Clamp intelligent control method and system based on pressure sensor array

Abstract

The application discloses a clamp intelligent control method and system based on a pressure sensor array, relates to the technical field of intelligent manufacturing equipment control, and comprises the following steps: combining linear static finite element analysis with topological sensitivity analysis, establishing an initial dynamic collaborative weight model based on the structural characteristics of a target part, and realizing mechanical basis modeling of a clamping strategy; combining pressure sensor array data fusion with a Sigmoid nonlinear mapping algorithm, constructing a contact state quality matrix reflecting the contact state of a clamp-workpiece, and improving the accuracy of clamping state identification; combining a sensitivity Jacobian matrix calibration mechanism with an improved conjugate gradient descent optimization algorithm, establishing a dynamic iterative solution system of clamping force distribution, realizing step-by-step optimization of dynamic collaborative weight, enabling adaptive coordination of clamping force distribution among multiple clamping points in the iterative process, and realizing optimal control of clamping force distribution.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing equipment control technology, and in particular to a method and system for intelligent control of fixtures based on pressure sensor arrays. Background Technology

[0002] With the rapid development of intelligent manufacturing and precision machining technologies, fixture technology is gradually evolving towards digitalization, networking and intelligence. In recent years, with the widespread application of sensor technology, finite element analysis and computational intelligent algorithms, intelligent fixture systems have begun to introduce pressure sensor arrays, distributed sensing modules and programmable logic controllers, enabling the distribution of contact force between the fixture and the workpiece to be acquired in the form of matrix data.

[0003] Existing intelligent clamp control technology still has shortcomings. Traditional independent adjustment strategies cannot accurately reflect the significant mechanical coupling relationship between each clamping point, resulting in excessive local pressure. It lacks a dynamic response mechanism to real-time disturbances. When the system is disturbed, manual intervention is often required, making it difficult to achieve efficient adaptive clamping. Existing methods have failed to construct a unified mathematical model of clamping force and contact quality, resulting in the inability to balance global convergence and real-time performance in multiple iterative adjustments. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a clamp intelligent control method and system based on a pressure sensor array to solve the problems of traditional independent adjustment strategies failing to accurately reflect the significant mechanical coupling relationship between clamping points, resulting in excessively high local pressure, lack of dynamic response mechanism to real-time disturbances, and the need for manual intervention when the system is disturbed, making it difficult to achieve efficient adaptive clamping. Existing methods have failed to construct a unified mathematical model of clamping force and contact quality, resulting in the inability to balance global convergence and real-time performance in multiple iterative adjustments.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a clamp intelligent control method based on a pressure sensor array, comprising,

[0008] Calculate the initial dynamic collaborative weights, collect pressure data on the target part using a pressure sensor array, and construct the contact state quality matrix.

[0009] Based on the contact state mass matrix, the target force at each clamping point of the target part is calculated, and the first round of real-time pressure map is generated. The error matrix between the first round of real-time pressure map and the ideal target pressure field matrix is ​​calculated. The square of the Frobenius norm of the error matrix is ​​used as the objective function value of the current iteration. The improved gradient descent method is used to iteratively optimize the initial dynamic collaborative weight. After reaching the maximum number of iterations, the clamping of the target part is completed, and the optimal dynamic collaborative weight is output.

[0010] The system determines the real-time disturbance of the target part and generates a disturbance rejection and recovery instruction sequence containing the optimal dynamic cooperative weights, and then performs disturbance rejection and recovery.

[0011] As a preferred embodiment of the clamp intelligent control method based on pressure sensor array described in this invention, the calculation of the initial dynamic cooperative weight includes:

[0012] Using linear statics finite element analysis combined with topology sensitivity analysis to set the initial dynamic synergistic weights, the clamping point position is set on the 3D CAD model of the target part to be clamped, and a unit normal clamping force is automatically applied. The equivalent stress of the entire workpiece under the action of this unit force system is calculated using a linear statics solver.

[0013] Based on equivalent stress, the topological sensitivity of any point in the model is calculated. For each clamping point, the topological sensitivity field in the influence area projected onto the workpiece surface is integrally integrated and averaged to obtain the comprehensive sensitivity index of the clamping point. The comprehensive sensitivity indices of all clamping points are normalized, and the absolute value of the normalized comprehensive sensitivity index is taken as the initial dynamic collaborative weight.

