Pose correction method, system, electronic device, and storage medium

Abstract

The application provides a posture correction method and system, an electronic device and a storage medium. The method comprises the following steps: acquiring thickness detection data, gap detection data, aerodynamic working condition data and posture image data in the process of pushing the board by a main propulsion device; performing feature fusion analysis on the thickness detection data, the gap detection data and the aerodynamic working condition data to determine the stacking state of the board, wherein the stacking state comprises a single board state and a multi-board state; performing feature extraction on the posture image data to obtain a position deviation value of the board and an angle deviation value of the board; generating a target control instruction based on the stacking state, the position deviation value and the angle deviation value, and controlling the main propulsion device and an auxiliary fine adjustment device to perform corresponding correction actions through the target control instruction.

Description

Technical Field

[0001] This application relates to the technical field of posture correction methods, and more particularly to a posture correction method, system, electronic device, and storage medium. Background Technology

[0002] In automated production lines for panel processing and furniture manufacturing, magazine-type feeders serve as crucial material flow and supply hubs, their operating principles and performance directly impacting the efficiency and stability of the entire production process. These devices typically employ a cylinder-based power drive mechanism, precisely controlling the reciprocating motion of the pusher plate to smoothly and steadily deliver pre-stacked panels one sheet at a time to downstream processing or assembly stations. With the increasing pursuit of production efficiency and the growing demand for flexible and customized production in modern manufacturing, production line operating cycles are accelerating, and task switching is becoming more frequent. Against this backdrop, unprecedentedly high standards and stringent requirements have been placed on the continuity and stability of panel transport, as well as the precision and synchronization between the feeder and downstream stations. Ensuring uninterrupted material flow and precise positioning has become a key factor in improving overall production line efficiency and guaranteeing consistent product quality.

[0003] Existing feeding systems often treat stack detection and attitude correction as independent processes and rely heavily on single-dimensional sensor data, making it difficult to resist interference signals in the dynamic environment of high-speed feeding. Summary of the Invention

[0004] This application provides a posture correction method, system, electronic device, and storage medium to solve the problems existing in related technologies. The technical solution is as follows: In a first aspect, embodiments of this application provide a posture correction method, including: Acquire thickness detection data, gap detection data, aerodynamic condition data, and attitude image data during the process of the main propulsion equipment pushing the plate; The thickness detection data, gap detection data and aerodynamic condition data are fused and analyzed to determine the stacking state of the sheet metal, which includes single sheet metal state and multiple sheet metal state. Feature extraction is performed on the posture image data to obtain the positional deviation and angular deviation values ​​of the plate. Based on the stacking status, position deviation value, and angle deviation value, target control commands are generated, and the main propulsion equipment and auxiliary fine-tuning equipment are controlled to perform corresponding correction actions through the target control commands.

[0005] Secondly, embodiments of this application provide a posture correction system, including: The first acquisition module is used to acquire thickness detection data, gap detection data, aerodynamic condition data and attitude image data during the process of the main propulsion equipment pushing the plate. The first determination module is used to perform feature fusion analysis on thickness detection data, gap detection data and aerodynamic condition data to determine the stacking state of the sheet metal. The stacking state includes single sheet metal state and multiple sheet metal state. The first module is used to extract features from the posture image data to obtain the positional deviation value and the angle deviation value of the plate. The first generation module is used to generate target control commands based on the stacking status, position deviation value, and angle deviation value, and to control the main propulsion equipment and auxiliary fine-tuning equipment to perform corresponding correction actions through the target control commands.

[0006] Thirdly, embodiments of this application provide an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the above-described posture correction method.

[0007] Fourthly, embodiments of this application provide a computer-readable storage medium that stores computer instructions, wherein when the computer instructions are executed on a computer, the methods in any of the above-described embodiments are performed.

[0008] The advantages or beneficial effects of the above technical solutions include at least the following: The posture correction method proposed in this embodiment acquires thickness detection data, gap detection data, aerodynamic condition data, and posture image data reflecting the material's dimensional dimensions during the main propulsion device's pushing of the sheet material. It then performs feature fusion analysis on the thickness detection data, gap detection data, and aerodynamic condition data, effectively filtering out local distortion interference from a single sensor in a dynamic pushing environment through cross-validation. This accurately analyzes and determines the current stacking state of the sheet material. Simultaneously, it extracts features from the posture image data to calculate the sheet material's positional and angular deviation values. Using the stacking state and spatial deviation values, it generates target control commands, providing the main propulsion device and auxiliary fine-tuning device with a global decision-making basis, enabling them to execute correction actions that strictly match the actual material conditions. The technical solution of this embodiment constructs a cross-validation mechanism between the external spatial dimensions of the material and the internal aerodynamic load of the actuator through thickness detection data, gap detection data, aerodynamic condition data, and posture image data. In dynamic conditions involving high-speed material feeding or vibration, when a sensor in one dimension generates interference signals due to local deformation or external factors, data from other dimensions can form physical state constraints. This effectively shields isolated noise disturbances at the algorithm level, ensuring the authenticity of the stacking state determination. Furthermore, it breaks the traditional disconnect between sensor detection links and execution pneumatic paths in control systems, improving the system's ability to remove redundant interference signals under complex conditions. Through unified dimensionality reduction decision-making of state and deviation quantities, it achieves smooth linkage of multiple execution devices, effectively solving the problem that existing feeding systems often treat stacking detection and attitude correction as independent processes, relying heavily on single-dimensional sensor data, making it difficult to resist interference signals in the dynamic environment of high-speed material feeding.

[0009] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of this application will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0010] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the various drawings denote the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings depict only some embodiments disclosed in this application and should not be construed as limiting the scope of this application.

[0011] Figure 1 This is a flowchart of a posture correction method according to an embodiment of this application.

[0012] Figure 2 This is a block diagram of an electronic device according to an embodiment of the present application.

[0013] Figure 3 A block diagram of the device architecture for implementing the posture correction method. Detailed Implementation

[0014] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of this application. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0015] Figure 1 A flowchart illustrating a posture correction method according to an embodiment of this application is shown. Figure 1 As shown, a posture correction method may include: S110: Acquire thickness detection data, gap detection data, aerodynamic condition data, and attitude image data during the process of the main propulsion equipment pushing the plate; S120: Perform feature fusion analysis on thickness detection data, gap detection data and aerodynamic condition data to determine the stacking state of the sheet metal. The stacking state includes single sheet metal state and multiple sheet metal state. S130: Perform feature extraction on the attitude image data to obtain the positional deviation value and angle deviation value of the plate; S140: Based on the stacking status, position deviation value, and angle deviation value, generate target control commands, and control the main propulsion equipment and auxiliary fine-tuning equipment to perform corresponding correction actions through the target control commands.

[0016] The posture correction method proposed in this embodiment acquires thickness detection data, gap detection data, aerodynamic condition data, and posture image data reflecting the material's dimensional dimensions during the main propulsion device's pushing of the sheet material. It then performs feature fusion analysis on the thickness detection data, gap detection data, and aerodynamic condition data, effectively filtering out local distortion interference from a single sensor in a dynamic pushing environment through cross-validation. This accurately analyzes and determines the current stacking state of the sheet material. Simultaneously, it extracts features from the posture image data to calculate the sheet material's positional and angular deviation values. Using the stacking state and spatial deviation values, it generates target control commands, providing the main propulsion device and auxiliary fine-tuning device with a global decision-making basis, enabling them to execute correction actions that strictly match the actual material conditions. The technical solution of this embodiment constructs a cross-validation mechanism between the external spatial dimensions of the material and the internal aerodynamic load of the actuator through thickness detection data, gap detection data, aerodynamic condition data, and posture image data. In dynamic conditions involving high-speed material feeding or vibration, when a sensor in one dimension generates interference signals due to local deformation or external factors, data from other dimensions can form physical state constraints. This effectively shields isolated noise disturbances at the algorithm level, ensuring the authenticity of the stacking state determination. Furthermore, it breaks the traditional disconnect between sensor detection links and execution pneumatic paths in control systems, improving the system's ability to remove redundant interference signals under complex conditions. Through unified dimensionality reduction decision-making of state and deviation quantities, it achieves smooth linkage of multiple execution devices, effectively solving the problem that existing feeding systems often treat stacking detection and attitude correction as independent processes, relying heavily on single-dimensional sensor data, making it difficult to resist interference signals in the dynamic environment of high-speed material feeding.

