structured data set according to time series, which can be in the following format: [timestamp, pressure (P), displacement (D), oil temperature (T), valve opening (u), settlement rate (dD / dt),...]. A response prediction model based on state variables can be constructed by deep reinforcement learning or long short-term memory network, and the structured data set converted from the working data of the jacking equipment can be used for training to obtain the response prediction model of the jacking equipment. The output of the response prediction model is compared with the expected effect of the preset control library, the policy gradient is calculated, and the trial and error learning is realized to optimize the parameters. The nonlinear mapping function converts the model error into the control parameter adjustment amount, updates the control algorithm weight coefficient in the preset control library, realizes the dynamic optimization of the control parameters, and solves the pain points of experience dependence and static strategy in traditional construction.

[0038] In one embodiment, the system further includes a settlement trend prediction module for: S12, extracting the time-frequency domain features of the pressure-displacement data matrix using wavelet transform algorithm; S13, constructing a LSTM-CNN branch deep learning model, analyzing time-frequency domain features, and generating time-frequency domain analysis information; S14, using the attention mechanism to fuse the time-frequency domain analysis information and generate the settlement trend prediction results; S15, dynamically set the pressure-displacement warning threshold using historical data, and generate a graded settlement warning in combination with the settlement trend prediction results.

[0039] Specifically, using the wavelet transform algorithm, the pressure-displacement data is converted from the time domain to the time-frequency domain to capture transient features. Through multi-scale analysis, the pressure-displacement data is decomposed into different frequency components while retaining time localization information to capture the non-stationary change characteristics of pressure and displacement during bridge construction, such as sudden loads and geological disturbances. The LSTM layer in the LSTM-CNN branch deep learning model is used to process the time dimension and capture the temporal correlations in the settlement process, such as continuous settlement acceleration; the CNN layer is used to extract spatial features and identify local anomalies in the pressure distribution, such as local stress concentration; the branch structure is used for two-stream parallel processing to fuse temporal and spatial information, enhance the model's representation ability, and generate time-frequency domain analysis information. The attention mechanism is used for fusion to dynamically weight key features, enhance the model's sensitivity to important signals, generate a weighted fusion feature vector, and obtain the settlement trend prediction result. It can be based on the statistical mean and standard deviation of the historical settlement curve, dynamically adjust the graded settlement warning threshold, combine the current construction stage, such as the pile foundation pouring period and the loading period, to adjust the sensitivity. Based on the adjusted threshold and combined with the settlement trend prediction result, a graded settlement warning is generated. The grading of the warning can be: green safety level, settlement rate ≤ 2 mm / h, cumulative settlement ≤ design allowable value; yellow warning level, settlement rate 2 - 5 mm / h, need to strengthen monitoring; red danger level, settlement rate > 5 mm / h or cumulative settlement exceeds the limit, trigger emergency measures.

[0040] In one embodiment, the system further includes a sensor fault diagnosis module for: S16, obtaining the operation data of the jacking equipment in real time and generating the deviation degree from the normal working condition using the Mahalanobis distance outlier detection algorithm; S17, judging whether the corresponding sensor fails according to the deviation degree from the normal working condition; S18, when the corresponding sensor fails, activating the corresponding Kalman filter and generating a failure warning.

[0041] Exemplarily, using the Mahalanobis distance outlier detection algorithm, the abnormal deviation of the sensor output is identified through statistical learning, and its difference from the normal working condition is quantified. It can be to perform mean normalization and covariance matrix calculation on multi-dimensional sensor data, measure the distance between the current data point and the mean, map the Mahalanobis distance to an outlier score, and generate the deviation degree from the normal working condition. It can be to set an initial threshold based on the 95% confidence interval of the historical normal working condition data to judge whether the corresponding sensor fails. When the corresponding sensor fails, use the output of the corresponding Kalman filter to generate a corrected control instruction, avoid control deviation caused by sensor failure, ensure the control stability during the failure period, and realize the closed-loop management from anomaly detection to fault tolerance control, improving the reliability and safety of the bridge construction system.