[0014] As a preferred embodiment of the clamp intelligent control method based on pressure sensor array described in this invention, the step of collecting pressure data on the target part using a pressure sensor array and constructing a contact state quality matrix includes:

[0015] After all the fixtures have pre-clamped the target part, a thin-film pressure sensor array embedded in the contact surface of the fixtures is used to collect pressure data on the target part.

[0016] The effective contact threshold is set based on sensor accuracy and noise level, and the saturation pressure threshold is set based on sensor range and workpiece safety.

[0017] The average of the effective contact threshold and the saturation pressure threshold is used as the center point of the curve;

[0018] Set the lower and upper limits of the target part quality coefficient based on processing requirements;

[0019] Based on the mathematical properties of the Sigmoid function, the relationship between the effective contact threshold and the saturation pressure threshold and the mass coefficient is expressed by its inverse function. Subtracting the two equations yields the steepness factor of the curve.

[0020] The pressure data collected by each pressure sensor is subjected to moving average filtering for noise reduction. Based on the steepness factor of the curve and the center point of the curve, the sigmoid function is used to continuously map the denoised pressure data to obtain the contact state quality coefficient of each sensor. The contact state quality coefficients of all sensors are then constructed into a contact state quality matrix according to the array position.

[0021] As a preferred embodiment of the clamping intelligent control method based on a pressure sensor array described in this invention, the step of calculating the target force at each clamping point of the target part and generating a first-round real-time pressure map includes:

[0022] Based on the contact state mass matrix, the sensor pressure values ​​are mapped to contact mass fractions using the ratio method, and the contact mass fractions are sorted in ascending order to generate an ordered clamping point processing queue.

[0023] Set a fixed window size, group the ordered clamping point processing queue, calculate the target force of each clamping point in each group based on the initial dynamic cooperative weight, define a fixed force value rise time window, and set the start time of each group.

[0024] Based on the start time of each group, the PLC requires that the clamping force be adjusted to the target force within the climbing time window, and that the first round of real-time pressure graph be generated after all clamping forces have been adjusted to the target force.

[0025] As a preferred embodiment of the clamp intelligent control method based on pressure sensor array described in this invention, the following steps are included: calculating the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, calculating the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, and iteratively optimizing the initial dynamic cooperative weights using an improved gradient descent method:

[0026] The ideal target pressure field matrix refers to the pressure field matrix obtained by using CAD finite element analysis on the target part;

[0027] The error matrix is ​​mapped to column vectors in column-major order. Based on the pre-calibrated sensitivity Jacobian matrix, the negative gradient vector for the current iteration is calculated using matrix-vector multiplication.

[0028] The pre-calibrated sensitivity Jacobian matrix establishes a linear relationship between the change in clamping point weight and the change in pressure sensor reading. The physical meaning of the matrix elements is: the amount of change in a sensor reading caused by a unit change in the weight of only one clamping point.

[0029] After obtaining the negative gradient vector, check the iteration counter. If the iteration counter is 0, the initial search direction is set to the negative gradient direction.

[0030] When the iteration counter is not 0, calculate the time series vector required to construct the conjugate gradient coefficients, including the gradient change and the dynamic co-weight change;

[0031] The conjugate coefficients are calculated using the function-valued enhanced conjugate gradient method, and a new search direction is synthesized based on the negative gradient vector of the current iteration and the search direction of the previous iteration.

[0032] The step size update phase includes initializing the search interval and parameters, and judging the nature of the trial step size in each iteration, including the following: the sufficient descent condition is not met, the sufficient descent condition is met but the curvature condition is not met, the strong Wolfe condition is met, and the sufficient descent condition is met but the curvature condition is not met and the derivative is positive.

[0033] The interval is shrunk based on the properties of the trial step size, and the optimal step size for the next iteration is determined.

[0034] If the sufficient descent condition is not met, then let the current iteration trial step size be the upper limit of the search interval for the next iteration, the current iteration search interval lower limit be the lower limit of the search interval for the next iteration, and set the lower limit of the search interval for the next iteration as the optimal step size for the next iteration.

[0035] If the sufficient descent condition is met but the curvature condition is not met, then the lower limit of the current iteration search interval is set to the lower limit of the search interval for the next iteration, and an extrapolation method is used to select a new trial step size in the search interval of the next iteration for the next step size update stage.

[0036] If the strong Wolfe condition is met, the search interval remains unchanged, and the trial step size of the current iteration is directly output as the optimal step size for the next iteration.

[0037] If the sufficient descent condition is met but the curvature condition is not met and the derivative is positive, then let the upper limit of the search interval for the next iteration be the current iteration's trial step size, and set the upper limit of the search interval for the next iteration as the optimal step size for the next iteration.