[0017] like Figure 3 As shown in the embodiments of this application, the apparatus for implementing the above-described posture correction method is specifically as follows: In terms of hardware physical topology, it mainly consists of a sensing and acquisition module, a main control and decision-making module, a pneumatic regulation module, and an actuator module working together. The signal output terminal of the sensing and acquisition module is connected to the input bus of the main control and decision-making module, the control signal output terminal of the main control and decision-making module is connected to the electrical receiving terminal of the pneumatic regulation module, and the fluid output terminal of the pneumatic regulation module is connected to the drive chamber of the actuator module through a pneumatic pipeline, thus forming a complete electro-pneumatic conversion and mechanical action closed loop.

[0018] Specifically, the sensing and acquisition module includes a laser displacement sensor, an ultrasonic sensor, a vision camera, and a cylinder pressure sensor. In terms of spatial installation, the laser displacement sensor and vision camera are fixed at a preset angle directly above the material conveying channel, their detection range covering the upper surface of the material. The ultrasonic sensor is embedded in the side wall of the material conveying channel, its probe aimed at the gap area on the side of the material. The cylinder pressure sensor is directly connected to the rodless chamber air supply line of the main pusher cylinder via a three-way connector. In terms of operation, the above four types of sensors maintain synchronous sampling during the material's propulsion stroke, converting the collected physical quantities into digital electrical signals and uploading them to the main control decision module in real time.

[0019] The main control decision module employs an industrial main control unit (such as a programmable logic controller or industrial control computer) with built-in data processing logic. Its operation involves receiving underlying thickness, gap, aerodynamic, and visual image data collected by the sensing and acquisition module, and internally executing feature fusion and deviation calculation algorithms. Based on the calculated stacking state and spatial deviation values, the industrial main control unit generates corresponding target control commands according to a preset control strategy and sends digital or analog commands to the lower-level aerodynamic adjustment module via a high-speed output port.

[0020] The pneumatic control module mainly consists of an electronically controlled proportional control valve and a proportional throttle valve assembly. Its electrical control terminal is connected to the output port of the industrial main control unit to receive target control commands; its air inlet connects to the factory's main air source, and its air outlet is connected via pressure-resistant hoses to the air inlet and outlet ports of each cylinder in the actuator module. In operation, the pneumatic control module acts as an electro-pneumatic conversion interface, linearly adjusting the valve core opening according to voltage or current changes in the main control commands, thereby achieving precise dynamic regulation of gas pressure and flow in the output air path.

[0021] The actuator module includes a main push plate cylinder, a station fine-tuning and correction cylinder, and a basic magazine discharge mechanism. The piston rod of the main push plate cylinder is fixedly connected to the push plate and arranged along the material conveying direction; the station fine-tuning and correction cylinder is installed perpendicular to the material conveying direction on the side of the channel. In the coordinated operation of the mechanical and pneumatic systems, under normal operating conditions, the main push plate cylinder receives rated air pressure to push the plate forward; when the system determines that correction is needed, the pneumatic adjustment module reduces the air supply pressure of the main push plate cylinder according to instructions, maintaining a low-friction sliding thrust; simultaneously, the side station fine-tuning and correction cylinder receives correction air pressure allocated by the pneumatic adjustment module, driving its piston rod to extend and laterally abut against the edge of the plate, changing the planar orientation of the plate through lateral thrust. Under the timing allocation of the main control unit, the two systems form an orthogonal force cooperation, completing the dynamic linkage of pushing and correction.

[0022] In step S110, thickness detection data, gap detection data, aerodynamic condition data, and attitude image data are acquired during the process of the main propulsion device pushing the plate.

[0023] In the actual operation of automated production lines, due to the physical deformation of the sheet material itself (such as edge warping and surface bulging) and environmental interference, the underlying data directly collected by sensors often contains a large amount of noise signals. If unprocessed noise is directly input into subsequent logic decision modules, it can easily lead to false alarms or control failures in the system.

[0024] In this embodiment, during the forward sliding and pushing of the sheet metal by the main propulsion device (such as the main push plate cylinder), a laser displacement sensor (LDS) installed at a preset angle above the material channel collects raw thickness detection data reflecting the height change of the sheet metal's upper surface, based on the high-frequency triangular reflection ranging principle. A high-frequency ultrasonic probe installed on the side wall of the material channel collects raw gap detection data reflecting the lateral microscopic state of the sheet metal. This data is mainly used to characterize potential gaps between stacked sheets. A pressure transmitter directly connected to the exhaust port of the rodless chamber of the main propulsion device collects real-time air pressure fluctuation values ​​inside the pipeline, serving as aerodynamic condition data reflecting the actual propulsion resistance load. A vision camera positioned directly above the material channel continuously captures images of the moving sheet metal at a set frame rate, obtaining attitude image data including the sheet metal's two-dimensional planar contour features. Simultaneously, a linear encoder or rope displacement sensor installed on the side of the main propulsion device's guide rail acquires real-time physical displacement data of the main propulsion device during its pushing stroke. All these sensors are controlled by the same system clock pulse, ensuring strict alignment of the collected data in the time dimension.

[0025] In actual material feeding operations, the cylinder's propulsion speed exhibits dynamic fluctuations due to the nonlinear changes in frictional resistance. Relying solely on time series analysis of the original thickness and gap detection data can lead to distortions caused by compression or stretching in spatial positioning. To address this issue, the system utilizes real-time physical displacement data to map the time-series-based original thickness and gap detection data to a pre-defined physical surface coordinate system. Using the designed starting point of the material's propulsion as the origin, the real-time physical displacement data serves as the longitudinal position coordinate (i.e., the direction of material feeding), while the lateral offset of the sensor's fixed installation is used as the lateral position coordinate. The original thickness and gap values ​​corresponding to each sampling time point are filled into the spatial grid corresponding to that coordinate, thereby generating a topological map of the surface thickness and gap distribution of the currently being pushed material. This topological map successfully converts one-dimensional time-flow data into a two-dimensional spatial numerical matrix that strictly corresponds to the actual physical dimensions.

[0026] Because industrial sheet metal is prone to localized mechanical bending or bulging during handling or resting, these localized physical deformations, when scanned by sensors, exhibit characteristics similar to the thickness increase or abrupt gap changes of two sheets, thus forming pseudo-double-sheet interference signals. The system detects numerically aberrant abrupt regions in the surface thickness spatial distribution topology map or the surface gap spatial distribution topology map. Specifically, the system uses a pre-defined threshold gradient comparison algorithm. When the numerical difference between adjacent spatial grids exceeds the normal tolerance zone, the region is marked as a numerically aberrant abrupt region, and the continuous connected area of ​​this region in the topology map is calculated. A local warping area threshold is pre-set, typically set to a value much smaller than the standard area of ​​the entire sheet metal. When the continuous area of ​​the numerically aberrant abrupt region is less than or equal to the pre-defined local warping area threshold, the system objectively determines, based on physical common sense, that the anomaly is not caused by the overlap of two complete sheets, but rather by localized physical deformation of the sheet metal itself (i.e., localized bulging or corner warping of a single sheet). The data portions corresponding to the numerically aberrant abrupt regions in the original thickness detection data and original gap detection data are then removed. Finally, the remaining data after cleaning and removal are output as the final thickness detection data and gap detection data.

[0027] The method in this embodiment introduces physical displacement data for spatial topological mapping and combines it with area thresholds for local deformation identification. The system effectively separates genuine double-sheet stacks from local deformation (pseudo-double sheets) of single sheets from the bottom-level data acquisition stage. The thickness detection data and gap detection data output after the above cleaning process, along with the synchronously acquired aerodynamic condition data and attitude image data, are all baseline data that have eliminated local condition interference.