[0042] In one embodiment, the system further includes a remote monitoring module for: S19. Based on the operation data of the jacking equipment, the actual displacement and pressure of the hydraulic cylinder in the jacking equipment are dynamically estimated through a state observer to obtain the dynamic estimation effect; S20. When the sensor fault diagnosis module does not detect a failure, compare the error between the control instruction and the dynamic estimation result; S21. When the sensor fault diagnosis module detects a failure, compare the error between the control instruction and the output data of the activated Kalman filter; S22. When the error exceeds the preset threshold, trigger the degradation control strategy and select the redundant control channel to execute.

[0043] Specifically, it can be by using a state observer such as a Luenberger observer or an extended Kalman filter combined with a system dynamics model such as a hydraulic cylinder flow-pressure model, and combining real-time operation data to estimate the actual displacement and pressure of the hydraulic cylinder. Compare the control instruction with the estimation result or the output data error of the Kalman filter. When the obtained error exceeds the preset threshold, such as when the Mahalanobis distance is abnormal or the Kalman filter is activated, trigger the degradation control strategy, such as: instruction correction, adjusting control parameters through fuzzy logic or a rule base, such as reducing the cylinder pressure; channel switching, enabling a backup sensor or control loop, such as switching to a redundant hydraulic valve; safety braking, urgently stopping the jacking and locking the current state. The redundant control channel can be designed with hardware redundancy, such as Arranging dual-channel pressure sensors, displacement sensors, and main / backup channel independent hydraulic control valve groups in multiple places; software redundancy, such as running the PID and MPC algorithms simultaneously and adopting a voting mechanism or weighted average decision fusion.

[0044] In one embodiment, the remote monitoring module is further configured to: S23. Obtain the construction video of the jacking equipment; S24. Real-time identify and detect the state of the jacking equipment through a target detection model to generate a component state detection result; S25. Use the SlowFast dual-path spatio-temporal network to extract the temporal correlation between the operation actions of construction workers and the vibration characteristics of the equipment to generate an operation compliance analysis result; S26. Integrate the component state detection result and the operation compliance analysis result, and generate a video analysis result through a gated attention mechanism.

[0045] Exemplarily, it can be to capture high-definition video streams in real time through cameras deployed at the construction site, covering the jacking equipment such as hydraulic cylinders, support frames, and the operation areas of construction workers. A lightweight object detection algorithm is used to detect key components of the equipment such as cylinders, sensors, bolts, pressure gauges, etc. in real time, and the equipment status is identified to generate component status detection results. Using the SlowFast dual-path spatio-temporal network, the Slow path processes videos with a low frame rate such as 1 frame per second to capture global operation actions such as personnel standing positions and tool usage, and the Fast path processes videos with a high frame rate such as 30 frames per second to extract local details such as gesture actions and equipment vibration frequencies, and analyze and identify: illegal operations such as not wearing safety helmets and overloading hoisting; equipment resonance or abnormal vibration; synchronize the operation actions with the vibration signals, analyze the causal relationship such as excessive vibration caused by high-intensity operations of workers, and attention weights can be assigned according to the severity of component status and the operation risk level to generate video analysis results.

[0046] In one of the embodiments, the system is also used for: S27, taking the operating parameters and environmental data of the jacking equipment as inputs, and taking the hydraulic valve opening adjustment instruction and motor speed control instruction of the jacking equipment as the action space, to construct a multi-dimensional decision-making model; S28, based on the multi-dimensional decision-making model, combined with construction progress constraints and safety penalties, using the DQN algorithm to generate an optimal energy supply strategy.

[0047] Specifically, a multi-dimensional decision-making model is constructed with the operating parameters of the jacking equipment such as the pressure of the hydraulic cylinder, oil temperature, displacement sensor data, oil pump flow rate, motor current, etc. and environmental data such as soil resistance, temperature, humidity, wind speed, and construction site video stream as inputs, and the corresponding hydraulic valve opening adjustment instruction and motor speed control instruction as outputs. The DQN (DeepQ-Network) algorithm, a reinforcement learning algorithm that combines deep learning and Q learning, replaces the traditional Q table through a neural network, solves the decision-making problem in a high-dimensional state space, and introduces an experience replay and target network mechanism to improve training stability. Using the DQN algorithm, a construction progress factor and a safety penalty factor are introduced to construct a composite reward function, mapping the multi-dimensional state of the jacking equipment and the control instructions to an optimal energy supply strategy, and achieving the construction goals of high efficiency and energy conservation under the premise of ensuring safety.