[0038] Check if the current interval length meets the preset tolerance. If it does, output the optimal step size; otherwise, proceed to the next step size update phase.

[0039] As a preferred embodiment of the intelligent control method for clamps based on a pressure sensor array described in this invention, the step of determining the real-time disturbance of the target part and generating an anti-disturbance recovery command sequence, and performing anti-disturbance recovery, includes:

[0040] Once the optimal dynamic collaborative weights are obtained, the disturbance observer is activated to observe the real-time disturbance distribution matrix for each control cycle, and the Frobenius norm of the real-time disturbance distribution matrix is ​​calculated as a quantitative indicator of the overall disturbance intensity.

[0041] The real-time disturbance distribution matrix refers to the difference between the real-time pressure matrix of the current period and the ideal target pressure field matrix;

[0042] The real-time pressure matrix is ​​obtained by the disturbance observer;

[0043] Based on historical review estimates, a disturbance threshold is set. If the quantitative indicator is less than the disturbance threshold, it is judged as normal fluctuation, and no active disturbance suppression is performed, only continuous monitoring is performed; otherwise, the disturbance suppression and recovery process is triggered immediately.

[0044] As a preferred embodiment of the intelligent control method for clamps based on pressure sensor arrays described in this invention, the disturbance rejection and recovery process refers to generating a disturbance rejection and recovery instruction sequence using an execution sequence generation function based on the optimal dynamic collaborative weight and the real-time pressure matrix of the current period. The motor immediately executes the disturbance rejection and recovery instruction sequence to clamp the target part.

[0045] Secondly, the present invention provides a clamp intelligent control system based on a pressure sensor array, comprising,

[0046] The computational building module is used to calculate the initial dynamic collaborative weights, collect pressure data on the target part using a pressure sensor array, and construct the contact state quality matrix.

[0047] The generation iteration module is used to calculate the target force at each clamping point of the target part based on the contact state mass matrix, generate the first round of real-time pressure map, calculate the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, calculate the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, and use the improved gradient descent method to iteratively optimize the initial dynamic collaborative weights. After reaching the maximum number of iterations, the clamping of the target part is completed, and the optimal dynamic collaborative weights are output.

[0048] The judgment and recovery module is used to judge the real-time disturbance of the target part and generate an anti-disturbance recovery instruction sequence containing the optimal dynamic cooperative weights to perform anti-disturbance recovery.

[0049] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the clamp intelligent control method based on a pressure sensor array as described in the first aspect of the present invention.

[0050] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the clamp intelligent control method based on a pressure sensor array as described in the first aspect of the present invention.

[0051] The beneficial effects of this invention are as follows: By combining linear static finite element analysis with topological sensitivity analysis, an initial dynamic collaborative weight model based on the structural characteristics of the target part is established, realizing the mechanical basis modeling of the clamping strategy. By combining pressure sensor array data fusion with the Sigmoid nonlinear mapping algorithm, a contact state quality matrix reflecting the contact state of the fixture and workpiece is constructed, improving the accuracy of clamping state identification. By combining the sensitivity Jacobian matrix calibration mechanism with the improved conjugate gradient descent optimization algorithm, a dynamic iterative solution system for clamping force distribution is established, realizing the gradual optimization of dynamic collaborative weights, enabling multiple clamping points to adaptively coordinate and distribute clamping force during the iteration process, and achieving optimal control of clamping force distribution. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart of the intelligent control method for a clamp based on a pressure sensor array in Example 1.

[0054] Figure 2 This is a schematic diagram of the clamp intelligent control system based on a pressure sensor array in Example 1.

[0055] Figure 3 This is a flowchart of the target force calculation and the first round of real-time pressure map generation in Example 1.

[0056] Figure 4 This is a system architecture diagram for solving dynamic collaborative weights using the improved gradient descent method in Example 1. Detailed Implementation

[0057] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0058] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0060] Example 1, referring to Figures 1 to 4 This is the first embodiment of the present invention, which provides a clamp intelligent control method based on a pressure sensor array, including the following steps:

[0061] S1. Calculate the initial dynamic collaborative weights, collect pressure data on the target part using a pressure sensor array, and construct the contact state quality matrix.

[0062] Preferably, the initial dynamic synergistic weight is set by using linear static finite element analysis combined with topology sensitivity analysis. On the three-dimensional CAD model of the target part to be clamped, the clamping point position is set, a unit normal clamping force is automatically applied, and the equivalent stress of the entire workpiece under the action of the unit force system is calculated using a linear static solver.

