Railway bridge erecting machine operation process digitization collaborative control method based on multi-working condition perception

Abstract

The application discloses a railway bridge erecting machine operation process digitalization collaborative control method based on multi-working condition sensing, relates to the field of bridge erecting machine control, and comprises the following steps: when a beam is moved and positioned, the measurement values of each working condition sensing node are synchronously collected, original working condition data frames carrying working procedure identifiers are generated, and the working procedure identifiers represent the current beam moving and positioning procedure; configuration normalization is performed on the original working condition data frames, and a standardized pose state vector independent of the sensing node installation configuration is obtained; the beam falling parameters are solved through multi-working condition sensing normalization, the driving signal is corrected in the frequency domain and is fed back in real time, low-speed crawling and disturbance are inhibited, abnormal sensing values are removed, stable positioning is determined through pose deviation and smoothness discrimination, the procedure parameters are smoothly connected, and the adaptive and anti-disturbance capabilities of operation procedures such as beam moving and beam falling are improved.

Description

Technical Field

[0001] This invention relates to the field of bridge erecting machine control technology, specifically to a digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception. Background Technology

[0002] Railway bridge erecting machines are used for the erection of precast beams for high-speed railways. The operation process includes beam crossing, beam feeding, beam movement, beam lowering, and outrigger conversion, involving the sequential coordination of multiple execution units such as front and rear outriggers, overhead cranes, and hoisting mechanisms. In complex mountainous environments, non-standard working conditions such as tunnel entrance clearances, small-radius curves, and steep longitudinal slopes place higher demands on equipment attitude and load control. Currently, intelligent sensing and digital twin technologies are used for key parameter monitoring and construction simulation, improving the level of visual monitoring of single-machine operations.

[0003] Chinese invention patent application number 202511659011.5 discloses an intelligent control system for automatic adjustment of a smart bridge-building machine. This application obtains the operating state vectors at multiple working conditions, performs pattern clustering and extracts typical working conditions, and then constructs an intelligent control model to achieve automatic adjustment.

[0004] However, after the bridge erecting machine moves the beam into place, before the beam dropping process begins, most existing bridge erecting machines still require manual measurement of the deviation between the actual posture of the beam and the designed posture, and manual adjustment of the beam dropping parameters accordingly. The large amount of working data collected by the sensors is only used for display and alarm, and is not automatically converted into control commands for the next process. The data flow between processes is interrupted by manual reading and manual input, making it difficult to achieve true continuous digital collaboration. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a digital collaborative control method for the operation process of railway bridge erecting machine based on multi-condition perception, which can effectively solve the problems of the existing technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions;

[0007] This invention discloses a digital collaborative control method for the operation process of a railway bridge erecting machine based on multi-condition sensing, including:

[0008] During beam placement, the measured values ​​of each working condition sensing node are synchronously collected to generate an original working condition data frame carrying a process identifier, which represents the current beam placement process. The original working condition data frame is normalized to obtain a standardized pose state vector that is independent of the installation configuration of the sensing nodes. Based on the preset mapping relationship between the standardized pose state vector and the beam dropping action parameters, the standardized pose state vector is directly converted into a beam dropping action parameter set, which includes the target stroke and action timing of each beam dropping actuator. Based on the dynamic response characteristics of the bridge erecting machine actuator, the beam dropping action parameter set is compiled into a drive signal sequence, which represents the time waveform of the control signal of each actuator. During the output of the drive signal sequence, the state feedback of each actuator is continuously collected, and the state feedback is compared point by point with the predetermined response state at the corresponding time in the drive signal sequence to generate a compensation correction amount.

[0009] The compensation and correction amount is superimposed on the drive signal sequence in real time to form a drive command for closed-loop regulation. The output continues until the status feedback of each driver reaches the condition for beam placement, and then the output stops.

[0010] Furthermore, the working condition sensing nodes include at least displacement sensing nodes installed at the front and rear supports of the main beam of the bridge erecting machine, tilt sensing nodes installed on both sides of the hoisting drum of the overhead crane, and pressure sensing nodes installed in the oil circuit of the beam lowering hydraulic cylinder.

[0011] The configuration normalization operation is as follows:

[0012] A global base coordinate system is constructed with the rotation center of the bridge erecting machine as the origin. Based on the factory installation calibration parameters of each sensing node, the rotation matrix of the local measurement coordinate system of each node relative to the global base coordinate system is obtained. With translation vector The local physical quantity measurements output by each node By transforming the relationship Unified representation quantity projected onto the global base coordinate system Then, the standardized pose state vector is obtained by weighting each unified representation quantity according to the node type. ,in Characterizes the spatial position components of the beam's centroid in the global base coordinate system. Characterizes the attitude angular components of the beam about the three coordinate axes of the global base coordinate system.

[0013] Furthermore, during the synchronous acquisition of measurement values ​​from each sensing node under various operating conditions, sensor data self-tests are performed in parallel. For the i-th sensing node, a Kalman filter is used to obtain the filtered estimate of the current moment based on its historical measurement sequence. And calculate the actual measured value. With the filtered estimate residuals between ;

[0014] Construct the real-time confidence weights for the sensing node based on the residuals:

[0015] ;

[0016] In the formula: The standard deviation of the measurement noise is statistically obtained for the i-th sensing node under offline calibration conditions;

[0017] when When the value is below the preset confidence threshold, the sensing node is marked as a failed node and its current measurement value is removed. At the same time, the measurement values ​​of the remaining valid sensing nodes are used for configuration normalization by weighted averaging.

[0018] When the number of effective sensing nodes is lower than the preset minimum number of nodes, the output drive signal sequence is paused and a safety alarm is triggered.

[0019] Furthermore, the preset mapping relationship is constructed as follows:

[0020] The beam dropping process feature database is constructed from historical process samples. Each historical sample records the corresponding historical standardized pose state vector, the support reaction force vector of each beam dropping actuator, the actual stroke vector, and the actual action timing vector.

[0021] When constructing the predefined mapping relationship, it is constructed as a function derived from the inverse kinematics. , Dynamic calibration subfunction With time-series planning subfunction The nonlinear mapping function F is formed by cascading; using historical samples as the training set, combined with the current beam self-weight and eccentric load distribution, the load distribution matrix of the actuator is fitted.

[0022] The Based on the deviation vector between the current attitude angle components and the target attitude angle components, the kinematic mapping relationship is solved to obtain the initial stroke value of each actuator;

[0023] The Based on the load distribution matrix and the current beam self-weight and eccentric load distribution, the support reaction force of each actuator is solved, and the elastic deformation compensation amount is determined according to the structural stiffness relationship. The sum of the initial stroke value and the elastic deformation compensation amount is used as the target stroke of each actuator.

[0024] The Using the startup time of each driver as the optimization variable, and based on the pose disturbance coupling relationship between each driver and the preset driver speed, acceleration and safety time interval constraints, the optimal startup time that minimizes the action conflict of each driver is solved, and the optimal startup time is used as the action timing output.

[0025] Furthermore, the dynamic response characteristics of the bridge erecting machine driver are characterized by the frequency response characteristic curve and step response rise time of the electro-hydraulic servo system of each driver.

[0026] When compiling the beam dropping motion parameter set into a drive signal sequence, for each beam dropping driver, first, based on its target stroke... and the start time in the action sequence Termination time Generate an ideal trapezoidal velocity profile, and then determine the phase hysteresis in the frequency response curve of the driver. and amplitude attenuation The ideal trapezoidal velocity profile is corrected in the frequency domain to obtain the corrected velocity command waveform. Finally, the speed command waveform According to the voltage-speed conversion factor of the driver control interface Converted to analog voltage signal waveform The analog voltage signal waveforms of each driver are aligned with the synchronous clock edge to form the drive signal sequence.

[0027] Furthermore, during the process of compiling the beam lowering action parameter set into a drive signal sequence, the real-time monitoring value of the oil temperature of the electro-hydraulic servo system of each driver is also used. Real-time monitoring value of system pressure For the voltage-speed conversion coefficient Real-time corrections are performed, and the correction relationship is as follows:

[0028] ;

[0029] In the formula: For the drive at the calibrated oil temperature and calibration system pressure The baseline conversion factor is as follows. and These are the oil temperature compensation coefficient and the pressure compensation coefficient, respectively.