[0048] An intelligent lifting system for bridge construction in the present application significantly improves the safety, accuracy, and efficiency of bridge construction through multi-module collaborative innovation. The system obtains a pressure-displacement data matrix in real time through a data acquisition module, and combines fuzzy PID, neural network prediction, and MPC algorithms in an intelligent control algorithm library to dynamically switch control strategies, realizing accurate calculation of the force and settlement state of bridge piles. The self-learning optimization module constructs a prediction model based on historical data and reversely optimizes control parameters to form a closed-loop adaptive ability. The settlement trend prediction module uses wavelet transform and LSTM-CNN models to analyze time-frequency domain features and combines dynamic thresholds to achieve hierarchical early warning. The sensor fault diagnosis module detects anomalies through Mahalanobis distance and activates Kalman filter fault tolerance to ensure system reliability. The remote monitoring module integrates video analysis and state observation to achieve equipment state recognition, operation compliance analysis, and redundant control. The multi-dimensional decision-making model generates an optimal energy supply strategy in combination with the DQN algorithm to balance construction progress and safety. The overall system solves problems such as excessive manual intervention, poor adaptability, and high safety hazards in traditional construction through the full-process intelligence of real-time perception, intelligent decision-making, dynamic optimization, and risk early warning, providing an efficient and reliable intelligent solution for bridge engineering.

[0049] To further illustrate the solution of the embodiments of the present application, a specific example is given below for explanation.

[0050] In this example, as Figure 2 shown, the intelligent lifting equipment for bridge construction consists of a host computer, an oil cylinder, a weighing sensor, a displacement sensor, and an integrated circuit, and is connected through the integrated circuit. The data collected by the weighing sensor and the displacement sensor are transmitted to the data acquisition module 101 through the integrated circuit. The host computer in the control calculation module 102 communicates with the hardware device through the integrated circuit and performs relevant calculations to generate control instructions. The execution drive module 103 receives the control instructions and drives the oil cylinder to complete the execution action: The data acquisition module 101 is used to obtain the pressure value data and displacement value data between the hydraulic cylinder and the bridge deck in real time; Among them, the pressure signal data between the hydraulic cylinder and the bridge deck can be obtained through a weighing sensor, and the distance signal data between the hydraulic cylinder and the bridge deck can be obtained in real time through a displacement sensor such as a self-resetting displacement sensor and transmitted to the control calculation module.

[0051] The motor control module 102 is used to judge the force and settlement state of the bridge pile based on the pressure value data and displacement value data, and generate the force and settlement state information of the bridge pile; Among them, it can be determined whether the bridge pile status meets the requirements by using an internal program based on the pressure signal data and the distance signal data. If the pressure value feedback by the sensor is less than the preset pressure value, or the distance value is less than the preset displacement value, it is determined that the pressure of the oil cylinder is insufficient, resulting in the settlement of the bridge pile where it is located, and the bridge pile status does not meet the requirements. If the feedback pressure value reaches the upper limit of the preset pressure value, it indicates that the oil cylinder is fully stressed on the bridge, the contact surface is not suspended, and it is in a normal state. If the distance value is greater than the preset displacement value, it indicates that the pressure given by the oil cylinder is too large. Based on this, the force and settlement status information of the bridge pile is generated.