[0063] Based on equivalent stress, the topological sensitivity at any point within the model is calculated using the following formula:

[0064] ,

[0065] in, For point Topological sensitivity at that location For point Equivalent stress at the point, The elastic modulus of the target part;

[0066] For each clamping point, the topological sensitivity field within its projected influence area on the workpiece surface is integrally integrated over the area, and the average is calculated to obtain the comprehensive sensitivity index of that clamping point. The formula is:

[0067]

[0068] in, For clamping point The overall sensitivity index, For the integration region area, To clamp around the contact point on the target part's contact surface The local area is the clamping point A local area is defined by setting a radius around the center, based on the actual dimensions of the pressure head contact surface;

[0069] The overall sensitivity index of all clamping points is normalized, and the absolute value of the normalized overall sensitivity index is taken as the initial dynamic collaborative weight.

[0070] By setting the clamping point positions on the 3D CAD model of the target part to be clamped and automatically applying a unit normal clamping force, the equivalent stress distribution under the action of the unit force system is calculated using a linear statics solver, thus realizing the basic analysis of the workpiece's stress characteristics. By normalizing the comprehensive sensitivity index of all clamping points and taking its absolute value as the initial dynamic collaborative weight, dynamic comparability between different clamping points under a unified dimension is achieved.

[0071] Furthermore, after all the fixtures have pre-clamped the target part, a thin-film pressure sensor array embedded in the contact surface of the fixtures is used to collect pressure data on the target part.

[0072] The effective contact threshold is set based on sensor accuracy and noise level, and the saturation pressure threshold is set based on sensor range and workpiece safety.

[0073] The average of the effective contact threshold and the saturation pressure threshold is used as the center point of the curve;

[0074] Based on processing requirements, lower and upper limits are set for the target part's quality coefficient. Based on the mathematical properties of the Sigmoid function, the relationship between the effective contact threshold, saturation pressure threshold, and quality coefficient is expressed by their inverse function, as shown in the formula:

[0075] ,

[0076] ,

[0077] in, and These are the effective contact threshold and the saturation pressure threshold, respectively. and These are the lower and upper limits of the target part's quality coefficient, respectively. The center point of the curve;

[0078] Subtracting the two equations, we can obtain the formula for calculating the kurtosis factor of the curve:

[0079] ,

[0080] in, is the kurtosis factor of the curve;

[0081] The pressure data collected by each pressure sensor is subjected to moving average filtering for noise reduction. Based on the steepness factor and center point of the curve, the sigmoid function is used to continuously map the denoised pressure data to obtain the contact state quality coefficient of each sensor. The formula is:

[0082]

[0083] in, For the position located at the The contact state quality coefficient of the pressure sensor in row j. It is a natural exponential function. and These represent the kurtosis factor of the curve and the center point of the curve, respectively. The first one after noise reduction Pressure data collected by the pressure sensor in row j;

[0084] The contact state quality coefficients of all sensors are then constructed into a contact state quality matrix according to the array position.

[0085] After all fixtures pre-clamp the target part, pressure data is collected using a thin-film pressure sensor array embedded in the contact surface of the fixtures, enabling real-time monitoring of the clamping state. By substituting the effective contact threshold and the saturation pressure threshold into the inverse Sigmoid function, the calculation formula for the curve steepness factor is derived, transforming the clamping quality control problem into an adjustable mathematical mapping problem. This allows the system to switch between clamping strategies with different machining accuracies through parameterized control.

[0086] S2. Calculate the target force at each clamping point of the target part, generate the first round of real-time pressure map, calculate the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, calculate the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, use the improved gradient descent method to iteratively optimize the initial dynamic collaborative weights, and after reaching the maximum number of iterations, the target part is clamped and the optimal dynamic collaborative weights are output.

[0087] Preferably, based on the contact state mass matrix, the sensor pressure value is mapped to a contact mass fraction using a ratio method, and the contact mass fractions are arranged in ascending order to generate an ordered clamping point processing queue;

[0088] The ratio method mapping refers to dividing the sensor pressure value by the maximum pressure value measured on the target part, and using this as the contact quality fraction.

[0089] Set a fixed window size (based on manual experience, such as a window grouping 2-3 clamping points), group the ordered clamping point processing queue, and calculate the target force for each clamping point in each group based on the initial dynamic collaborative weights, using the formula:

[0090] ,

[0091] in, Let i be the target force at the i-th clamping point. The base preload is set manually. Let i be the initial dynamic cooperative weight of the i-th clamping force point. The saturation pressure threshold. The contact mass fraction at the i-th clamping force point;

[0092] Define a fixed force value climb time window, set the start time for each group, and use the following formula:

[0093] ,

[0094] in, Let m be the start time of the m-th group. This is the starting time for the first group. The climbing time window is determined based on the time required to complete the target force at each clamping point within a set.