[0030] Using the corrected real-time conversion coefficients The analog voltage signal waveform is recalculated, and the signal amplitude of the corresponding driver at the corresponding time in the drive signal sequence is replaced with the updated analog voltage signal waveform.

[0031] Furthermore, when generating compensation and correction amounts by comparing the state feedback with the predetermined response states at corresponding times in the driving signal sequence point by point, the current sampling time is set as... The collected state feedback vector is In the driving signal sequence The predetermined response state vector at time t is Define the synchronization error vector ;

[0032] Based on the aforementioned synchronization error vector, the following compensation correction amount is constructed. Generative law:

[0033] ;

[0034] In the formula: ⊙ represents the Hadamard product; These are the proportional gain vector, integral gain vector, and differential gain vector, respectively. The j-th sampling interval; N is the number of sampling points within the integration backtracking window; This is the first-order difference estimate of the synchronization error. is the nonlinear friction compensation gain vector; sign(⋅) is the sign function; β is a pre-defined nonlinear exponential coefficient between 0 and 1;

[0035] The width of the integral backtracking window The adjustment is adaptive based on the beam movement speed during the current process;

[0036] Among them, the sign function The value can be:

[0037] when When >0, take +1; when When < 0, take -1; when When =0, take 0;

[0038] The nonlinear exponential coefficient β ranges from 0.3 to 0.7. Considering the prominent low-speed crawling phenomenon of hydraulic cylinders during the bridge erecting machine's beam lowering operation, the specific value of β is determined based on the ratio of the lowest stable motion speed measured by the actuator under both no-load and full-load conditions. The larger the ratio, the closer β is to 0.3; the smaller the ratio, the closer β is to 0.7. Intermediate ratios are determined by linear interpolation.

[0039] The width of the integral backtracking window Based on the beam movement speed of the current process Adjust according to the following segmented adaptive rules:

[0040] when When the speed exceeds the preset high-speed threshold, the width is taken as the first preset time length T1. When the speed is less than the preset low-speed threshold, the width is taken as the second preset time length T2, where T2 is greater than T1. When the width is between a preset low-speed threshold and a preset high-speed threshold, the width is continuously taken between T1 and T2 using a linear interpolation method.

[0041] Furthermore, the beam placement conditions are determined by a multi-dimensional stable convergence discriminant;

[0042] Set the target position as and Current sampling time The actual position in the state feedback after closed-loop regulation is The actual posture is Define the position deviation norm With attitude deviation norm ;

[0043] Simultaneously define a stability convergence index:

[0044] ;

[0045] In the formula: and These are the position component and the attitude angle component, respectively. and These are preset coefficients corresponding to the weights for the rate of change of position and the acceleration of attitude change, respectively. The sampling period for state feedback;

[0046] when ≤ , ≤ and ≤ Simultaneously established and maintained for more than the preset stable time window. When the conditions for lowering the beam into place are met, the beam is deemed to have been lowered into place. , , These are the position deviation threshold, attitude deviation threshold, and stability convergence threshold, respectively.

[0047] Furthermore, after determining that the beam placement condition has been met and terminating the current output drive signal sequence, the current process identifier is switched from the beam placement process to the next preset process identifier. At the same time, the standardized pose state vector at the termination time, the final stroke position of each actuator, and the compensation correction amount at the termination time are packaged into a process transition state package.

[0048] When switching to the next preset process, the final travel position in the process transition state package is used as the initial reference benchmark for the beam dropping action parameter set in the next preset process, and the compensation correction amount at the termination time is used as the feedforward compensation initial value at the start time of the next preset process and superimposed on the first drive signal sequence of the next preset process.

[0049] Furthermore, in addition to carrying the process identifier, the original working condition data frame also carries a global timestamp, frame sequence number, and current job cycle number;

[0050] Before synchronously collecting the measured values ​​of each working condition sensing node, a distributed clock synchronization reference is first established. The distributed clock synchronization reference eliminates the local clock drift of each node through master-slave clock message exchange, so that the acquisition time of each sensing node and the output time of the drive signal sequence share a global time base.

[0051] During the output of the drive signal sequence, the control quantity at each moment in the output drive signal sequence of each frame is bound to its corresponding frame number and global timestamp, and the local timestamp attached to the state feedback at the sampling moment is converted into a synchronization timestamp under a unified time base through the global time base. The point-by-point comparison is only performed between data pairs with the same frame number and a synchronization timestamp deviation less than a preset synchronization deviation tolerance.

[0052] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:

[0053] This invention normalizes the measured values ​​of multiple sensing nodes under various working conditions into a standardized pose state vector through configuration, and directly solves the beam dropping motion parameters based on a preset mapping relationship. This avoids parameter tuning that relies on manual experience or repeated trial and error. It utilizes the dynamic response characteristics of the actuator to correct the frequency domain of the speed command, making the control signal fit the actual capability of the actuator and reducing signal distortion errors. When outputting the drive signal, it continuously compares the state feedback and generates a correction quantity with nonlinear friction compensation. With the help of an adaptive integral backtracking window, it suppresses low-speed crawling of the hydraulic cylinder and external disturbances. Real-time calculation of sensor confidence weights and elimination of failed nodes ensure the reliability of the measured values ​​involved in the control, avoiding instability caused by single sensor anomalies. A multi-dimensional stability convergence discriminator examines position deviation, attitude deviation, and motion stability. A continuous holding time window is set to ensure that the beam is positioned to meet the positioning accuracy and has static stability. The recording and transmission of process transition state packets ensures smooth connection of control parameters between adjacent processes, reducing the impact of process switching and the time spent on readjustment. The entire control process can maintain high adaptability and anti-interference performance under various working conditions. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0055] Figure 1 A flowchart illustrating a digital collaborative control method for railway bridge erecting machine operations based on multi-condition perception.

[0056] Figure 2 This is a comparison diagram of the frequency domain correction of the speed command waveform of the beam drop driver in this invention;

[0057] Figure 3 This is a graph showing the changes in the multi-dimensional stability convergence index during the beam placement process in this invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0059] The present invention will be further described below with reference to embodiments.

[0060] Example:

[0061] This embodiment presents a digital collaborative control method for the operation process of a railway bridge erecting machine based on multi-condition perception, such as... Figure 1 As shown, it includes:

[0062] The entire digital collaborative control system is divided into a three-layer hardware architecture: the perception layer consists of the aforementioned displacement, tilt, and pressure sensing nodes and synchronous acquisition modules; the control layer uses an ARM Cortex-A9 industrial motion controller, which has 8 built-in 4~20mA analog output channels, an RS485 bus interface, an IEEE1588 clock synchronization module, power-off retention SRAM, and a solid-state drive; the execution layer is equipped with 4 electro-hydraulic servo proportional valves, with each analog output channel corresponding one-to-one with a single hydraulic cylinder servo valve. The controller outputs a corrected analog voltage signal to directly drive the servo valve core, adjusting the lifting speed of the hydraulic cylinder. The three hardware layers are connected by shielded cables to avoid signal distortion caused by electromagnetic disturbances at the construction site.

[0063] When the beam is moved into place, the measurement values ​​of each working condition sensing node are collected synchronously to generate the original working condition data frame carrying the process identifier. The process identifier represents the current beam moving and positioning process.

[0064] The working condition sensing nodes include at least displacement sensing nodes installed at the front and rear supports of the main beam of the bridge erecting machine, tilt sensing nodes installed on both sides of the hoisting drum of the overhead crane, and pressure sensing nodes installed in the oil circuit of the beam lowering hydraulic cylinder.