[0052] Execute the driving module 103, which is used to generate corresponding control instructions based on the force and settlement status information by using a preset control library. Among them, the control mechanism in the preset control library can be that, according to the force and settlement status information of the bridge pile, when the pressure value feedback by the sensor is less than the preset pressure value and the distance value is less than the preset displacement value at the same time, an instruction to control the hydraulic cylinder to rise is generated; when the pressure value feedback by the sensor reaches the upper limit value of the preset pressure value or it is detected that the displacement value is about to reach the preset displacement value, an instruction to control the hydraulic cylinder to stop rising is generated. Similarly, when it is detected that the displacement value is greater than the preset displacement value, an instruction to control the hydraulic cylinder to descend is generated, and when the pressure value feedback by the sensor reaches the preset pressure value range, an instruction to control the hydraulic cylinder to stop descending is generated. By this method, the control accuracy can be improved.

[0053] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are displayed in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0054] Based on the same inventive concept, the embodiments of the present application also provide a method for implementing the above-mentioned intelligent jacking system for bridge construction. The solution provided by this method to solve the problem is similar to the solution described in the above system. Therefore, the specific limitations in one or more embodiments of the following intelligent jacking method for bridge construction can refer to the limitations on the intelligent jacking system for bridge construction in the above text, and will not be repeated here.

[0055] In an exemplary embodiment, as Figure 3As shown, an intelligent jacking method for bridge construction is provided, including: S1. Obtain the pressure value data and displacement value data between the hydraulic cylinder and the bridge deck in real time, and generate a pressure-displacement data matrix; S2. Based on the pressure-displacement data matrix, use the intelligent control algorithm library to calculate the force and settlement state of the bridge pile, and generate the force and settlement state information of the bridge pile; S3. Based on the force and settlement state information, use the preset control library to generate corresponding control instructions; S4. Construct a learning model according to the working data of the jacking equipment, and dynamically update the preset control library through the learning model and the parameter mapping mechanism.

[0056] In one embodiment, the method further includes: S5. Based on the intelligent control algorithm library, combined with the real-time working condition information, use the following formula to switch the intelligent control algorithm for calculating the force and settlement state of the bridge pile: ; where, represents the control algorithm selected at time t, i represents the index of the control algorithm, i = 1 corresponds to the fuzzy PID control algorithm, i = 2 corresponds to the neural network predictive control algorithm, i = 3 corresponds to the model predictive control MPC algorithm, represents the weight of the i-th control algorithm for the j-th working condition parameter, n represents the total number of working condition parameters, represents the normalized value of the j-th working condition parameter at time t: ; where, represents the real-time value of the j-th working condition parameter, represents the minimum value of the j-th working condition parameter, represents the maximum value of the j-th working condition parameter.

[0057] In one embodiment, the method further includes: S6. Use the following model predictive control MPC algorithm to calculate the force and settlement state of the bridge pile: ; where, represents the optimal control input, represents the predicted output value, represents the desired output value, represents the weight of the control increment, and represent the constraints of the output, and represent the constraints of the input.

[0058] In one embodiment, the neural network predictive control algorithm in the control computing module 102 realizes optimized rolling control and online learning update through the following steps: S7. Optimize the rolling control using the following formula: ; where represents a two-channel prediction network with inputs of pressure P, displacement D, and control quantity u, represents the dynamic safety threshold, the term represents the prediction of the bridge pile state for the next 5 steps, the term represents the penalty control quantity and the deviation from the expected value ; S8. Update online learning using the following formula: ; where represents the mean square error between the actual state and the predicted state , represents the regularization applied to the neural network weights activated by .

[0059] In one embodiment, the method further includes: S9. Real-time obtain the working data of the jacking equipment and fuse it to generate a working history dataset of the jacking equipment; S10. Construct a learning model based on the state variables of the jacking equipment and iteratively train it using the working history dataset to obtain a jacking equipment response prediction model; S11. Based on the error backpropagation between the jacking equipment response prediction model and the preset control library, update the control algorithm weight coefficients in the preset control library using a parameter mapping function.

[0060] In one embodiment, the method further includes: S12. Use the wavelet transform algorithm to extract the time-frequency domain features of the pressure-displacement data matrix; S13. Construct an LSTM-CNN branch deep learning model to analyze the time-frequency domain features and generate time-frequency domain analysis information; S14. Use the attention mechanism to fuse the time-frequency domain analysis information to generate a settlement trend prediction result; S15. Dynamically set the pressure-displacement warning threshold using historical data and combine it with the settlement trend prediction result to generate a hierarchical settlement warning.