[0095] Based on the start time of each group, the PLC (Programmable Logic Controller) is required to adjust the clamping force to the target force within the climbing time window, and generate the first round of real-time pressure map after all clamping forces are adjusted to the target force.

[0096] The value of each clamping force point in the first round of real-time pressure graph is given by the formula:

[0097] ,

[0098] in, For the first round of real-time pressure maps The pressure value at the i-th clamping point in the equation. The number of clamping force points. To iterate through the indices of the clamping force points, In time The instantaneous pressure value at the i-th clamping point is measured by a sensor. This is the end time.

[0099] By using the initial dynamic collaborative weights and contact state quality matrix, the target force of each clamping point is calculated. Then, the clamping force is controlled by PLC to generate the first round of real-time pressure map, ensuring that each clamping point can achieve "differentiated" loading in the first round of loading based on the surface contact characteristics of the part and the differences in local pressure distribution.

[0100] Furthermore, the ideal target pressure field matrix refers to the pressure field matrix obtained by using CAD finite element analysis on the target part;

[0101] The error matrix is ​​mapped to column vectors in column-major order. Based on the pre-calibrated sensitivity Jacobian matrix, the negative gradient vector for the current iteration is calculated using matrix-vector multiplication.

[0102] The pre-calibrated sensitivity Jacobian matrix establishes a linear relationship between the change in clamping point weight and the change in pressure sensor reading. The physical meaning of the matrix elements is: the amount of change in a sensor reading caused by a unit change in the weight of only one clamping point.

[0103] Acquisition method: Install a rigid calibration block on the machine tool worktable, control all electric actuators to apply a uniform reference clamping force, keep the force of all other clamping points unchanged, apply a small and known force increment to only one clamping point, calculate the pressure change vector of the clamping point based on the sensor, and normalize it as an element in the sensitivity Jacobian matrix;

[0104] Once the negative gradient vector is obtained, check the iteration counter. If the iteration counter is 0 (first iteration), the initial search direction is set to the negative gradient direction.

[0105] When the iteration counter is not 0, calculate the time series vector required to construct the conjugate gradient coefficients, including the gradient change and the dynamic co-weight change;

[0106] The formula for calculating the change in dynamic collaborative weight is as follows:

[0107] ,

[0108] in, This represents the change in dynamic collaborative weights between two adjacent iterations. For the first The step size for each iteration is determined empirically, with the initial step size set based on empirical methods. For the first The search direction for the next iteration;

[0109] The conjugate coefficients are calculated using the function-valued enhanced conjugate gradient method, as shown in the formula:

[0110] ,

[0111] in, Let be the conjugate coefficient of the k-th iteration. Let be the negative gradient vector of the k-th iteration. This represents the gradient change between two consecutive iterations. It is the transpose symbol. This represents the objective function value for the k-th iteration.

[0112] Based on the negative gradient vector of the current iteration and the search direction of the previous iteration, a new search direction is synthesized using the conjugate coefficient, as shown in the formula:

[0113]

[0114] in, This represents the search direction for the k-th iteration;

[0115] The step size update phase includes initializing the search interval and parameters, including the initial step size upper limit, the initial trial step size, the sufficient descent condition constant, the curvature condition constant, the search interval, and defining the scalar function;

[0116] The scalar function is defined by the following formula:

[0117] ,

[0118] in, For about scalar functions, For the first Dynamic collaborative weights in the next iteration;

[0119] Each iteration determines the nature of the trial step size, including: the sufficient descent condition is not met, the sufficient descent condition is met but the curvature condition is not met, the strong Wolfe condition is met, and the sufficient descent condition is met but the curvature condition is not met and the derivative is positive.

[0120] The interval is shrunk based on the properties of the trial step size, and the optimal step size for the next iteration is determined.

[0121] The sufficient descent condition is not met, formula:

[0122] ,

[0123] in, For about scalar functions, To be within the initial search interval, the first A tentative step size To test the meaning of the step size, for scalar function at point, For the constant of the fully descent condition, for Following the search direction directional derivative;

[0124] Let the current iteration's trial step size be the upper limit of the search interval for the next iteration, the current iteration's search interval's lower limit be the lower limit of the search interval for the next iteration, and set the lower limit of the search interval for the next iteration as the optimal step size for the next iteration.