[0028] In step S120, the thickness detection data, gap detection data, and aerodynamic condition data are subjected to feature fusion analysis to determine the stacking state of the sheet metal. The stacking state includes single sheet metal state and multiple sheet metal state.

[0029] In traditional feeding equipment, the determination of stacked materials (i.e., multiple sheets of material being pushed out simultaneously due to sticking together due to oil, static electricity, etc.) usually relies on a single thickness sensor. However, in the dynamic vibration environment of high-speed feeding, the signal from a single sensor is easily distorted; at the same time, if only pneumatic resistance is relied upon for judgment, it is easy to misjudge the jamming of the mechanical guide rail as the sticking of the sheet material.

[0030] In this embodiment, the single pushing stroke of the pushing cylinder is divided into several time windows, and the currently operating time window is defined as the current sampling period. The system extracts the thickness fluctuation variance of the thickness detection data after cleaning within the current sampling period, and simultaneously extracts the gap fluctuation variance of the gap detection data after cleaning within the current sampling period. The fluctuation variance characterizes the dispersion and stability of the sensor-acquired signal within the current time window. If there are dense micro-rough textures on the surface of the board or if the sensor mounting bracket experiences high-frequency resonance, the data variance within this period will increase.

[0031] The more severe the signal fluctuations, the lower the confidence level of its representation of the current physical reality. Therefore, the system calculates the real-time confidence weights for thickness data and gap data based on the thickness fluctuation variance and gap fluctuation variance, respectively. Mathematically, the specific value of the fluctuation variance is inversely proportional to the corresponding real-time confidence weight (i.e., the larger the variance, the smaller the assigned weight coefficient), thereby automatically attenuating the influence of single sensor data in a high-frequency oscillation state on the overall judgment at the algorithm level.

[0032] During the entire physical process of the sheet metal being pushed by the main propulsion equipment, the force state of the sheet metal is not constant. At the moment of startup, the cylinder needs to overcome the maximum static friction of the equipment and the sheet metal, at which time a sharp high-pressure peak will be generated in the pneumatic circuit; as the sheet metal begins to slide, the friction force transforms into a relatively small dynamic friction force, and the air pressure drops and tends to stabilize. If the high starting pressure caused by static friction is included in the stacking judgment, the system will generate false alarms.

[0033] To address this technical issue, in this embodiment, the system extracts the continuous pressure change curve from the pneumatic operating data. Using a slope abrupt change detection algorithm, it identifies and removes the initial peak data in the startup phase of the continuous pressure change curve. After stripping away this transient high-pressure characteristic, the system retains the smooth pressure data during the steady-state feeding phase (i.e., the uniform or stable acceleration sliding phase of the sheet material). Subsequently, the system determines the load reliability weight of the pneumatic operating data based on the slope of the smooth pressure data change. Specifically, if the slope of the pressure data approaches zero during the steady-state feeding phase (i.e., the pressure curve is stable), it indicates that the underlying mechanical guide rail is operating smoothly, and the increase in pneumatic resistance accurately reflects the increase in the material's own weight (i.e., material stacking occurs). In this case, a higher load reliability weight is assigned. Conversely, if the slope exhibits violent and irregular oscillations, it indicates interference or poor lubrication of the mechanical guide rail itself. In this case, the air pressure fluctuation cannot accurately reflect the material's state, and the system will proportionally reduce the load reliability weight for this dimension.

[0034] Based on the calculated real-time reliability weights (including thickness and gap weights) and load reliability weights, the system dynamically weights and sums the thickness detection data, gap detection data, and smoothing pressure data after cleaning. This weighted summation process integrates the thickness and gap, which characterize the external spatial dimensions of the material, with the smoothing pressure, which characterizes the internal resistance characteristics of the material, into a normalized comprehensive feature evaluation value. Subsequently, the system compares this weighted summation result with a preset double-sheet judgment benchmark value. This double-sheet judgment benchmark value is calibrated based on the historical multidimensional comprehensive values ​​of a standard single sheet under normal working conditions. If the weighted summation result is greater than the preset double-sheet judgment benchmark value, the system determines that the current comprehensive physical quantity exceeds the reasonable tolerance range of a single sheet, and thus determines the stacking state of the sheets as a multi-sheet state. Conversely, if the weighted summation result is less than or equal to the double-sheet judgment benchmark value, the system determines the stacking state of the sheets as a single sheet state.

[0035] When the optical sensor experiences data drift due to reflections from the sheet surface, the system automatically reduces its weight, relying more on aerodynamic drag and ultrasonic gap data for judgment. When the pneumatic system generates transient pulses due to overcoming static friction, the system can accurately remove these false features. This feature fusion analysis of multi-source heterogeneous data overcomes the drawbacks of isolated sensor operation in existing technologies, converging multi-dimensional sensing signals under complex working conditions into a clear and highly accurate Boolean variable of the stacked material state (single or multiple sheets).

[0036] In step S130, feature extraction is performed on the attitude image data to obtain the positional deviation value and angle deviation value of the plate.

[0037] After determining the stacking status (thickness and resistance anomalies) of the sheet material in the vertical Z-axis dimension, the system simultaneously performs two-dimensional geometric quantitative analysis on the spatial pose of the sheet material in the XY horizontal conveying plane. During the high-speed pushing process of the cylinder-driven pusher, due to the uneven distribution of dynamic friction on both sides of the mechanical guide rail or the slight tilt of the initial placement of the sheet material, a single sheet material is very prone to positional slippage and angular rotation along the conveying plane.

[0038] In this embodiment, the two-dimensional pixel array acquired by the sensing and acquisition module at the front end is converted into mechanical geometric error parameters that can be referenced by the pneumatic actuator. Specifically: Due to variations in lighting in industrial settings or the presence of oil stains reflecting off the material surface, raw images often contain redundant background information. The system first performs grayscale conversion and threshold segmentation on the posture image data, binarizing and separating the material from the background of the conveyor channel. Then, it uses an edge detection operator to identify sets of boundary points with drastic changes in pixel grayscale gradients, thereby extracting the edge contour lines that enclose the true physical boundaries of the material from the posture image data. These edge contour lines completely replicate the top-view geometry of the material being pushed in a two-dimensional pixel coordinate system.

[0039] The system extracts the parallel long side contours of the two sides based on the edge contour lines. By calculating the set of midpoints of the lines connecting the corresponding pixels of these two long sides, a straight line is fitted using the least squares method to fit a target center line that runs through the length of the board. This target center line physically represents the actual current travel axis of the board.

[0040] The system calculates the centroid of the closed polygon enclosed by the edge contour lines (i.e., the arithmetic mean of the coordinates of all contour pixels) and identifies this centroid coordinate as the geometric center point of the material. This geometric center point represents the current absolute dwell position of the material in the transport plane.

[0041] To convert the actual attitude parameters into error values ​​that can be used for aerodynamic correction, a preset baseline is pre-calibrated in the global coordinate system. This preset baseline physically coincides with the theoretical ideal pushing center axis of the main pusher cylinder, representing the calibrated trajectory of the material sheet moving forward without any friction interference.

[0042] To address translational deviation, the system calculates the vertical Euclidean distance from the geometric center point to the preset baseline, and converts this distance span and offset direction (leftward or rightward) into a positional deviation value. This value quantifies the severity of lateral slippage of the material from the central conveyor track.

[0043] To address rotational deviation, the system calculates the angle formed by the intersection of the target centerline and the preset baseline in a two-dimensional plane. The magnitude of this angle and the direction of rotation (clockwise or counterclockwise) are then defined as the angular deviation value. This value quantifies the degree of yaw rotation of the material around its geometric center.

[0044] In this embodiment, by reconstructing geometric features based on physical contours, the complex two-dimensional spatial composite motion error is successfully reduced in dimensionality and decoupled mathematically into two independent vectors: a positional deviation value representing lateral translation error and an angular deviation value representing spin error. This feature decoupling extraction method avoids the limitations of previous correction systems that relied solely on edge photoelectric switches for coarse boundary crossing alarms.