[0065] The configuration normalization operation is as follows:

[0066] A global base coordinate system is constructed with the rotation center of the bridge erecting machine as the origin. Based on the factory installation calibration parameters of each sensing node, the rotation matrix of the local measurement coordinate system of each node relative to the global base coordinate system is obtained. With translation vector The local physical quantity measurements output by each node By transforming the relationship Unified representation quantity projected onto the global base coordinate system Then, the standardized pose state vector is obtained by weighting each unified representation quantity according to the node type. ,in Characterizes the spatial position components of the beam's centroid in the global base coordinate system. Characterizes the attitude angular components of the beam about the three coordinate axes of the global base coordinate system;

[0067] During the synchronous acquisition of measurement values ​​from each sensing node under various operating conditions, sensor data self-tests are performed in parallel. For the i-th sensing node, a Kalman filter is used to obtain the current filtered estimate based on its historical measurement sequence. And calculate the actual measured value. With the filtered estimate residuals between ;

[0068] Construct the real-time confidence weights for the sensing node based on the residuals:

[0069] ;

[0070] In the formula: The standard deviation of the measurement noise is statistically obtained for the i-th sensing node under offline calibration conditions;

[0071] The above formula fits the physical characteristic that the sensor measurement noise follows a Gaussian distribution. It combines the aforementioned residual with the standard deviation of the sensor offline calibration noise to construct a continuously changing confidence weight. Unlike the simple binary judgment logic of traditional sensors that is effective / failed, it can dynamically adjust the proportion of single-node data in pose fusion calculation according to the degree of data deviation. Sensors with slight abnormalities can still retain low weights to participate in the calculation, thereby reducing the frequency of whole-machine control interruption caused by small disturbances of a single sensor.

[0072] when When the value is below the preset confidence threshold, the sensing node is marked as a failed node and its current measurement value is removed. At the same time, the measurement values ​​of the remaining valid sensing nodes are used for configuration normalization by weighted averaging.

[0073] When the number of effective sensing nodes is lower than the preset minimum number of nodes, the output drive signal sequence is paused and a safety alarm is triggered.

[0074] The system distinguishes between two levels of alarms:

[0075] A single sensor failure triggers a Level 1 early warning alarm, which only displays the failed sensor number and real-time confidence level in a pop-up window on the human-machine interface, without interrupting the current process.

[0076] If the number of valid nodes falls below the preset minimum threshold (the minimum number of valid nodes in this embodiment is 5), a level 2 shutdown safety alarm is triggered. All beam dropping driver servo valve outputs are immediately locked, and the current stroke position of the hydraulic cylinder remains stationary. At the same time, the failure node number, the current full sensor data, and the process status package are written to the fault log. After the fault is cleared and manually confirmed for reset, the controller re-executes the sensor self-test process. Once the confidence level of all nodes meets the standard, the beam dropping control is restored by reading the process transition status package before shutdown, without the need to re-collect pose data from zero.

[0077] Wherein, the pre-set confidence threshold for the i-th sensing node is denoted as ,according to Confirmed, among which The residual tolerance coefficient, which is pre-defined for this node type, characterizes the maximum number of times that the measurement residual of this node is allowed to deviate from its noise standard deviation.

[0078] The sensor Kalman filter iteration is performed in two steps:

[0079] Prediction step: Q represents the process noise covariance;

[0080] Update step: Kalman gain Filtered estimate Update the covariance P = (1-K)P;

[0081] In this embodiment, the displacement sensor has a residual tolerance coefficient λ=3, an inclination angle λ=2.5, and a pressure λ=2.

[0082] displacement sensor Taking 0.02mm as an example, the confidence cutoff threshold is calculated by substituting it into the threshold formula. If the weight is lower than the calculated value in real time, the sensor data in the current frame will be directly removed.

[0083] The above formula sets a residual tolerance coefficient separately based on the measurement accuracy differences of different types of sensing nodes, and derives the confidence cutoff critical value of each node in reverse based on the Gaussian distribution function. Instead of using the general judgment rule of a unified threshold for all types of sensors, displacement, tilt, and pressure sensing nodes can be matched with the measurement deviation range allowed by their respective working conditions, thereby completing the differentiated data screening of multiple types of heterogeneous sensing nodes.

[0084] The original working condition data frame is normalized to obtain a standardized pose state vector that is independent of the installation configuration of the sensing node;

[0085] Based on the preset mapping relationship between the standardized pose state vector and the beam dropping motion parameters, the standardized pose state vector is directly converted into a beam dropping motion parameter set, which includes the target stroke and motion sequence of each beam dropping actuator.

[0086] The default mapping relationship is constructed as follows:

[0087] The beam dropping process feature database is constructed from historical process samples. Each historical sample records the corresponding historical standardized pose state vector, the support reaction force vector of each beam dropping actuator, the actual stroke vector, and the actual action timing vector.

[0088] When constructing the predefined mapping relationship, it is constructed as a function derived from the inverse kinematics. , Dynamic calibration subfunction With time-series planning subfunction The nonlinear mapping function F is formed by cascading; using historical samples as the training set, combined with the current beam self-weight and eccentric load distribution, the load distribution matrix of the actuator is fitted.

[0089] Based on the deviation vector between the current attitude angle components and the target attitude angle components, the kinematic mapping relationship is solved to obtain the initial stroke value of each actuator;

[0090] Based on the load distribution matrix and the current beam self-weight and eccentric load distribution, the support reaction force of each actuator is solved, and the elastic deformation compensation amount is determined according to the structural stiffness relationship. The sum of the initial stroke value and the elastic deformation compensation amount is used as the target stroke of each actuator.

[0091] Taking the start moment of each driver as the optimization variable, the optimal start moment that minimizes the action conflict of each driver is solved according to the pose disturbance coupling relationship between the drivers and the preset constraints of driver speed, acceleration and safety time interval, and the optimal start moment is output as the action timing;

[0092] where the historical sample record is , where is the historical normalized pose state vector, is the measured support reaction force vector of each beam lowering driver, is the corresponding actual stroke vector of each driver, is the corresponding actual action timing vector, and m is the total number of beam lowering drivers;

[0093] When constructing the nonlinear mapping function F, historical samples are first used as the training set, and the locally weighted regression method is adopted to fit the current beam dead weight G and eccentric load distribution lower driver load distribution matrix , the eccentric load distribution in are respectively the offsets of the beam centroid relative to the design centroid in the x-direction and y-direction of the global base coordinate system;

[0094] where the beam lowering process feature database is stored in the industrial solid-state drive matched with the controller, and a single historical sample complete storage fields include: operation date, beam model, curve radius, longitudinal slope gradient, beam dead weight G, eccentricity offset , global timestamp, normalized pose vector , m driver support reaction force vectors , each driver actual stroke vector , driver start / stop timing vector ;

[0095] Sample collection only retains the working condition data of normal beam lowering, no sensor failure, and reaching the positioning deviation standard, and abnormal samples with pose deviation exceeding 2 times the threshold are eliminated after collection; after each complete beam lowering process is completed, the system automatically generates a new sample and incrementally writes it into the database, and automatically executes locally weighted regression during the low-load period in the early morning of each month, and refits and updates the driver load distribution matrix based on all new samples ;

[0096] Inverse kinematics sub-function , set the attitude angle component in the current normalized pose state vector as , and the target beam lowering positioning attitude as , then the attitude deviation vector is Construct the forward kinematics function of the bridge erecting machine's end-effector posture using the stroke of each actuator as input variables. And the forward kinematics function in the current travel A first-order Taylor expansion is performed to obtain the linear mapping relationship between attitude deviation and stroke adjustment. ,in Given the kinematic Jacobian matrix, solve the system of linear equations using the least squares solution. Obtain the initial travel values ​​for each driver. ,in for The Moore-Penrose generalized inverse matrix, This represents the stroke adjustment amount for the i-th driver;

[0097] In this embodiment, the number of beam-dropping actuators, m, is four, forming a four-point redundant support structure; the kinematic Jacobian matrix... Numerical solution is achieved by discrete differentiation of central difference: a small perturbation ΔL=0.01mm is applied to the stroke of a single actuator, and the difference in attitude angle of the end of the bridge erecting machine before and after the perturbation is solved. The ratio of the attitude difference to the small perturbation is used as the corresponding row and column elements of the Jacobian matrix.

[0098] Moore-Penrose generalized inverse In the embedded controller, the Singular Value Decomposition (SVD) algorithm is used for programming implementation. Components with singular values ​​less than 1e-6 are truncated to eliminate matrix singularity problems, and the resulting travel adjustment amount is obtained. Superimpose the current actual travel of the driver to output the initial travel value without overload equalization. ;

[0099] The above formula uses the Moore-Penrose generalized inverse matrix to solve the Jacobi linear equation system at least squares. It is suitable for redundant drive structures where the number of beam drop actuators is greater than the degree of freedom of attitude adjustment. The initial adjustment stroke is obtained by superimposing the current actual stroke after the stroke increment is corrected by multiple actuators. This can balance the attitude adjustment task of each hydraulic cylinder when there is redundant drive, and avoid overload of a single cylinder.