[0061] In one embodiment, the method further includes: S16. Obtain the operation data of the jacking equipment in real time, and generate the deviation degree from the normal working condition by using the Mahalanobis distance outlier detection algorithm; S17. Judge whether the corresponding sensor fails according to the deviation degree from the normal working condition; S18. When the corresponding sensor fails, activate the corresponding Kalman filter and generate a failure warning.

[0062] In one embodiment, the method further includes: S19. Based on the operation data of the jacking equipment, dynamically estimate the actual displacement and pressure of the hydraulic cylinder in the jacking equipment through a state observer to obtain the dynamic estimation effect; S20. When the sensor fault diagnosis module does not detect a failure, compare the error between the control instruction and the dynamic estimation result; S21. When the sensor fault diagnosis module detects a failure, compare the error between the control instruction and the output data of the activated Kalman filter; S22. When the error exceeds the preset threshold, trigger the downgrade control strategy and select the redundant control channel to execute.

[0063] In one embodiment, the method further includes: S23. Obtain the construction video of the jacking equipment; S24. Real-time identify and detect the state of the jacking equipment through the target detection model to generate the component state detection result; S25. Use the SlowFast dual-path spatio-temporal network to extract the temporal correlation between the operation actions of the construction personnel and the vibration characteristics of the equipment to generate the operation compliance analysis result; S26. Integrate the component state detection result and the operation compliance analysis result, and generate the video analysis result through the gated attention mechanism.

[0064] In one embodiment, the method further includes: S27. Take the operation parameters and environmental data of the jacking equipment as the input, and take the hydraulic valve opening adjustment instruction and the motor speed control instruction of the jacking equipment as the action space to construct a multi-dimensional decision-making model; S28. Based on the multi-dimensional decision-making model, combine the construction progress constraint and the safety penalty, and use the DQN algorithm to generate the optimal energy supply strategy.

[0065] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of an intelligent jacking system for bridge construction as described above are implemented.

[0066] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0067] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the descriptions of the method embodiments. The device embodiments described above are only illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. A person of ordinary skill in the art can understand and implement it without creative work.

[0068] The above embodiments only represent several implementation manners of the embodiments of the present application. The descriptions are relatively specific and detailed, but should not be construed as a limitation on the patent scope of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. An intelligent lifting system for bridge construction, characterized in that, Including: A data acquisition module, which is used to obtain the pressure value data and displacement value data between the hydraulic cylinder and the bridge deck in real time, and generate a pressure-displacement data matrix; A control calculation module, which is used to calculate the force and settlement state of the bridge pile based on the pressure-displacement data matrix by using an intelligent control algorithm library, and generate the force and settlement state information of the bridge pile; An execution drive module, which is used to generate corresponding control instructions based on the force and settlement state information by using a preset control library; A self-learning optimization module, which is used to construct a learning model according to the working data of the jacking equipment, and dynamically update the preset control library through the learning model and the parameter mapping mechanism.

2. The intelligent lifting system for bridge construction according to claim 1, wherein The control calculation module is further used for: Based on the intelligent control algorithm library, combined with the real-time working condition information, use the following formula to switch the intelligent control algorithm for calculating the force and settlement state of the bridge pile: ； Among them, represents the control algorithm selected at time t, i represents the index of the control algorithm, i = 1 corresponds to the fuzzy PID control algorithm, i = 2 corresponds to the neural network predictive control algorithm, and i = 3 corresponds to the model predictive control MPC algorithm, represents the weight of the i-th control algorithm for the j-th operating condition parameter, and n represents the total number of operating condition parameters, represents the normalized value of the j-th operating condition parameter at time t: ； Among them, represents the real-time value of the j-th operating condition parameter, represents the minimum value of the j-th operating condition parameter, represents the maximum value of the j-th operating condition parameter.