[0045] In step S140, a target control command is generated based on the stacking state, position deviation value, and angle deviation value, and the main propulsion device and auxiliary fine-tuning device are controlled to perform corresponding correction actions through the target control command.

[0046] The system obtains the stacking state of the plate material in the Z-axis direction based on multi-source fusion analysis, and calculates the positional deviation and angular deviation values ​​of the plate material in the XY conveying plane through visual feature contour reconstruction.

[0047] In this embodiment, the multi-dimensional judgment results at the software level are converted into the connection point of mechanical and physical work at the hardware level. The system first performs a first-priority judgment on the stacked material state at the main control logic layer, which determines the physical flow of subsequent materials. If, after cross-validation, it is determined that the stacked material state consists of multiple sheets, the system, based on industrial equipment protection, determines that the current sheet material has two or more sheets overlapping and adhering, exceeding normal processing properties. If planar position correction is performed and the material continues to be pushed to the downstream station, the thickness of the overlapping sheet material will directly cause hard cracking of the subsequent stamping or bonding molds, and may even cause the core transmission mechanism of the equipment to jam and be scrapped.

[0048] Therefore, under this abnormal operating condition, the system disables the verification logic for planar position and angle, and issues the highest priority emergency stop and waste discharge command as the target control command. This command acts on the pneumatic adjustment circuit that controls the main propulsion equipment (such as the main pusher cylinder). The specific physical control actions include: quickly shutting off the air supply proportional valve on the air inlet side (rodless chamber) of the main propulsion equipment, while fully opening the exhaust side to release pressure; and immediately introducing high-pressure gas into the return side (rod chamber), forcibly driving the piston rod of the main pusher cylinder to return at high speed to the initial zero-position safety area. Simultaneously, the main control system triggers the external audible and visual alarms and the feed interlock signal of the downstream station through the input / output bus, putting the equipment in a safe interruption state awaiting waste discharge.

[0049] Only when the stacked state of the sheet material is determined to be a single sheet can the system determine that the current material is an acceptable carrier in the Z-axis thickness direction, and then unlock and enter the second-level planar position and angle verification process. During high-speed pushing, a single sheet may drift laterally or spin within the conveying plane due to local airflow interference or slight asymmetric sliding friction of the guide rail. To prevent excessive mechanical intervention on sheets that are within the acceptable tolerance zone but have slight fluctuations (i.e., over-control leading to unstable production line cycle), the system pre-stores mechanically calibrated tolerance extreme values, namely the position deviation threshold and the angle deviation threshold. The system compares the positional deviation value with a positional deviation threshold in real time, and simultaneously compares the angle deviation value with an angle deviation threshold. If both the positional deviation value and angle deviation value are less than or equal to their respective thresholds, the plate is considered to be in a centered position, and the system does not issue any additional control commands; the main propulsion device maintains its rated air pressure. Conversely, if either the positional deviation value or the angle deviation value exceeds its threshold, the system determines that the slippage or rotation of the plate threatens the gripping accuracy of the downstream docking station. In this case, the system immediately generates a flexible attitude correction command as the target control command to initiate the subsequent closed-loop correction execution process.

[0050] For working conditions requiring correction, avoiding damage to the material or causing it to jam in the guide rail during the pushing process is a bottleneck that traditional equipment cannot overcome. In this embodiment, the rigid control mode of single-axis forceful pushing in traditional equipment is avoided. The flexible attitude correction command at the software level is precisely decoupled and split into two underlying sub-commands at the hardware driver layer that are synchronized in time logic but orthogonal and cooperative in physical force space.

[0051] According to basic mechanics principles, when the sheet metal advances in the guide rail, the static and sliding friction forces that it needs to overcome laterally are proportional to the normal pressure and acceleration inertia force applied longitudinally to the sheet metal by the main propulsion equipment. If, while the main propulsion equipment is pushing the sheet metal at high speed and forcefully, the lateral auxiliary mechanism directly and forcibly pushes the edge of the sheet metal, the sheet metal is prone to inelastic deformation (such as edge compression and curling) under the pressure of rigid forces in two perpendicular directions, and may even tilt and jam within the channel.

[0052] To resolve this fundamental mechanical conflict, upon triggering the correction, the system first issues a main thrust air pressure reduction command to the proportional solenoid valve controlling the main propulsion equipment circuit. This command, through a linear downward adjustment of the air supply proportional valve opening, significantly reduces the air supply pressure to the rodless chamber of the main propulsion equipment within milliseconds. After the air pressure reduction, the main propulsion equipment switches from the original high-pressure acceleration and strong thrust mode to a low-pressure, low-resistance sliding mode. In this mode, the longitudinal thrust output by the main thrust cylinder is attenuated to a critical state that is only sufficient to maintain the plate at a uniform forward sliding speed or overcome the dynamic friction of the guide rail base, eliminating excessive compressive stress. By transforming the mechanical state of the plate within the conveyor into a state similar to suspension or low-resistance free movement, the resistance to lateral sliding is significantly reduced, clearing the mechanical obstacles for the intervention of the lateral correction mechanism.

[0053] Furthermore, while completing drag reduction and decoupling in the main pushing direction, the system issues lateral dynamic thrust distribution commands to the auxiliary fine-tuning equipment (i.e., fine-tuning cylinder groups distributed on the side of the conveying channel and arranged perpendicular to the pushing direction) based on the error type. In order to complete the correction efficiently and accurately, the main control unit allocates completely different pneumatic drive modes according to the specific error attributes.

[0054] When data comparison results show that the current error is mainly lateral translation (i.e., the positional deviation exceeds the standard, but the angular deviation is within the safe threshold), since the plate itself has not undergone spatial rotation, the lateral offset of each vertex is consistent. At this time, the system controls the proportional regulating valve to output a continuous pneumatic thrust with corresponding pressure to the auxiliary fine-tuning device according to the magnitude of the positional deviation. This continuous thrust is smoothly pressed against the side of the plate through the pusher of the fine-tuning cylinder, driving the plate, which is sliding in a low-resistance state, to move laterally as a whole, directly translating until the geometric center point is aligned with the preset baseline.

[0055] When data comparison results show that the current main deviation is rotational deviation (i.e., the angle deviation value significantly exceeds the standard), it indicates that the plate has yawed and rotated around its own center. At this time, if a continuous static thrust is used, it is easy to push the entire plate off the track. Based on the principle of physical torque, the system controls the auxiliary fine-tuning device to switch the drive mode and output a high-frequency pulsed aerodynamic thrust to the lever arm end of the side of the plate that has deflected (usually the front or rear corner edge of the plate). Through the striking pulse momentum, a discontinuous eccentric force is instantaneously applied to the edge of the plate. This force uses the fitted geometric center point of the plate as the physical rotation fulcrum, generating a strong and controllable reverse restoring torque (rotational axis torque). Each pulse adjustment drives the plate to undergo a slight angle reversal, cleverly using momentum to overcome the plate's own rotational inertia. This can quickly correct the centerline angle of the plate and avoid excessive overshoot or reverse rotational oscillation that may be caused by continuous thrust.

[0056] This embodiment achieves intelligent classification and response by organically combining single- and multi-sheet state gating with bipolar deviation threshold gating, enabling immediate interception of anomalies, safe release of minor anomalies, and precise intervention for out-of-range anomalies. From an execution perspective, through a longitudinal active unloading and drag reduction and a lateral dynamic differentiated force application electro-pneumatic decoupling and collaborative mechanism, it cleverly utilizes continuous thrust to solve translational slippage problems and pulse torque to solve angular spin problems. The entire correction process is executed under dynamic conditions where the sheet material continues to slide forward without interrupting the high-speed feeding cycle of the production line, solving the industry challenges of mechanical jamming and non-destructive material protection during the process.