[0100] Dynamic calibration subfunction In the middle, according to the load allocation matrix and the current beam self-weight G and eccentric load distribution Solve for the current support reaction force of each actuator according to the static equilibrium equation. And construct the structural stiffness matrix. Structural stiffness matrix It is an m×m symmetric positive definite matrix, where the diagonal elements represent the vertical stiffness of each actuator support point, and the off-diagonal elements represent the load transfer coupling stiffness between the support points; This represents the vector of elastic deformation compensation for each actuator, obtained by solving a system of linear equations. get and with As the target journey of the i-th driver, where ;

[0101] It should be noted that:

[0102] Structural stiffness matrix The static load test was used to calibrate the following: Under no-load conditions, the rated load of 20%, 40%, 60%, 80%, and 100% was applied to each hydraulic cylinder for beam lowering in stages, and the vertical compression deformation of the four hydraulic cylinders was collected simultaneously.

[0103] When a single cylinder is pressurized, the ratio of its own load to its own deformation is the diagonal vertical stiffness of the matrix, while the ratio of the deformation of the other pressurized cylinders to the load of the pressurized cylinder is the non-diagonal coupled stiffness. A 4×4 symmetric positive definite stiffness matrix is ​​obtained by linear fitting of all loading data and is fixed into the control program. During operation, the support reaction force of each cylinder is read in real time. Matrix inversion operation The elastic deformation compensation of each cylinder is output, and the final target stroke is obtained by superimposing the initial stroke value. ;

[0104] The above formula is based on a symmetric positive definite stiffness matrix that includes the inter-cylinder coupling stiffness. It combines the actual bearing support reaction force of each hydraulic cylinder to solve the elastic deformation compensation caused by the structure under pressure. The deformation compensation is superimposed on the initial stroke obtained by kinematic solution to form the final target stroke. It balances the two types of influencing factors: geometric attitude adjustment and load-induced structural deformation, and eliminates the positioning error caused by relying solely on kinematic solution under heavy load beam dropping conditions.

[0105] Timing Planning Subfunction In this context, the coupling influence coefficient matrix C is defined as an m×m matrix, whose elements... This represents the pose perturbation rate generated at the end of the i-th driver when the j-th driver operates at a unit speed; This represents the velocity-time function vector of each actuator during the common action period, where Let be the real-time speed of the i-th drive, and let be the startup time of each drive. To optimize the design variables, the velocity-weighted conflict energy of each actuator during the common action period is used as the objective function. and at the maximum permissible speed of each drive. Maximum permissible acceleration Given the constraint that the absolute value of the time difference between the start-up times of adjacent drivers is not less than a preset safety time interval, a sequential quadratic programming method is used to iteratively solve the objective function. Minimize the optimal startup time of each driver The optimal start time is used as the action timing output.

[0106] Specifically:

[0107] In this embodiment, the maximum permissible speed of the single-beam hydraulic cylinder is set to 0.8 mm / s, the maximum acceleration is set to 0.2 mm / s², and the minimum safe time interval between the start-up times of any two actuators is set to 500 ms; the coupling influence coefficient matrix C is calibrated through a single-cylinder motion disturbance test.

[0108] Individually control a single actuator to move at a constant speed, and collect the beam attitude disturbance rate corresponding to all other actuators. The ratio of the disturbance rate to the single cylinder's motion speed is used as a matrix correspondence. Elements; the termination condition for the sequential quadratic programming iteration is set as the difference between two iterations of the objective function being less than 1e-8, the maximum number of iterations is limited to 200, and the iteration output is the optimal start time that satisfies all constraints. As the timing of the beam lowering action;

[0109] The above objective function introduces the pose disturbance coupling coefficient matrix between the actuators, integrates and statistically calculates the comprehensive pose disturbance energy generated by the synchronous action of multiple cylinders throughout the entire action period, and takes the minimum disturbance energy as the core objective of timing optimization. This is different from the simple timing constraint that only restricts the action time interval. It optimizes the start time of each cylinder from the root cause of beam attitude disturbance and reduces the beam sway amplitude during the multi-cylinder linkage process.

[0110] Based on the dynamic response characteristics of the bridge erecting machine's driver, the set of parameters for the beam dropping action is compiled into a sequence of driving signals, which characterizes the time waveform of the control signals of each driver.

[0111] The dynamic response characteristics of the bridge erecting machine's drive system are characterized by the frequency response curves and step response rise times of the electro-hydraulic servo system of each drive system.

[0112] When compiling the beam dropping motion parameter set into a drive signal sequence, for each beam dropping driver, first, based on its target stroke... and the start time in the action sequence Termination time Generate an ideal trapezoidal velocity profile, and then determine the phase hysteresis in the frequency response curve of the driver. and amplitude attenuation The ideal trapezoidal velocity profile is corrected in the frequency domain to obtain the corrected velocity command waveform. Finally, the speed command waveform According to the voltage-speed conversion factor of the driver control interface Converted to analog voltage signal waveform The analog voltage signal waveforms of each driver are aligned with the synchronous clock edge to form a drive signal sequence;

[0113] At the implementation level:

[0114] The frequency response curve of the driver was calibrated using a sinusoidal sweep frequency test.

[0115] Input a constant amplitude sinusoidal speed command of 0.01~10Hz, record the actual output speed amplitude and phase of the hydraulic cylinder, and plot the frequency response curve. The -3dB cutoff frequency is taken as 3Hz in this embodiment.

[0116] The frequency domain correction software process is as follows:

[0117] 1. Discretize the ideal trapezoidal velocity profile into a time-domain sequence at 10ms sampling intervals;

[0118] 2. Perform a DFT transform on the discrete sequence to obtain each frequency component;

[0119] 3. Each frequency component is multiplied by the corresponding frequency amplitude attenuation. Superimposed phase lag ;

[0120] 4. The corrected speed command waveform, which eliminates servo hysteresis, is obtained by inverse DFT reconstruction. Then it is converted into an analog voltage drive signal;

[0121] In the frequency domain correction, the frequency range is the effective operating frequency band of each driver electro-hydraulic servo system, which is determined by the -3dB cutoff frequency of the frequency response characteristic curve.

[0122] During the process of compiling the beam lowering action parameter set into a drive signal sequence, the real-time monitoring value of the oil temperature of the electro-hydraulic servo system of each driver is also used. Real-time monitoring value of system pressure For voltage-speed conversion coefficient Real-time corrections are performed, and the correction relationship is as follows:

[0123] ;

[0124] In the formula: For the drive at the calibrated oil temperature and calibration system pressure The base conversion factor is below. and These are the oil temperature compensation coefficient and the pressure compensation coefficient, respectively.

[0125] The above formula is based on the voltage-speed conversion reference coefficient under the calibration working condition, and superimposed with the linear compensation term of the difference between oil temperature, system pressure and calibration working condition. It incorporates the servo response drift caused by changes in hydraulic oil viscosity and system oil supply pressure into the real-time correction range, gets rid of the limitation that fixed conversion coefficient cannot adapt to variable temperature and pressure working environment, and continuously maintains the correspondence accuracy between voltage command and hydraulic cylinder movement speed.

[0126] Using the corrected real-time conversion coefficients Recalculate the analog voltage signal waveform and replace the signal amplitude of the corresponding driver at the corresponding moment in the drive signal sequence with the updated analog voltage signal waveform;

[0127] Among them, oil temperature compensation coefficient By testing the electro-hydraulic servo system at the calibrated oil temperature Constant-amplitude speed command step response tests were conducted under multiple different oil temperature conditions in the vicinity. The relative change rate of the measured steady-state speed at each test oil temperature to the measured steady-state speed at the calibrated oil temperature was used as the dependent variable, and the difference between each test oil temperature and the calibrated oil temperature was used as the independent variable. A univariate linear regression was used to fit the slope coefficients; pressure compensation coefficients were also included. By calibrating the electro-hydraulic servo system under system pressure The constant amplitude velocity command step response test was carried out under multiple different system pressure conditions in the vicinity. The relative change rate of the steady-state velocity measured under each test pressure to the steady-state velocity measured under the calibrated pressure was used as the dependent variable, and the difference between each test pressure and the calibrated pressure was used as the independent variable. The slope coefficient was obtained by fitting the data using univariate linear regression.