[0125] The sufficient descent condition is met, but the curvature condition is not met. The formula is:

[0126] ,

[0127] in, Here is the curvature condition constant;

[0128] Then set the lower limit of the current iteration search interval to the lower limit of the search interval for the next iteration, and use extrapolation to select a new trial step size in the search interval for the next iteration, and proceed to the next step size update stage.

[0129] The strong Wolfe condition holds, as shown in the formula:

[0130] ,

[0131] If the search interval remains unchanged, the trial step size of the current iteration is directly output as the optimal step size for the next iteration.

[0132] The sufficient descent condition holds, but the curvature condition does not, and the derivative is positive. Formula:

[0133] ,

[0134] Then set the upper limit of the search interval for the next iteration to the current iteration's trial step size, and set the upper limit of the search interval for the next iteration to the optimal step size for the next iteration;

[0135] Check if the current interval length meets the preset tolerance (based on empirical methods, such as...). Its accuracy is far higher than the sensitivity of the system's physical sensing. If the preset tolerance is met, the optimal step size is output; otherwise, the next step size update phase begins.

[0136] By comparing the real-time pressure distribution measured by sensors with the ideal target pressure field matrix obtained from CAD finite element simulation, an error matrix is ​​constructed. The square of its Frobenius norm is used as the current iterative objective function value to quantify the overall deviation between the entire pressure field and the ideal state, providing a unified and continuously differentiable objective function for the optimization algorithm. Based on the error matrix, the sensitivity Jacobian matrix, and the negative gradient vector, the optimal dynamic collaborative weight is iteratively obtained using the function value-enhanced conjugate gradient method, realizing a dynamic self-learning multi-point collaborative control mechanism. An improved step size update mechanism and a strong Wolfe condition control iterative convergence are used to improve the robustness and convergence efficiency of the algorithm.

[0137] S3. Determine the real-time disturbance of the target part and generate a disturbance rejection and recovery command sequence to perform disturbance rejection and recovery;

[0138] Preferably, after obtaining the optimal dynamic collaborative weights, the disturbance observer is activated to observe the real-time disturbance distribution matrix of each control cycle, and the Frobenius norm of the real-time disturbance distribution matrix is ​​calculated as a quantitative indicator of the overall disturbance intensity.

[0139] Based on historical review estimates, a disturbance threshold is set. If the quantitative indicator is less than the disturbance threshold, it is judged as normal fluctuation, and no active disturbance suppression is performed, only continuous monitoring is performed; otherwise, the disturbance suppression and recovery process is triggered immediately.

[0140] By activating the disturbance observer after obtaining the optimal dynamic cooperative weights, the difference between the "real-time pressure matrix of the current period and the ideal target pressure field matrix" is calculated in real time, thus realizing an accurate spatial characterization of the deviations that occur in the system in each control period.

[0141] Furthermore, the disturbance rejection and recovery process refers to generating a disturbance rejection and recovery instruction sequence based on the optimal dynamic collaborative weight and the real-time pressure matrix of the current period using an execution sequence generation function. The motor immediately executes the disturbance rejection and recovery instruction sequence to clamp the target part.

[0142] By combining the optimal dynamic collaborative weights with the current periodic real-time pressure matrix in the disturbance rejection recovery process and using the execution sequence generation function to generate a recovery instruction sequence, a coordinated, weight-driven recovery command planning among multiple execution units is realized.

[0143] This embodiment also provides a clamp intelligent control system based on a pressure sensor array, including:

[0144] The computational building module is used to calculate the initial dynamic collaborative weights, collect pressure data on the target part using a pressure sensor array, and construct the contact state quality matrix.

[0145] The generation iteration module is used to calculate the target force at each clamping point of the target part based on the contact state mass matrix, generate the first round of real-time pressure map, calculate the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, calculate the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, and use the improved gradient descent method to iteratively optimize the initial dynamic collaborative weights. After reaching the maximum number of iterations, the clamping of the target part is completed, and the optimal dynamic collaborative weights are output.

[0146] The judgment and recovery module is used to judge the real-time disturbance of the target part and generate an anti-disturbance recovery instruction sequence containing the optimal dynamic cooperative weights to perform anti-disturbance recovery.

[0147] This embodiment also provides a computer device applicable to the intelligent control method of clamps based on pressure sensor arrays, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent control method of clamps based on pressure sensor arrays as proposed in the above embodiment.