[0057] In some embodiments of this application, thickness detection data, gap detection data, and aerodynamic condition data are fused and analyzed to determine the stacking state of the sheet metal. The stacking state includes single sheet metal state and multiple sheet metal state, including: Based on the thickness measurement data, determine the thickness fluctuation variance within the current sampling period; Based on the gap detection data, determine the gap fluctuation variance within the current sampling period; Based on the thickness fluctuation variance and the gap fluctuation variance, the real-time reliability weights of the thickness data and the gap data are obtained. Among them, the fluctuation variance is inversely proportional to the corresponding real-time reliability weight. Based on aerodynamic operating data, determine the continuous pressure variation curve; The initial peak data in the start-up stage of the continuous pressure change curve is removed to obtain smooth pressure data in the steady-state feeding stage. The load confidence weight of the aerodynamic operating condition data is determined based on the slope of the change in the smoothed pressure data. Based on real-time reliability weight and load reliability weight, the thickness detection data, gap detection data and smoothing pressure data are dynamically weighted and summed to obtain the weighted summation result; If the weighted summation result is greater than the preset double-sheet judgment benchmark value, the stacked state of the board is determined to be a multi-sheet state.

[0058] During high-speed, dynamic feeding processes accompanied by mechanical vibrations, any single-dimensional sensor is susceptible to transient interference from the external physical environment (such as oil stains and guide rail resonance), causing the output value to drift.

[0059] In this embodiment, during the continuous pushing of the sheet material, the system divides the time axis into continuous and equally long sampling windows. For the time window where the logical operation is currently being performed (i.e., the current sampling period), the system extracts all sampling point data within this period for statistical analysis of time-domain characteristics. Specifically, based on the thickness detection data, the system uses a variance statistical algorithm to determine the thickness fluctuation variance within the current sampling period; simultaneously, based on the gap detection data, the system determines the gap fluctuation variance within the current sampling period.

[0060] In this operational logic, the fluctuation variance characterizes the dispersion of the signal collected by a single-size sensor within a specific time period. For example, when severe local oil contamination on the surface of the board causes high-frequency signal drift in the optical thickness sensor, or when high-frequency resonance of the mechanical guide rail causes jitter in the echo signal of the ultrasonic ranging sensor, the calculated thickness fluctuation variance or gap fluctuation variance will increase significantly. Based on the objective mapping relationship of the above physical phenomena, the system, based on the calculated thickness fluctuation variance and gap fluctuation variance, obtains the real-time reliability weights for the thickness data and gap data respectively through the underlying weight allocation logic. In the specific mathematical mapping rules, the system sets the fluctuation variance to be inversely proportional to the corresponding real-time reliability weight. That is, the more drastic the fluctuation of a certain dimension's size detection data within the current sampling period (the larger the variance value), the more severe the interference from the external environment, and the lower the confidence level of the data in reflecting the true thickness state of the board. The system then automatically attenuates and assigns it a smaller real-time reliability weight; conversely, if the data's time-domain characteristics are stable (the variance value is small), a larger real-time reliability weight is assigned. This weighting mechanism enables adaptive and interference-resistant identification of multi-source size data.

[0061] The sliding of the sheet metal within the conveying channel is not always at a constant speed. According to basic physical and mechanical principles, at the instant the main propulsion device contacts the sheet metal and begins pushing, the cylinder needs to output a huge transient thrust to overcome the maximum static friction between the sheet metal and the mechanical guide rail. This physical process causes the pressure in the pneumatic circuit to rise sharply, forming a pressure peak. However, once the sheet metal leaves the stationary state and begins to slide, the friction force transforms into a relatively small dynamic friction force, and the air pressure subsequently drops and tends to level off.

[0062] To avoid misinterpreting normal pressure surges caused by static friction as excessive material weight or adhesion (stacking), the system extracts and identifies a continuous pressure change curve from the collected pneumatic operating data over time. Subsequently, the system uses a peak feature recognition algorithm to strictly identify and remove the initial peak data in the start-up phase of the continuous pressure change curve. After stripping away this transient pulse high-pressure characteristic, the system extracts and retains smooth pressure data from the steady-state pushing phase (i.e., the uniform or stable acceleration sliding phase of the material). After obtaining the pure smooth pressure data, the system further calculates the derivative of the curve sequence, i.e., determines the load reliability weight of the pneumatic operating data based on the slope of the smooth pressure data change.

[0063] During the steady-state feeding stage, if the slope of the pressure data changes close to zero (i.e., the pressure remains within a stable range), it indicates that the underlying mechanical guide rail is operating smoothly. At this time, the magnitude of the pneumatic resistance can accurately reflect the gravity load of the material itself, so the system assigns it a high load reliability weight. Conversely, if the slope of the change shows violent and irregular oscillations, it indicates that the change in pneumatic resistance is caused by a lack of lubrication of the guide rail or mechanical jamming. At this time, the air pressure data cannot accurately represent the weight state of the material, and the system proportionally reduces its load reliability weight.

[0064] After obtaining the real-time reliability weights for the size dimension and the load reliability weights for the dynamic dimension, the system dynamically weights and sums the thickness detection data, gap detection data, and smoothing pressure data based on the real-time reliability weights and load reliability weights, and finally obtains the weighted summation result.

[0065] The mathematical relation is: Where S represents the weighted summation result obtained by the system through calculation, which is a dimensionless comprehensive evaluation value that characterizes the current composite state of the board. This represents the thickness detection data acquired by the system; The real-time reliability weight of the thickness data dynamically calculated by the system based on the thickness fluctuation variance; This represents the gap detection data acquired by the system; The real-time reliability weight of the gap data is represented by the system dynamically calculated based on the gap fluctuation variance. This represents the smoothed pressure data obtained after the system removes the initial peak data. The load reliability weight represents the aerodynamic condition data dynamically calculated by the system based on the slope of the change in smooth pressure data.

[0066] After completing the aforementioned dynamic weighted summation, the system logically compares the obtained weighted summation result with a preset dual-sheet judgment benchmark value within the main control unit. This preset dual-sheet judgment benchmark value is a comprehensive feature upper limit boundary reference value obtained by normalizing and calibrating a large amount of historical working condition prior data under normal equipment operation conditions, based on a standard single sheet of board. If the comparison result shows that the weighted summation result is greater than the preset dual-sheet judgment benchmark value, the comprehensive physical feature quantity of the current multi-dimensional sensors after confidence dynamic filtering and weighted fusion has exceeded the reasonable tolerance boundary allowed for a single sheet of board. At this time, the system triggers the final judgment condition, determining that the stacked state of the currently pushed boards is a multi-sheet state.

[0067] This application overcomes the fundamental flaw in traditional automated equipment where single-sensor static threshold alarms are prone to misjudgment due to mechanical vibration or light interference. By introducing fluctuation variance to assess the purity of spatial dimensional data and introducing de-starting peak and slope judgments to assess the authenticity of dynamic load data, a cross-validation adaptive data cleaning and weighting mechanism is constructed. When faced with complex dynamic material feeding conditions, the system can automatically filter out abnormal signals in the distorted dimension and strengthen the effective signals in the steady-state dimension. Without increasing the cost of additional high-precision measurement hardware, the algorithmic logic can converge multi-source heterogeneous signals into a highly reliable physical state judgment result, thereby ensuring the accuracy of identifying multiple overlapping sheets.

[0068] In some embodiments of this application, feature extraction of the attitude image data to obtain the positional deviation value and angle deviation value of the plate includes: In the attitude image data, a first projection axis and a second projection axis that are perpendicular to the direction of travel of the plate are determined. The pixel gradient distribution of the attitude image data on the first projection axis and the second projection axis is calculated to obtain the gradient peak region. The gradient peak region is extracted to obtain discrete edge pixel coordinates. Based on the coordinates of all edge pixels, determine the residual distances from all edge pixel coordinates to the fitted center line; Edge pixel coordinates with a distance residual greater than a preset noise threshold are removed to obtain an effective set of edge feature points; The actual outline line of the board is obtained based on the effective edge feature point set; Based on the geometric relationship between the actual contour line and the preset reference coordinate system, the positional deviation and angular deviation values ​​are obtained. In the event that feature extraction from attitude image data fails, extract multiple historical position deviation values ​​and historical angle deviation values ​​that the main propulsion device has successfully calculated consecutively within a preset historical time window. Based on multiple historical positional deviation values ​​and multiple historical angle deviation values, a trend evolution trajectory is generated; Based on the trend evolution trajectory, extrapolate and calculate the estimated position deviation and estimated angle deviation values ​​for the current working cycle. Based on the estimated position deviation and estimated angle deviation values, the position deviation and angle deviation values ​​are determined to trigger the generation of specific degraded control commands to limit the maximum permissible propulsion speed of the main propulsion equipment and to coordinate the extension of the physical guide limit device.