[0128] Among them, oil temperature compensation coefficient The calibration test range was set at hydraulic oil temperatures of 30℃, 40℃, 50℃, 60℃, and 70℃, with the system pressure fixed at the calibration value. =16MPa, with the same constant speed command input at each oil temperature, the steady-state hydraulic cylinder movement speed is collected; and the oil temperature difference is used as the data. With the independent variable being velocity and the relative rate of change being the dependent variable, the slope is calculated using a two-point linear regression. ;

[0129] Pressure compensation coefficient Fixed calibrated oil temperature At 45℃, system pressures of 10MPa, 13MPa, 16MPa, 19MPa, and 22MPa were used. Steady-state velocity was collected under the same constant speed command, and the pressure difference was linearly fitted to the velocity change rate. ;

[0130] This embodiment =-0.002 / ℃, =0.015 / MPa, solidified until the control program corrects the voltage-speed conversion coefficient in real time;

[0131] During the output of the drive signal sequence, the state feedback of each driver is continuously collected, and the state feedback is compared with the predetermined response state at the corresponding time in the drive signal sequence point by point to generate the compensation correction amount.

[0132] When generating the compensation correction amount by comparing the state feedback with the predetermined response state at corresponding times in the driving signal sequence point by point, the current sampling time is set as... The collected state feedback vector is In the driving signal sequence The predetermined response state vector at time t is Define the synchronization error vector ;

[0133] Based on the synchronization error vector, the following compensation correction amount is constructed. Generative law:

[0134] ;

[0135] In the formula: ⊙ represents the Hadamard product; These are the proportional gain vector, integral gain vector, and differential gain vector, respectively. The j-th sampling interval; N is the number of sampling points within the integration backtracking window; This is the first-order difference estimate of the synchronization error. is the nonlinear friction compensation gain vector; sign(⋅) is the sign function; β is a pre-defined nonlinear exponential coefficient between 0 and 1;

[0136] The above formula integrates three linear correction terms: proportional, integral, and derivative, and adds a friction compensation term with a nonlinear exponent. It uses the Hadamard product to independently correct the gain vector and error vector of the multi-driver separately. The integral term adapts to different beam movement speeds through a variable-length backtracking window. The nonlinear term is specifically designed to offset the control distortion caused by the low-speed crawling of the hydraulic cylinder. Unlike traditional PID controllers, it can simultaneously suppress steady-state deviation, dynamic lag, and low-speed friction disturbance.

[0137] Width of the integral backtracking window The adjustment is adaptive based on the beam movement speed during the current process;

[0138] Among them, the sign function The value can be:

[0139] when When >0, take +1; when When < 0, take -1; when When =0, take 0;

[0140] The nonlinear exponential coefficient β ranges from 0.3 to 0.7. Considering the prominent low-speed crawling phenomenon of hydraulic cylinders during the bridge erecting machine's beam lowering operation, the specific value of β is determined based on the ratio of the lowest stable motion speed measured by the actuator under both no-load and full-load conditions. The larger the ratio, the closer β is to 0.3; the smaller the ratio, the closer β is to 0.7. Intermediate ratios are determined by linear interpolation.

[0141] In the controller program, the symbolic functions are implemented using segmented conditional statements. Output +1 if greater than 0, output -1 if less than 0, output 0 if equal to 0; power terms Logarithmic exponential conversion is used to avoid floating-point overflow;

[0142] Step-by-step rules for β interpolation calculation:

[0143] First step: minimum stable speed under no-load Minimum stable speed under full load The ratio of the two is calculated and denoted as R;

[0144] The second step sets the upper limit of the ratio to 5 and the lower limit to 1, corresponding to a lower limit of 0.3 and an upper limit of 0.7 for β;

[0145] The third step is if 1≤R≤5, ;

[0146] When R is greater than 5, β is fixed at 0.3; when R is less than 1, β is fixed at 0.7.

[0147] Width of the integral backtracking window Based on the beam movement speed of the current process Adjust according to the following segmented adaptive rules:

[0148] when When the speed exceeds the preset high-speed threshold, the width is taken as the first preset time length T1. When the speed is less than the preset low-speed threshold, the width is taken as the second preset time length T2, where T2 is greater than T1. When the width is between the preset low-speed threshold and the preset high-speed threshold, the width continuously takes values ​​between T1 and T2 according to a linear interpolation method.

[0149] The preset high-speed threshold is preferably 40% of the rated maximum speed of the bridge beam during the beam dropping process of the bridge erecting machine; the preset low-speed threshold is preferably 1.5 times the lowest speed that the electro-hydraulic servo system can stably control; T1 is preferably 1 time the adjustment time of the step response of the electro-hydraulic servo system under rated operating conditions; and T2 is preferably 5 times the adjustment time of the step response of the electro-hydraulic servo system under rated operating conditions.

[0150] Optionally, in this embodiment, the rated maximum speed of beam dropping is 0.8 mm / s, the high-speed threshold is 40% of the rated speed, i.e., 0.32 mm / s, and the low-speed threshold is 1.5 times the minimum stable speed of the servo, i.e., 0.03 mm / s.

[0151] The measured step response adjustment time of the electro-hydraulic servo is 2s, therefore T1=2s and T2=10s;

[0152] When the beam speed When the speed is between 0.03 mm / s and 0.32 mm / s, the width of the integration window The closer the speed is to the low-speed threshold, the closer the integral window width is to 10s, thus enhancing the low-speed disturbance suppression effect.

[0153] The compensation and correction amount is superimposed on the drive signal sequence in real time to form a drive command for closed-loop regulation. The output continues until the status feedback of each driver reaches the condition for beam placement, and then the output stops.

[0154] The system has a preset single-drive synchronization error limit threshold of 0.2mm. When the synchronization error of any drive... When the limit threshold is exceeded for 20 consecutive sampling periods, the controller provides two levels of protection:

[0155] The first stage automatically reduces the speed of all drives to 20% of the rated speed, and simultaneously increases the length of the integral backtracking window to enhance disturbance suppression;

[0156] If the error still exceeds the limit after 300ms of low-speed operation, the secondary protection is executed, the drive voltage output is cut off, the hydraulic cylinder is locked and the audible and visual alarm is triggered, and the position, correction amount and sensor data at the time of exceeding the limit are recorded to the fault log. The beam position needs to be manually checked, reset and restart the process.

[0157] The conditions for beam placement are determined by a multi-dimensional stable convergence discriminant.

[0158] Set the target in-position pose as and Current sampling time The actual position in the state feedback after closed-loop regulation is The actual posture is Define the position deviation norm With attitude deviation norm ;

[0159] Simultaneously define a stability convergence index:

[0160] ;

[0161] In the formula: and These are the position component and the attitude angle component, respectively. and These are preset coefficients corresponding to the weights for the rate of change of position and the acceleration of attitude change, respectively. The sampling period for state feedback;

[0162] The above formula extracts the translational velocity and rotational angular acceleration of the beam through the first-order difference of position and the second-order difference of attitude, respectively. It is then weighted and summed with the weight coefficients that are inversely matched with the equivalent mass of the structure and the moment of inertia to comprehensively evaluate the overall vibration convergence of the beam. It does not rely solely on static position deviation to determine the positioning state, but can identify unstable working conditions where the position values ​​meet the standards but the beam continues to sway slightly.

[0163] when ≤ , ≤ and ≤ Simultaneously established and maintained for more than the preset stable time window. When the conditions for lowering the beam into place are met, the beam is deemed to have been lowered into place. , , These are the position deviation threshold, attitude deviation threshold, and stability convergence threshold, respectively.

[0164] Optional, location change rate weighting coefficient The weighting coefficients are set inversely proportional to the equivalent mass of the bridge erecting machine structure along each coordinate axis; that is, the larger the equivalent mass, the smaller the weighting coefficient for the rate of change of position along the coordinate axis. The values ​​of all values ​​are greater than 0 and not greater than 1. The minimum value of the position change rate weight coefficient among the six degrees of freedom is 1, and the position change rate weight coefficients of the other coordinate axes are determined according to this inverse proportional relationship.