3. An intelligent lifting system for bridge construction according to claim 2, characterized in that, The control calculation module is further used for: Use the following model predictive control MPC algorithm to calculate the force and settlement state of the bridge pile: ； wherein, represents the optimal control input, represents the predicted output value, represents the desired output value, represents the weight of the control increment, and represents the output constraint, and represents the input constraint.

4. An intelligent jacking system for bridge construction according to claim 2, characterized in that, The neural network predictive control algorithm in the control calculation module realizes optimized rolling control and online learning update through the following steps: Use the following formula to optimize rolling control: ； Among them, represents a two-channel prediction network with input pressure P, displacement D, and control quantity u, represents the dynamic safety threshold, The item represents the prediction of the bridge pile state in the next 5 steps, The item represents the penalty control quantity and the deviation from the expected value ; Use the following formula for online learning update: ； Among them, represents the actual state and the predicted state of the mean square error, represents the neural network weights applied activation regularization.

5. An intelligent jacking system for bridge construction according to claim 1, characterized in that, The self-learning optimization module is further used for: Obtain the working data of the jacking equipment in real time, and fuse to generate a working history data set of the jacking equipment; Construct a learning model based on the state variables of the jacking equipment, and iteratively train the working history data set to obtain a jacking equipment response prediction model; Based on the error backpropagation between the jacking equipment response prediction model and the preset control library, use a parameter mapping function to update the control algorithm weight coefficient in the preset control library.

6. The intelligent jacking system for bridge construction according to claim 1, wherein It further includes a settlement trend prediction module, which is used for: Use the wavelet transform algorithm to extract the time-frequency domain characteristics of the pressure-displacement data matrix; Construct an LSTM-CNN branch deep learning model, analyze the time-frequency domain characteristics, and generate time-frequency domain analysis information; Use the attention mechanism to fuse the time-frequency domain analysis information to generate a settlement trend prediction result; Dynamically set the pressure-displacement warning threshold by using historical data, and combine the settlement trend prediction result to generate a hierarchical settlement warning.

7. An intelligent lifting system for bridge construction according to claim 4, characterized in that, It further includes a sensor fault diagnosis module, which is used for: Obtain the operation data of the jacking equipment in real time, and use the Mahalanobis distance outlier detection algorithm to generate the deviation degree of the normal working condition; Judge whether the corresponding sensor fails according to the deviation degree of the normal working condition; When the corresponding sensor fails, activate the corresponding Kalman filter and generate a failure warning.

8. An intelligent jacking system for bridge construction according to claim 7, characterized in that, It further includes a remote monitoring module, which is used for: Based on the operation data of the jacking equipment, dynamically estimate the actual displacement and pressure of the hydraulic cylinder in the jacking equipment through a state observer to obtain a dynamic estimation effect; When the sensor fault diagnosis module does not detect a failure, compare the error between the control instruction and the dynamic estimation result; When the sensor fault diagnosis module detects a failure, compare the error between the control instruction and the output data of the activated Kalman filter; When the error exceeds the preset threshold, trigger a downgraded control strategy and select a redundant control channel to execute.

9. The intelligent jacking system for bridge construction according to claim 8, characterized in that, The remote monitoring module is further used for: Obtain the construction video of the jacking equipment; The state of the jacking equipment is recognized and detected in real time through the target detection model, and the component state detection result is generated; The temporal correlation between the operation actions of the construction workers and the vibration characteristics of the equipment is extracted by using the SlowFast dual-path spatio-temporal network, and the operation compliance analysis result is generated; The component state detection result and the operation compliance analysis result are fused, and the video analysis result is generated through the gated attention mechanism.

10. The intelligent jacking system for bridge construction according to claim 1, characterized in that, The system is also used for: Taking the operating parameters and environmental data of the jacking equipment as inputs, and taking the hydraulic valve opening adjustment instruction and the motor speed control instruction of the jacking equipment as the action space, a multi-dimensional decision-making model is constructed; Based on the multi-dimensional decision-making model, combined with the construction progress constraint and safety penalty, the optimal energy supply strategy is generated by using the DQN algorithm.

Images