[0069] In this embodiment, after acquiring the posture image data, the system establishes a first projection axis (typically corresponding to the horizontal coordinate axis of the image plane) and a second projection axis (typically corresponding to the vertical coordinate axis of the image plane) perpendicular to the physical travel direction of the board in the two-dimensional coordinate system of the image. Subsequently, the system calculates the pixel gradient distribution of the posture image data along the first and second projection axes. In the underlying physical mapping of machine vision, there is a significant difference in optical reflectivity between the board body and the background of the conveyor rail below it. This physical difference manifests in the image data as a sharp abrupt change in grayscale values ​​between adjacent pixels. The system calculates the grayscale change rate across the entire image region using a difference algorithm, and identifies the set of pixels where the change rate reaches a local extremum (i.e., the most drastic grayscale jump) as the gradient peak region. Furthermore, the system performs geometric center extraction and binarization processing on the aforementioned gradient peak regions, stripping away redundant background pixels to obtain a series of discrete edge pixel coordinate points. Ideally, these discrete points should conform to the physical edges of the board.

[0070] However, in real industrial processing environments, the edges of sheet metal inevitably contain tiny burrs left from mechanical stamping, or their surfaces may have oil stains reflecting light, as well as localized shadows caused by ambient light. These physical disturbances can cause some of the extracted edge pixel coordinates to deviate from the actual physical boundaries of the sheet metal. If these noisy points are used directly for attitude calculation, the resulting correction action will suffer from severe overshoot or erroneous movements.

[0071] To eliminate the aforementioned interference, the system introduces an adaptive filtering mechanism based on residual analysis. First, the system uses the least squares method or random sample consensus algorithm to perform initial straight-line fitting on the extracted edge pixel coordinates, constructing a fitting center line as a mathematical benchmark. Then, based on all edge pixel coordinates, the system iterates through and calculates the distance residuals from all edge pixel coordinates to the fitting center line. Geometrically, this distance residual represents the perpendicular normal distance from each discrete pixel to the fitting center line.

[0072] The system logically compares all calculated distance residuals with a pre-set noise threshold. This pre-set noise threshold is a safety boundary value derived from physical calibration based on the maximum allowable burr size in sheet metal processing and camera resolution. The system performs forced removal of edge pixel coordinates with distance residuals greater than the pre-set noise threshold, identifying them as false edges or environmental noise interference. After this high-throughput filtering, the set of points retained by the system is the high-purity, high-confidence effective edge feature point set. Based on the effective edge feature point set after removing interference noise, the system performs secondary analytical geometric fitting to obtain a true contour line that reflects the physical boundary orientation of the sheet metal.

[0073] Finally, during a normal vision cycle, the system performs a spatial vector comparison based on the geometric relationship between the actual contour line and the preset reference coordinate system (i.e., the standard physical center line of the equipment guide rail). By calculating the difference in the vertical intercepts of the actual contour line and the reference coordinate system in the translation direction, the positional deviation value representing the lateral slip is obtained; by calculating the vector angle between the actual contour line and the reference coordinate system, the angular deviation value representing the rotational yaw is obtained.

[0074] This embodiment not only focuses on high-precision calculation under normal conditions, but also on constructing a low-level physical fault-tolerant mechanism to address the vulnerability of optical systems. In industrial settings, sudden changes in ambient lighting caused by welding arc light or external obstruction caused by falling debris can easily occur, which can directly lead to the system's failure to extract features from the attitude image data (i.e., the inability to extract a sufficient set of effective edge feature points).

[0075] In the event of an abnormal situation where feature extraction fails as described above, to prevent the system from going out of control and crashing due to loss of visual feedback, or from causing production line stagnation by taking a simple and crude emergency shutdown, the system automatically triggers a predictive degradation branch process: First, the system accesses the internal cache and extracts multiple historical position deviation values ​​and historical angle deviation values ​​that the main propulsion device has successfully calculated consecutively within a preset historical time window (e.g., the time period 0.5 seconds before visual failure).

[0076] Based on the principle of kinematic inertia, the system assumes that the sliding and rotational postures of a plate in a high-speed sliding state have physical continuity over a short period of time. Therefore, the system generates the trend evolution trajectory of the plate in the conveying plane using time series analysis algorithms (such as Kalman filtering or linear regression) based on multiple historical position deviation values ​​and multiple historical angle deviation values.

[0077] Based on the constructed trend evolution trajectory, the system performs forward mathematical extrapolation according to the current time step of visual gaps to extrapolate and calculate the estimated positional deviation and estimated angle deviation values ​​for the current working cycle. At this point, the system uses the estimated positional deviation and estimated angle deviation values ​​to replace the originally missing real-time calculation data as the final output positional deviation and angle deviation values. The system is aware that the current deviation values ​​are estimates based on historical data, not absolutely precise real-time measurements. As the extrapolation time increases, the estimation error will inevitably accumulate and amplify. To prevent excessive aerodynamic correction actions based on the estimated values ​​from causing the material to fly off the guide rail, the system triggers the generation of a specific degradation control command simultaneously with the output of the estimated deviation values. The degradation control command is a composite command that links software and hardware across domains: at the software and pneumatic control level, the command forcibly limits the maximum permissible propulsion speed of the main propulsion device (for example, by controlling the proportional valve to reduce the main propulsion air pressure, reducing the kinetic energy of the plate movement, and reserving physical buffer time to deal with potential collisions); at the mechanical execution level, the command triggers the physical guide and limit devices (such as rigid nylon rollers driven by limit cylinders) distributed on both sides of the guide rail to extend instantaneously, shrinking the originally open flexible correction channel into a rigid limit channel with physical forced constraints.

[0078] This embodiment employs a dual-layer visual algorithm—image gradient coarse extraction and residual culling fine fitting—to avoid interference from burrs and oil contaminants in industrial environments, ensuring the purity of position and angle deviation values. Furthermore, through a historical trend extrapolation prediction mechanism following visual failure, the system can maintain continuous production line operation even under extreme conditions of temporary blindness. Simultaneously, the system uses degraded control commands to extend physical guide and limit devices, seamlessly switching from high-precision flexible pneumatic correction to low-speed rigid physical limit collision avoidance mode. This fundamental mechanical structure compensates for the error gap in the estimation algorithm, achieving a balance between efficiency and safety. This broadens the adaptability of automated equipment to various operating conditions, no longer limited by ideal optical inspection environments.

[0079] In some embodiments of this application, based on the stacking state, position deviation value, and angle deviation value, a target control command is generated, and the main propulsion device and auxiliary fine-tuning device are controlled to perform corresponding correction actions through the target control command, including: When the stacked material is in the state of a single sheet, and the positional deviation and angle deviation values ​​are within the preset fine-tuning correction range, a flexible correction command is generated. Based on the flexible correction command, the main propulsion equipment is controlled to reduce the air supply pressure to the preset maintenance pressure threshold so that the plate is pushed in a low-friction state. When the plate is in a low-friction pushing state, the control auxiliary fine-tuning device applies a lateral thrust to the side of the plate according to the position deviation value and angle deviation value until the plate moves to the target posture position. After detecting that the plate has reached the target posture position, control the main propulsion equipment to restore normal pushing air supply pressure; When the stacked state consists of multiple sheets of material, an emergency waste discharge command is generated. Based on the emergency exhaust command, the main air intake channel of the main propulsion equipment is controlled to exhaust air urgently, and the reverse return air passage is controlled to open to the maximum degree of ventilation so that the main propulsion equipment can quickly return to the initial position. When the main propulsion equipment retracts, the pneumatic flap located at the bottom of the material channel is opened, allowing multiple overlapping sheets to fall into the waste collection area. After the waste is discharged, the locking system is reset. Acquire pneumatic operating condition data for multiple consecutive working cycles within a preset time window, and acquire smoothed pressure data corresponding to each working cycle; The slope of the mean change trend of all smoothed pressure data is obtained by calculating the time sliding window. If the slope of the mean change trend indicates that the overall pressure is slowly increasing and the increase does not exceed the preset mechanical limit threshold, it is determined that the main propulsion equipment is in a state of natural wear and aging. Based on the slope of the mean change trend, the pneumatic pressure judgment baseline for double sheet determination in subsequent work cycles is updated proportionally with positive compensation.