[0165] Attitude change acceleration weighting coefficient The values ​​are set inversely proportional to the magnitude of the equivalent moment of inertia of the bridge erecting machine structure along each coordinate axis. That is, the larger the equivalent moment of inertia along a coordinate axis, the smaller the weighting coefficient for attitude change acceleration. The values ​​of all values ​​are greater than 0 and not greater than 1. The minimum value of the attitude change acceleration weight coefficient among the six degrees of freedom is 1. The attitude change acceleration weight coefficients in the other coordinate axes are determined according to this inverse proportional relationship.

[0166] Sampling period of state feedback The time constant should be set to no more than half of the overall closed-loop frequency of the bridge erecting machine's electrical and hydraulic control system.

[0167] Maintaining stability for longer than the preset stable time window The determination method is as follows:

[0168] When the number of sampling points in a continuous time series where all three inequalities mentioned above are met reaches more than 95% of the total number of sampling points within a preset stable time window, it is considered to have remained stable for more than the preset stable time window. It is taken as 1.5 times the settling time of the step response of the electro-hydraulic servo system under rated operating conditions;

[0169] The equivalent mass of the six degrees of freedom was calculated by static load displacement test of the beam. Standard loads were applied to the X, Y and Z directions to collect displacements, and the load-to-displacement ratio was the equivalent mass of each axis.

[0170] The equivalent moment of inertia was calculated by a torsional pendulum test. Torque was applied around the three axes and the torsional angle was collected. The ratio of torque to angular displacement was the moment of inertia.

[0171] Position deviation threshold The attitude deviation threshold is set according to the absolute value of the upper limit of the final positioning error of the bridge girder during the bridge erecting machine's beam lowering process. The stability convergence threshold is set according to the absolute value of the upper limit of the final positioning attitude angle error of the bridge girder during the bridge erection machine's beam lowering process. The stability convergence index is set according to the maximum value of the static fluctuation amplitude under the noise level of the state feedback sensor signal.

[0172] After determining that the beam placement condition has been met and terminating the current output drive signal sequence, the current process identifier is switched from the beam placement process to the next preset process identifier. At the same time, the standardized pose state vector at the termination time, the final stroke position of each actuator, and the compensation correction amount at the termination time are packaged into a process transition state package.

[0173] When switching to the next preset process, the final travel position in the process transition state package is used as the initial reference benchmark for the beam dropping action parameter set in the next preset process, and the compensation correction amount at the termination time is used as the feedforward compensation initial value at the start time of the next preset process and superimposed on the first drive signal sequence of the next preset process.

[0174] In addition to the process identifier, the raw work condition data frame also carries a global timestamp, frame number, and current job cycle number;

[0175] The original operating condition data frames and drive signal sequence frames are uniformly stored on the controller's solid-state drive, using a cyclic overwrite storage mechanism. Each frame occupies 256 bytes of storage space, completely preserving the operation frame data of the most recent 72 hours. When the storage space occupancy rate reaches 90%, the earliest historical frames of the past 24 hours are automatically deleted to release storage resources. It supports local export via USB and remote export via Ethernet. The exported file adopts the CSV standard format, and the fields include process identifier, global timestamp, sensor value, drive voltage, compensation and correction amount, and status feedback value, which facilitates later process review and model iteration updates.

[0176] Before synchronously collecting the measured values ​​of each sensing node under various operating conditions, a distributed clock synchronization reference is first established. The distributed clock synchronization reference eliminates the local clock drift of each node through master-slave clock message exchange, so that the acquisition time of each sensing node and the output time of the drive signal sequence share a global time base.

[0177] Distributed clock synchronization uses a simplified version of IEEE 1588 message exchange. The master controller acts as the clock master, sending a synchronization message every 200ms. Sensor nodes act as slaves, recording the local timestamps for message transmission and reception. The clock drift compensation for each slave is calculated based on the round-trip time of the messages. The formula for converting the local timestamp to the global synchronization timestamp is as follows:

[0178] , This is the real-time calculated slave clock drift compensation value; the synchronization deviation tolerance is set to 1 / 10 of the sampling period ΔT, i.e., 1ms. Synchronization error is only executed when the sequence number of the drive command frame and the status feedback frame are the same and the difference in the global synchronization timestamp is ≤1ms. Calculate and correct quantities, discarding time-mismatched data;

[0179] During the output of the drive signal sequence, the control quantity at each moment in the output drive signal sequence of each frame is bound to its corresponding frame number and global timestamp, and the local timestamp attached to the state feedback at the sampling moment is converted into a synchronization timestamp under a unified time base through the global time base. The point-by-point comparison is only performed between data pairs with the same frame number and a synchronization timestamp deviation less than the preset synchronization deviation tolerance.

[0180] The preset synchronization deviation tolerance can be optionally set to one-tenth of the sampling period of the state feedback.

[0181] In this embodiment, during the beam positioning stage, displacement, tilt, and pressure sensor node data are first synchronously collected from key locations such as the main beam support, the hoisting drum of the overhead crane, and the hydraulic circuit of the beam lowering cylinder. This generates an original working condition data frame carrying a process identifier. This multi-source sensing mechanism can comprehensively capture the spatial pose and load state of the beam, providing a real-time data foundation for subsequent precise control. Subsequently, considering the differences in the installation configuration of each sensing node, a global base coordinate system is established with the rotation center of the bridge erecting machine as the origin. Based on the factory calibration parameters of each node, the local measurements are unified to the global coordinate system through rotation and translation transformations, and weighted averaged to form a standardized pose state vector. This eliminates the influence of the installation configuration on data consistency and ensures the universality of subsequent mapping relationships.

[0182] Among them, the displacement sensing nodes of the front and rear support points of the main beam are magnetostrictive linear displacement sensors, model GY-250, which are fixed to the outside of the hydraulic cylinder of the main beam support leg using a clamp-type rigid bracket. The measurement range is 0~300mm, and the output is 4~20mA analog signal. The tilt angle sensing nodes on both sides of the hoist drum of the crane are MEMS dual-axis tilt sensors SST200, which are installed by fitting the end face of the hoist drum with a flange. The measurement range is ±5°, and the output is RS485 digital signal. The oil circuit pressure sensing node of the beam lowering hydraulic cylinder is a diffused silicon pressure transmitter PT124G, which is directly connected to the pressure measuring port of the rodless cavity of the hydraulic cylinder with a thread. The range is 0~32MPa, and the output is 4~20mA analog signal. All sensing nodes are equipped with an 8-channel analog acquisition module and an RS485 bus acquisition module. The synchronous sampling period ΔT=10ms is uniformly set. The acquisition module has a built-in hardware synchronous trigger circuit. The system main controller outputs a synchronous pulse signal to trigger all nodes to sample simultaneously, eliminating the timing difference of multi-sensor acquisition.

[0183] When constructing the global coordinate system on site, a total station was used in conjunction with reflective targets to calibrate the rotation center: four targets were evenly distributed on the inner and outer rings of the bridge erecting machine's slewing support. The slewing mechanism was controlled to rotate 90° in both the forward and reverse directions to collect the target coordinates. The intersection of the four sets of coordinates was fitted to determine the origin of the rotation center. The three-axis forward definition rules were as follows: the X-axis was along the longitudinal direction of the main beam of the bridge erecting machine, pointing in the direction of beam erection; the Y-axis was horizontal and perpendicular to the main beam, pointing in the inner side of the track curve; and the Z-axis was vertical and upward, perpendicular to the track plane. When each sensing node was calibrated at the factory, the sensors were installed on a standard fixture, with the fixture reference aligned with the three axes of the global coordinate system. Multiple sets of sensor readings were collected under the fixture reference. The rotation matrix and translation vector of each node's local coordinate system relative to the global coordinate system were solved by least squares fitting and stored in the controller's local Flash storage unit. During operation, the coordinate projection transformation was directly called up.