[0148] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0149] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the intelligent control method for a clamp based on a pressure sensor array as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0150] In summary, this invention establishes an initial dynamic collaborative weight model based on the structural characteristics of the target part by combining linear static finite element analysis with topological sensitivity analysis, thereby realizing the mechanical foundation modeling of the clamping strategy. By combining pressure sensor array data fusion with the Sigmoid nonlinear mapping algorithm, a contact state quality matrix reflecting the contact state between the fixture and the workpiece is constructed, improving the accuracy of clamping state identification. By combining the sensitivity Jacobian matrix calibration mechanism with the improved conjugate gradient descent optimization algorithm, a dynamic iterative solution system for clamping force distribution is established, realizing the gradual optimization of dynamic collaborative weights. This allows multiple clamping points to adaptively and coordinately distribute clamping force during the iteration process, achieving optimal control of clamping force distribution.

[0151] It should be noted that 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A clamp intelligent control method based on a pressure sensor array, characterized in that: include, Calculate the initial dynamic collaborative weights, collect pressure data on the target part using a pressure sensor array, and construct the contact state quality matrix. Calculate the target force at each clamping point of the target part, generate the first round of real-time pressure map, calculate the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, calculate the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, use the improved gradient descent method to solve the dynamic cooperative weight, and after reaching the maximum number of iterations, the clamping of the target part is completed, and the optimal dynamic cooperative weight is output. The system determines the real-time disturbance of the target part and generates a disturbance rejection and recovery command sequence to perform disturbance rejection and recovery. The calculation of the initial dynamic collaborative weights includes: Using linear statics finite element analysis combined with topology sensitivity analysis to set the initial dynamic synergistic weights, the clamping point position is set on the 3D CAD model of the target part to be clamped, and a unit normal clamping force is automatically applied. The linear statics solver is used to calculate the equivalent stress of the entire target part under the action of the unit normal clamping force. Based on equivalent stress, the topological sensitivity of each clamping point in the 3D CAD model domain is calculated. For each clamping point, the topological sensitivity field in the region of its projection onto the surface of the target part is integrated by the area integral and averaged to obtain the comprehensive sensitivity index of the clamping point. The comprehensive sensitivity indices of all clamping points are normalized, and the absolute value of the normalized comprehensive sensitivity index is taken as the initial dynamic collaborative weight.

2. The intelligent control method for a clamp based on a pressure sensor array as described in claim 1, characterized in that: The process of collecting pressure data on the target part using a pressure sensor array and constructing a contact state quality matrix includes: After all the fixtures have pre-clamped the target part, a thin-film pressure sensor array embedded in the contact surface of the fixtures is used to collect pressure data on the target part. The effective contact threshold is set based on sensor accuracy and noise level, and the saturation pressure threshold is set based on sensor range and target part safety. The average of the effective contact threshold and the saturation pressure threshold is used as the center point of the curve; Set the lower and upper limits of the target part quality coefficient based on processing requirements; Based on the mathematical properties of the Sigmoid function, the relationship between the effective contact threshold and the saturation pressure threshold and the mass coefficient is expressed by its inverse function. Subtracting the two equations yields the steepness factor of the curve. The pressure data collected by each pressure sensor is subjected to moving average filtering for noise reduction. Based on the steepness factor of the curve and the center point of the curve, the sigmoid function is used to continuously map the denoised pressure data to obtain the contact state quality coefficient of each sensor. The contact state quality coefficients of the pressure sensors are then constructed into a contact state quality matrix according to the array position.

3. The intelligent control method for a clamp based on a pressure sensor array as described in claim 2, characterized in that: The calculation of the target force at each clamping point of the target part generates the first round of real-time pressure maps, including: Based on the contact state mass matrix, the sensor pressure values ​​are mapped to contact mass fractions using the ratio method, and the contact mass fractions are sorted in ascending order to generate an ordered clamping point processing queue. Set a fixed window size, group the ordered clamping point processing queue, calculate the target force of each clamping point in each group based on the initial dynamic cooperative weight, define a fixed force value climbing time window, and set the start time of each group. Based on the start time of each group, the PLC requires that the clamping force be adjusted to the target force within the force value ramp-up time window, and that the first round of real-time pressure graphs be generated after all clamping forces have been adjusted to the target force.