[0080] In this embodiment, when the system determines that the current stacking state is a normal single sheet state, and the detected position deviation value and angle deviation value are both within the preset fine-tuning correction range, the system determines that the current sheet has a slight tilt, but does not constitute a serious risk of jamming, and then triggers the generation of a flexible correction command.

[0081] In traditional rigid alignment, the auxiliary cylinder often applies lateral force force while the main propulsion cylinder is fully loaded, which can easily lead to edge deformation or surface scratches on the sheet metal under strong bidirectional extrusion. To solve this problem, this embodiment, based on the flexible alignment command, first controls the main propulsion device to actively reduce its air supply pressure to a preset maintaining pressure threshold. The thrust provided by this maintaining pressure threshold is just enough to overcome the dynamic friction of the system foundation to maintain the low-speed inertial sliding of the sheet metal, but not enough to generate additional rigid compressive stress.

[0082] The equation for balancing sliding friction is: ;in, This indicates the output thrust of the main propulsion system; To maintain the pressure threshold; The effective force-bearing cross-sectional area of ​​the cylinder for the main propulsion equipment; This refers to the basic dynamic friction between the sheet metal and the conveyor rail. This is achieved by reducing the air supply pressure to... The system ensures that the output thrust is exactly equal to the system's basic dynamic friction force, thereby putting the plate in a low-friction state of being pushed at a constant speed or with slight deceleration.

[0083] When the sheet material is in a low-friction pushing state, the normal pressure between the sheet material and the side physical limiters approaches zero, allowing the system to maintain this state. In this state, the normal pressure between the sheet material and the bottom conveyor rail and side limiters is significantly reduced, minimizing sliding resistance. Subsequently, the system controls auxiliary fine-tuning devices (such as pneumatic push rods driven by high-frequency proportional solenoid valves distributed on the side of the channel) to apply precisely calculated lateral thrust to the sides of the sheet material based on positional and angular deviation values. Because the sheet material is in a low-friction pushing state, this lateral thrust can propel the sheet material to lateral translation and yaw rotation with extremely low resistance until the sheet material moves in the plane and aligns with the ideal target posture position. After detecting that the sheet material has reached the target posture position, the system confirms that the posture correction is complete and then controls the main propulsion equipment to restore normal pushing air pressure, allowing the sheet material to regain rated acceleration and enter the next processing stage. This flexible pressure reduction fine-tuning mechanism avoids physical damage during the correction process.

[0084] If the system determines that the stacked material is in a state of multiple sheets (i.e., severe adhesion and stacking), continuing to push the material will lead to expensive mold damage or serious mechanical jamming. Therefore, the system immediately generates a high-priority emergency waste discharge command. Based on this emergency waste discharge command, the system executes a transient reversal action in the pneumatic control loop: on the one hand, it controls the main air intake channel of the main propulsion equipment to perform an emergency exhaust action, quickly releasing the pressure in the main propulsion chamber and cutting off the propulsion power to continue forward; on the other hand, it simultaneously controls its reverse return air path to open to the maximum opening. By creating a huge transient reverse pressure difference, the main propulsion equipment is forced to overcome inertia and quickly return to the initial position in a very short time, thereby freeing up physical space for subsequent waste discharge.

[0085] The aerodynamic model of this process can be expressed as: ; in, For recoil acceleration; This is the maximum inflation pressure for the reverse return air path; The residual exhaust pressure in the main intake passage; This represents the cross-sectional area of ​​the reverse air chamber; For the equivalent mass of the feeding mechanism, It is the inherent resistance of mechanical equipment (the force that pulls backward).

[0086] By generating a huge transient reverse pressure difference, the main propulsion equipment is forced to overcome mechanical inertia and quickly return to its initial position with maximum acceleration, thereby spatially detaching from contact with the abnormally overlapping plates.

[0087] Once it is confirmed that the main propulsion equipment has retracted and disengaged from the material, the system controls the pneumatic flap at the bottom of the material channel to open. Without bottom physical support, multiple overlapping sheets fall under gravity into the pre-designated waste collection area below. After the photoelectric sensor confirms that the waste discharge channel is empty and waste discharge is complete, the system controls the pneumatic flap to reset and relocks the system, quickly restoring the production line to standby mode.

[0088] This embodiment not only solves the problem of transient operating condition control, but also introduces an adaptive calibration mechanism for the physical evolution of the equipment throughout its entire life cycle. During long-term high-frequency operation of the equipment, the natural wear of the mechanical guide rails and the drying of the lubricating grease will cause the inherent basic frictional resistance of the system to gradually increase. In order to push a single sheet of the same weight, the steady-state air pressure required by the system will inevitably increase accordingly. If the system adopts a static and fixed judgment criterion, this long-term air pressure increase will be incorrectly accumulated in the load characteristics, ultimately causing the system to misclassify a normal single sheet of material as multiple sheets.

[0089] To overcome the misjudgment caused by this long-term evolution, the system acquires aerodynamic condition data for multiple consecutive work cycles within a preset time window (e.g., nearly 10,000 continuous work cycles) and extracts the smoothed pressure data corresponding to each work cycle. The system performs time-series regression calculations on all smoothed pressure data within the time window to obtain the slope of the mean change trend of all smoothed pressure data. This trend can be characterized by the following univariate linear regression formula: ; in, The slope of the mean change trend is the desired value; N is the total number of samples in the preset time window. This is the time series index for the i-th working period. The average value of the time index; Let be the mean of the smoothed pressure data corresponding to the i-th working cycle. This is the overall average of all smoothed pressure data within the entire time window. The slope of the mean change trend is... If the overall pressure shows a slow upward trend (i.e., the value is a continuous small positive value) and it is determined that the absolute increase has not exceeded the preset mechanical limit threshold (i.e., the physical damage threshold that requires the equipment to be shut down for major repair), the system determines based on the objective physical evolution law that the main propulsion equipment and the conveying channel are currently in a state of natural wear and aging.

[0090] To adapt to this physical change, the system calculates the frictional resistance increment based on the slope of the mean change trend, and proportionally updates the aerodynamic pressure judgment baseline of the double-sheet judgment benchmark in subsequent working cycles.

[0091] The compensation formula can be expressed as: ; in, For the updated baseline, As the initial or previous stage baseline, This is the compensation amount calculated based on the slope.

[0092] As mechanical wear causes the base propulsion air pressure to rise, the system automatically raises the pressure threshold baseline that triggers the condition determination of multiple sheet metal sheets. Through this underlying adaptive baseline compensation mechanism, the system eliminates the interference of long-term mechanical aging on thrust data, ensuring that the multi-source data fusion determination algorithm has high accuracy and anti-aging robustness throughout the entire life cycle of the equipment.

[0093] In some embodiments of this application, controlling the auxiliary fine-tuning device to apply a lateral thrust to the side of the plate according to the position deviation value and the angle deviation value until the plate moves to the target posture position includes: The rate of reduction of positional deviation of the plate during the application of lateral thrust is obtained; If the current position deviation is greater than the preset pulse switching threshold, the auxiliary fine-tuning device is controlled to start the continuous air supply mode in order to perform a continuous approaching action toward the target attitude position. If the current position deviation value is reduced to equal to or less than the pulse switching threshold, and the target attitude position has not yet been reached, the continuous air supply mode will be switched to the pulse jog air supply mode. In pulse-driven air supply mode, the transient residual air pressure in the auxiliary fine-tuning device is released by the air cut-off interval between pulses, and the plate is controlled to move in a step-by-step manner until it reaches the target posture position.