[0184] Furthermore, based on a preset nonlinear mapping function, the standardized pose state vector is directly converted into a set of beam dropping motion parameters. This mapping function consists of a cascade of three levels of sub-functions: inverse kinematics, dynamic calibration, and timing planning. First, the inverse kinematics is solved based on the deviation between the current attitude and the target attitude to obtain the initial values ​​of the stroke of each actuator. Then, combined with the beam's self-weight and eccentric load distribution, the elastic deformation compensation is calculated through the load distribution matrix and the structural stiffness matrix to correct the target stroke, thereby overcoming the positioning error caused by load variations and structural elastic deformation. Finally, taking the start-up time of each actuator as the optimization variable, considering the effects of motion coupling and safety interval constraints, the timing scheme with the minimum conflict energy is solved to ensure the stability and safety of the multi-actuator coordinated action.

[0185] After obtaining the beam dropping action parameters, an ideal speed profile is generated for the target stroke and timing based on the frequency response characteristics of each driver electro-hydraulic servo system. Phase and amplitude compensation is then performed, and the result is converted into an analog voltage signal waveform. The drive signal sequence is formed by aligning the synchronous clock edge. During this process, the voltage-speed conversion coefficient is dynamically corrected based on the real-time monitored oil temperature and pressure to compensate for the influence of temperature and pressure changes on the servo valve characteristics, thereby improving the environmental adaptability and execution accuracy of the control commands.

[0186] During the drive signal output process, the status feedback of each driver is continuously collected and compared point by point with the predetermined response status at the corresponding time in the drive signal sequence to generate a compensation correction amount. This correction amount integrates proportional, integral, derivative and nonlinear and friction compensation terms, which can effectively cope with disturbances such as low-speed crawling and nonlinear friction in the hydraulic system. The width of the integral backtracking window is adaptively adjusted according to the beam movement speed to ensure that disturbances are suppressed at low speeds and fast response at high speeds. Then the compensation amount is superimposed on the drive signal sequence in real time to form a closed-loop adjustment command, which is continuously output until the status feedback meets the beam placement conditions.

[0187] The beam placement determination adopts a multi-dimensional stable convergence discriminator, which simultaneously detects position deviation, attitude deviation and stability convergence index. The beam is determined to be in place only after all three conditions are continuously met within a preset time window. This multi-condition joint determination mechanism can avoid false triggering caused by instantaneous interference and improve the reliability of placement determination. After placement, the process identifier is switched to the next preset process and the process transition state package is packaged. The final travel position is used as the initial reference for the next process, and the compensation correction amount is used as the initial value of feedforward compensation, so that the control parameters between processes are smoothly inherited and respond quickly.

[0188] Process transition status packets are stored in the controller's power-down retainable SRAM memory area, containing four types of data:

[0189] Global timestamp, standardized pose vector, final stroke of 4 actuators, and real-time compensation and correction vector;

[0190] When the beam placement condition is met and the drive signal output is terminated, the process identifier switching logic is automatically triggered, switching the process identifier from beam placement code 001 to the code corresponding to the next beam feeding / leg conversion process; when the next process drive signal sequence is generated, the compensation and correction amount stored in the transition package is directly superimposed on the voltage command waveform at the first sampling moment, and the stroke value is used as the initial reference for the target stroke calculation of the new process, without the need to recalibrate the driver position from zero.

[0191] See Figure 2 As shown in the figure, the curves before and after frequency domain correction of the speed command of the beam drop driver are compared. The solid line is the unprocessed ideal trapezoidal speed profile (before correction), and the dashed line is the corrected speed command waveform after servo frequency domain compensation. The electro-hydraulic servo system has inherent amplitude attenuation and phase lag characteristics. Directly outputting the ideal trapezoidal wave will cause speed trajectory distortion.

[0192] This method first discretizes the trapezoidal velocity time-domain sequence, decomposes it into frequency components using DFT, compensates for each component based on the amplitude attenuation and phase lag of the driver's frequency response curve, and then generates a corrected waveform using inverse DFT. The figure clearly shows that the corrected waveform (dashed line) exhibits phase shift and amplitude contraction relative to the ideal contour (solid line), adapting to the system's 3Hz-3dB cutoff frequency response and offsetting servo dynamic lag errors. Combined with real-time correction of the voltage-speed conversion coefficient based on oil temperature and pressure, it further reduces waveform drift caused by environmental conditions, suppresses low-speed crawling of the hydraulic cylinder, and ensures that the actual movement trajectory of the hydraulic cylinder conforms to the planned target.

[0193] See Figure 3 As shown in the figure, this graph contains three sets of synchronous monitoring curves, corresponding to the changes in position deviation norm, attitude deviation norm, and stability convergence index with the beam dropping time, respectively. The gray horizontal line represents the judgment threshold for each index, and the gray interval represents the stability maintenance time window. In the initial stage of beam dropping, the beam swayed significantly, and the values ​​of position, attitude deviation, and vibration convergence index were all significantly higher than the threshold. With continuous correction by the closed-loop nonlinear PID, the three indices decreased synchronously and continuously.

[0194] Traditional control relies solely on position and orientation deviations to determine positioning, which can easily lead to misjudgments when numerical values ​​meet the standards but the beam continues to vibrate slightly. This invention adds a stability convergence index, quantifying the vibration amplitude through first-order position difference and second-order attitude difference. Position deviation, attitude deviation, and stability convergence index are all simultaneously below thresholds and remain below the preset stability time window before the beam is considered properly positioned. The curve shows that all three indices enter the threshold range and stabilize near the judgment time, avoiding premature termination caused by instantaneous disturbances and improving positioning accuracy and static stability.

[0195] In summary, the method described in the above embodiments normalizes the measured values ​​of multiple sensing nodes into a standardized pose state vector through configuration, and directly solves the beam dropping action parameters based on a preset mapping relationship. This avoids parameter tuning that relies on manual experience or repeated trial and error. The dynamic response characteristics of the actuator are used to correct the frequency domain of the speed command, making the control signal fit the actual capability of the actuator and reducing signal distortion error. When outputting the drive signal, the state feedback is continuously compared and a correction quantity containing nonlinear friction compensation is generated. With the help of an adaptive integral backtracking window, low-speed crawling of the hydraulic cylinder and external disturbances are suppressed. The real-time calculation of sensor confidence weights and the elimination of failed nodes ensure the reliability of the measured values ​​involved in the control, avoiding instability caused by single sensor anomalies. The position deviation, attitude deviation and motion stability are examined through a multi-dimensional stability convergence discriminator. A continuous holding time window is set to ensure that the beam is positioned to meet the positioning accuracy and has static stability. The process transition state package is recorded and transmitted to smoothly connect the control parameters of adjacent processes, reducing the impact of process switching and the time spent on readjustment. The entire control process can maintain high adaptability and anti-interference performance under various working conditions.

[0196] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital collaborative control method for the operation process of railway bridge erecting machines based on multi-condition perception, characterized in that, include: When the beam is moved into place, the measurement values ​​of each working condition sensing node are collected synchronously to generate an original working condition data frame carrying a process identifier, which represents the current beam moving and positioning process. The original working condition data frame is normalized to obtain a standardized pose state vector that is independent of the installation configuration of the sensing node; Based on the preset mapping relationship between the standardized pose state vector and the beam dropping motion parameters, the standardized pose state vector is directly converted into a beam dropping motion parameter set, which includes the target stroke and motion sequence of each beam dropping actuator. Based on the dynamic response characteristics of the bridge erecting machine's actuators, the set of parameters for the beam dropping action is compiled into a sequence of driving signals, which characterizes the time waveform of the control signals of each actuator. During the output of the drive signal sequence, the state feedback of each driver is continuously collected, and the state feedback is compared with the predetermined response state at the corresponding time in the drive signal sequence point by point to generate the compensation correction amount. The compensation and correction amount is superimposed on the drive signal sequence in real time to form a drive command for closed-loop regulation. The output continues until the status feedback of each driver reaches the condition for beam placement, and then the output stops.

2. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 1, characterized in that, The working condition sensing nodes include at least displacement sensing nodes installed at the front and rear supports of the main beam of the bridge erecting machine, tilt sensing nodes installed on both sides of the hoisting drum of the overhead crane, and pressure sensing nodes installed in the oil circuit of the beam lowering hydraulic cylinder. The configuration normalization operation is as follows: A global base coordinate system is constructed with the rotation center of the bridge erecting machine as the origin. Based on the factory installation calibration parameters of each sensing node, the rotation matrix of the local measurement coordinate system of each node relative to the global base coordinate system is obtained. With translation vector The local physical quantity measurements output by each node By transforming the relationship Unified representation quantity projected onto the global base coordinate system Then, the standardized pose state vector is obtained by weighting each unified representation quantity according to the node type. ,in Characterizes the spatial position components of the beam's centroid in the global base coordinate system. Characterizes the attitude angular components of the beam about the three coordinate axes of the global base coordinate system.

3. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 2, characterized in that, During the synchronous acquisition of measurement values ​​from each sensing node under various operating conditions, sensor data self-tests are performed in parallel. For the i-th sensing node, a Kalman filter is used to obtain the current filtered estimate based on its historical measurement sequence. And calculate the actual measured value. With the filtered estimate residuals between ; Construct the real-time confidence weights for the sensing node based on the residuals: ； In the formula: The standard deviation of the measurement noise is statistically obtained for the i-th sensing node under offline calibration conditions; when When the value is below the preset confidence threshold, the sensing node is marked as a failed node and its current measurement value is removed. At the same time, the measurement values ​​of the remaining valid sensing nodes are used for configuration normalization by weighted averaging. When the number of effective sensing nodes is lower than the preset minimum number of nodes, the output drive signal sequence is paused and a safety alarm is triggered.

4. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 1, characterized in that, The preset mapping relationship is constructed as follows: The beam dropping process feature database is constructed from historical process samples. Each historical sample records the corresponding historical standardized pose state vector, the support reaction force vector of each beam dropping actuator, the actual stroke vector, and the actual action timing vector. When constructing the predefined mapping relationship, it is constructed as a function derived from the inverse kinematics. , Dynamic calibration subfunction With time-series planning subfunction The nonlinear mapping function F is formed by cascading; using historical samples as the training set, combined with the current beam self-weight and eccentric load distribution, the load distribution matrix of the actuator is fitted. The Based on the deviation vector between the current attitude angle components and the target attitude angle components, the kinematic mapping relationship is solved to obtain the initial stroke value of each actuator; The Based on the load distribution matrix and the current beam self-weight and eccentric load distribution, the support reaction force of each actuator is solved, and the elastic deformation compensation amount is determined according to the structural stiffness relationship. The sum of the initial stroke value and the elastic deformation compensation amount is used as the target stroke of each actuator. The Using the startup time of each driver as the optimization variable, and based on the pose disturbance coupling relationship between each driver and the preset driver speed, acceleration and safety time interval constraints, the optimal startup time that minimizes the action conflict of each driver is solved, and the optimal startup time is used as the action timing output.

5. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 1, characterized in that, The dynamic response characteristics of the bridge erecting machine driver are characterized by the frequency response curves and step response rise times of the electro-hydraulic servo system of each driver. When compiling the beam dropping motion parameter set into a drive signal sequence, for each beam dropping driver, first, based on its target stroke... and the start time in the action sequence Termination time Generate an ideal trapezoidal velocity profile, and then determine the phase hysteresis in the frequency response curve of the driver. and amplitude attenuation The ideal trapezoidal velocity profile is corrected in the frequency domain to obtain the corrected velocity command waveform. Finally, the speed command waveform is... According to the voltage-speed conversion factor of the driver control interface Converted to analog voltage signal waveform The analog voltage signal waveforms of each driver are aligned with the synchronous clock edge to form the drive signal sequence.

6. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 5, characterized in that, During the process of compiling the beam lowering action parameter set into a drive signal sequence, the real-time monitoring value of the oil temperature of the electro-hydraulic servo system of each driver is also used. Real-time monitoring value of system pressure For the voltage-speed conversion coefficient Real-time corrections are performed, and the correction relationship is as follows: ； In the formula: For the drive at the calibrated oil temperature and calibration system pressure The base conversion factor is below. and These are the oil temperature compensation coefficient and the pressure compensation coefficient, respectively. Using the corrected real-time conversion coefficients The analog voltage signal waveform is recalculated, and the signal amplitude of the corresponding driver at the corresponding time in the drive signal sequence is replaced with the updated analog voltage signal waveform.

7. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 1, characterized in that, When generating the compensation correction amount by comparing the state feedback with the predetermined response state at corresponding times in the driving signal sequence point by point, the current sampling time is set as... The collected state feedback vector is In the driving signal sequence The predetermined response state vector at time t is Define the synchronization error vector ; Based on the aforementioned synchronization error vector, the following compensation correction amount is constructed. Generative law: ； In the formula: ⊙ represents the Hadamard product; These are the proportional gain vector, integral gain vector, and differential gain vector, respectively. The j-th sampling interval; N is the number of sampling points within the integration backtracking window; This is the first-order difference estimate of the synchronization error. is the nonlinear friction compensation gain vector; sign(⋅) is the sign function; β is a pre-defined nonlinear exponential coefficient between 0 and 1; The width of the integral backtracking window The adjustment is adaptive based on the beam movement speed during the current process; Among them, the sign function The value can be: when When >0, take +1; when When < 0, take -1; when When =0, take 0; The nonlinear exponential coefficient β ranges from 0.3 to 0.

7. Considering the prominent low-speed crawling phenomenon of hydraulic cylinders during the bridge erecting machine's beam lowering operation, the specific value of β is determined based on the ratio of the lowest stable motion speed measured by the actuator under both no-load and full-load conditions. The larger the ratio, the closer β is to 0.3; the smaller the ratio, the closer β is to 0.

7. Intermediate ratios are determined by linear interpolation. The width of the integral backtracking window Based on the beam movement speed of the current process Adjust according to the following segmented adaptive rules: when When the speed exceeds the preset high-speed threshold, the width is taken as the first preset time length T1. When the speed is less than the preset low-speed threshold, the width is taken as the second preset time length T2, where T2 is greater than T1. When the width is between a preset low-speed threshold and a preset high-speed threshold, the width is continuously taken between T1 and T2 using a linear interpolation method.

8. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 1, characterized in that, The beam placement conditions are determined by a multi-dimensional stable convergence discriminant. Set the target in-position pose as and Current sampling time The actual position in the state feedback after closed-loop regulation is The actual posture is Define the position deviation norm With attitude deviation norm ; Simultaneously define a stability convergence index: ； In the formula: and These are the position component and the attitude angle component, respectively. and These are preset coefficients corresponding to the weights for the rate of change of position and the acceleration of attitude change, respectively. The sampling period for state feedback; when ≤ , ≤ and ≤ Simultaneously established and maintained for more than the preset stable time window. When the conditions for lowering the beam into place are met, the beam is deemed to have been lowered into place. , , These are the position deviation threshold, attitude deviation threshold, and stability convergence threshold, respectively.

9. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 8, characterized in that, After determining that the beam placement condition has been met and terminating the current output drive signal sequence, the current process identifier is switched from the beam placement process to the next preset process identifier. At the same time, the standardized pose state vector at the termination time, the final stroke position of each actuator, and the compensation correction amount at the termination time are packaged into a process transition state package. When switching to the next preset process, the final travel position in the process transition state package is used as the initial reference benchmark for the beam dropping action parameter set in the next preset process, and the compensation correction amount at the termination time is used as the feedforward compensation initial value at the start time of the next preset process and superimposed on the first drive signal sequence of the next preset process.

10. The digital collaborative control method for railway bridge erecting machine operation process based on multi-condition perception as described in claim 1, characterized in that, In addition to carrying the process identifier, the original working condition data frame also carries a global timestamp, frame sequence number, and current job cycle number. Before synchronously collecting the measured values ​​of each working condition sensing node, a distributed clock synchronization reference is first established. The distributed clock synchronization reference eliminates the local clock drift of each node through master-slave clock message exchange, so that the acquisition time of each sensing node and the output time of the drive signal sequence share a global time base. During the output of the drive signal sequence, the control quantity at each moment in the output drive signal sequence of each frame is bound to its corresponding frame number and global timestamp, and the local timestamp attached to the state feedback at the sampling moment is converted into a synchronization timestamp under a unified time base through the global time base. The point-by-point comparison is only performed between data pairs with the same frame number and a synchronization timestamp deviation less than a preset synchronization deviation tolerance.

Images