4. The intelligent control method for a clamp based on a pressure sensor array as described in claim 3, characterized in that: The calculation of the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, the calculation of the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, and the solution of the dynamic collaborative weights using the improved gradient descent method include: The ideal target pressure field matrix refers to the pressure field matrix obtained by using CAD finite element analysis on the target part; The error matrix is ​​mapped to column vectors in column-major order. Based on the pre-calibrated sensitivity Jacobian matrix, the negative gradient vector for the current iteration is calculated using matrix-vector multiplication. The pre-calibrated sensitivity Jacobian matrix establishes a linear relationship between the change in clamping point weight and the change in pressure sensor reading. The physical meaning of the matrix elements is: the amount of change in a sensor reading caused by a unit change in the weight of only one clamping point. After obtaining the negative gradient vector, check the iteration counter. If the iteration counter is 0, the initial search direction is set to the negative gradient direction. When the iteration counter is not 0, calculate the time series vector required to construct the conjugate gradient coefficients, including the gradient change and the dynamic co-weight change; The conjugate coefficients are calculated using the function-valued enhanced conjugate gradient method, and a new search direction is synthesized based on the negative gradient vector of the current iteration and the search direction of the previous iteration. The step size update phase includes initializing the search interval and parameters, and judging the nature of the trial step size in each iteration, including the following: the sufficient descent condition is not met, the sufficient descent condition is met but the curvature condition is not met, the strong Wolfe condition is met, and the sufficient descent condition is met but the curvature condition is not met and the derivative is positive. The interval is shrunk based on the properties of the trial step size, and the optimal step size for the next iteration is determined. If the sufficient descent condition is not met, then let the current iteration trial step size be the upper limit of the search interval for the next iteration, the current iteration search interval lower limit be the lower limit of the search interval for the next iteration, and set the lower limit of the search interval for the next iteration as the optimal step size for the next iteration. If the sufficient descent condition is met but the curvature condition is not met, then the lower limit of the current iteration search interval is set to the lower limit of the search interval for the next iteration, and an extrapolation method is used to select a new trial step size in the search interval of the next iteration for the next step size update stage. If the strong Wolfe condition is met, the search interval remains unchanged, and the trial step size of the current iteration is directly output as the optimal step size for the next iteration. If the sufficient descent condition is met but the curvature condition is not met and the derivative is positive, then let the upper limit of the search interval for the next iteration be the current iteration's trial step size, and set the upper limit of the search interval for the next iteration as the optimal step size for the next iteration. Check if the current interval length meets the preset tolerance. If it does, output the optimal step size; otherwise, proceed to the next step size update phase.

5. The intelligent control method for a clamp based on a pressure sensor array as described in claim 4, characterized in that: The process of determining real-time disturbances to the target part and generating a disturbance rejection and recovery command sequence for disturbance rejection and recovery includes: Once the optimal dynamic collaborative weights are obtained, the disturbance observer is activated to observe the real-time disturbance distribution matrix for each control cycle, and the Frobenius norm of the real-time disturbance distribution matrix is ​​calculated as a quantitative indicator of the overall disturbance intensity. Based on historical review estimates, a disturbance threshold is set. If the quantitative indicator is less than the disturbance threshold, it is judged as normal fluctuation, and no active disturbance suppression is performed, only continuous monitoring is performed; otherwise, the disturbance suppression and recovery process is triggered immediately.

6. The intelligent control method for a clamp based on a pressure sensor array as described in claim 5, characterized in that: The disturbance rejection and recovery process refers to generating a disturbance rejection and recovery instruction sequence based on the optimal dynamic collaborative weight and the real-time pressure matrix of the current cycle using an execution sequence generation function. The motor immediately executes the disturbance rejection and recovery instruction sequence to clamp the target part.

7. A clamp intelligent control system based on a pressure sensor array, based on the clamp intelligent control method based on a pressure sensor array according to any one of claims 1 to 6, characterized in that: include, The computational building module is used to calculate the initial dynamic collaborative weights, collect pressure data on the target part using a pressure sensor array, and construct the contact state quality matrix. The generation iteration module is used to calculate the target force at each clamping point of the target part, generate the first round of real-time pressure map, calculate the error matrix between the first round of real-time pressure map and the ideal target pressure field matrix, calculate the square of the Frobenius norm of the error matrix as the objective function value of the current iteration, use the improved gradient descent method to solve the dynamic cooperative weight, and after reaching the maximum number of iterations, the clamping of the target part is completed, and the optimal dynamic cooperative weight is output. The judgment and recovery module is used to judge the real-time disturbance of the target part and generate an anti-disturbance recovery command sequence to perform anti-disturbance recovery.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the clamp intelligent control method based on pressure sensor array as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the clamp intelligent control method based on pressure sensor array as described in any one of claims 1 to 6.

Images