[0094] The system controls the main propulsion device to reduce pressure, putting the sheet metal in a low-friction, pushed state, and instructs the auxiliary fine-tuning device to begin applying lateral thrust to the side of the sheet metal. Due to the inherent hysteresis phenomenon caused by the compressibility of gas in the pneumatic actuator system, if a single constant thrust is used for correction, inertial overshoot can easily occur when the sheet metal approaches the target position, i.e., overshoot in the control system domain, resulting in repeated back-and-forth correction oscillations.

[0095] In this embodiment, during the initial and intermediate processes of applying lateral thrust by the control auxiliary fine-tuning device, the system obtains the current kinematic state of the plate in real time and at high frequency through the underlying sensor array (such as high-frequency visual sampling or laser displacement feedback). Specifically, this includes real-time calculation and acquisition of the reduction rate of position deviation of the plate during the application of lateral thrust.

[0096] The physical meaning of this positional deviation reduction rate characterizes the approach velocity of the plate's lateral translation. It can be obtained by differentiating the deviation difference between the current cycle and the previous cycle with respect to time. The calculation formula is as follows: ; in, The rate at which the positional deviation is reduced; This refers to the current position deviation value extracted by the system in real time. This refers to the tiny sampling period of the control system.

[0097] The system continuously monitors the current position deviation value in the control loop. If the current position deviation exceeds the preset pulse switching threshold (this threshold is typically dynamically set or preset to a fixed safety distance based on the material quality, the current position deviation reduction rate, and the inherent braking inertia of the mechanical system), the system determines that the material is still far from the final target, and the primary objective is to quickly eliminate large-scale errors. Under this condition, the system controls the auxiliary fine-tuning device to activate continuous air supply mode. In continuous air supply mode, the solenoid valve controlling the air circuit remains open, and the air source provides uninterrupted positive air pressure to the cylinder of the auxiliary fine-tuning device. This allows the system to overcome the basic sliding friction with maximum acceleration and deceleration capabilities, performing continuous approach movements towards the target position, thus ensuring the overall time efficiency of the correction process.

[0098] As the continuous approach motion proceeds, the lateral displacement of the plate increases. When the system monitors in real time that the current position deviation has decreased to or below the pulse switching threshold, and confirms that the plate has not yet fully reached the target posture position, the system determines that the plate has entered the precision fine-tuning zone. At this point, if continuous air supply continues, the large amount of compressed gas accumulated in the pneumatic circuit will generate a pneumatic spring effect. Once the plate reaches the target position and the valve is closed, the expansion force of the residual gas will inevitably push the plate past the target position.

[0099] To eliminate the risk of overshoot, the system immediately switches the operating state of the auxiliary fine-tuning device from continuous air supply mode to pulse-jog air supply mode the instant the pulse switching threshold is exceeded. In pulse-jog air supply mode, the system controls the pneumatic reversing valve by outputting high-frequency pulse width modulation (PWM) commands. In this embodiment, the air cut-off interval between high-frequency pulse air supplies (i.e., the dead time of valve closure) is used to forcibly open the exhaust pressure relief channel, thereby quickly releasing the transient residual air pressure in the execution air chamber of the auxiliary fine-tuning device.

[0100] The dynamic principle of this process is as follows: within the air supply pulse width, the briefly established air pressure provides a small momentum increment, accelerating the plate to produce a minimal displacement; during the air cut-off interval, the transient residual air pressure is released, and the propulsion thrust instantly drops to zero. At this point, under the low-friction pushing state, the only remaining weak mechanical friction force is applied in the opposite direction to the plate as the only horizontal force, so that the small lateral kinetic energy of the plate is quickly consumed, thereby achieving forced rapid micro-braking.

[0101] The system controls the sheet metal to move towards the target in a highly precise, intermittent stepping motion (i.e., a stop-and-go micro-displacement mode). This control method avoids the cumulative effect of elastic potential energy in the pneumatic system, ensuring that each pulse action achieves a definite and controllable micro-step. The system maintains this intermittent air supply mode until it detects that the sheet metal has finally and precisely reached the target's position.

[0102] By combining continuous air supply mode with micro-pulse jogging air supply mode, and releasing residual air pressure during the jogging phase using air cut-off intervals, the system effectively balances the correction speed and final positioning accuracy using pure pneumatic components without adding any complex and costly electric servo drive hardware, thus avoiding the problem of repeated adjustments caused by mechanical overshoot.

[0103] In some embodiments of this application, obtaining thickness detection data and gap detection data during the process of the main propulsion device pushing the plate includes: Acquire raw thickness detection data, raw gap detection data, and real-time physical displacement data of the main propulsion device during the pushing process; Based on real-time physical displacement data, the original thickness detection data and original gap detection data collected based on time series are mapped to the preset physical surface coordinate system to generate the surface thickness spatial distribution topology map and the surface gap spatial distribution topology map of the plate. Detect regions of abrupt numerical changes in the spatial distribution topology of surface thickness or surface gap. If the continuous area of ​​the numerically abnormal mutation region is less than or equal to the preset local warping area threshold, the numerically abnormal mutation region is determined to be a local physical deformation of the board. The data portions corresponding to the regions of numerical abnormal changes in the original thickness detection data and original gap detection data are removed to obtain the remaining data after cleaning and removal. Based on the remaining data after cleaning and removal, the thickness detection data and gap detection data are determined.

[0104] In industrial stamping and feeding environments, the edges of individual sheet metal sheets often develop burrs due to preceding shearing processes, or experience localized physical bending and warping during stacking and handling. These localized deformations generate extremely high data peaks when detected by sensors. If the control system directly uses the raw time-series data for averaging, these local extreme values ​​will significantly inflate the overall data level, causing the system to misjudge it as multiple sheet metal sheets overlapping (i.e., a "pseudo-double sheet" alarm).

[0105] To eliminate background noise interference caused by local deformation and extract macroscopic features that characterize the true state of the plate, in this embodiment, during the initial data acquisition phase of system operation, the underlying sensors and control bus acquire heterogeneous data in real time from three dimensions: first, raw thickness detection data acquired based on time series (e.g., time-thickness sequence signals output by high-frequency laser rangefinders or eddy current sensors); second, raw gap detection data acquired based on time series (time-gap sequence signals characterizing the distance between the plate and the reference surface); and third, real-time physical displacement data of the main propulsion device during the pushing process (this data is usually fed back in real time by the absolute encoder of the servo motor or a high-precision grating ruler).

[0106] Since the original thickness and gap data only have a time-domain attribute, they cannot directly reflect the physical morphological distribution of the board surface. Therefore, the system uses real-time physical displacement data and executes a spatiotemporal synchronization algorithm to accurately map the original thickness detection data and original gap detection data collected based on time series to a preset physical surface coordinate system.

[0107] This spatiotemporal mapping process can be characterized by the following kinematic integral model: Let the fixed coordinate reference point of the sensor be... The instantaneous operating speed of the main propulsion equipment at time t is ), then at any sampling time The horizontal coordinate of the physical surface of the board currently scanned by the sensor for: ; Combining the fixed width of the board material or the lateral distribution of the sensor array (denoted as the Y-axis coordinate), the system will assign each timestamp The original thickness value Th ( ) and the original gap value Gap ( Assign spatial coordinates respectively This process transforms a one-dimensional time signal into a two-dimensional physical space signal, thereby generating a topological map of the surface thickness spatial distribution and the surface gap spatial distribution that objectively reflect the current physical form of the board. This topological map is essentially a data matrix containing spatial coordinates and corresponding physical characteristic quantities.

[0108] After generating the above topology map, the system uses spatial digital filtering and feature recognition algorithms to detect regions with abrupt changes in numerical values ​​in the surface thickness spatial distribution topology map or the surface gap spatial distribution topology map.

[0109] Specifically, the system calculates the spatial first derivative (i.e., spatial gradient) of each data node in the topology graph. Taking a thickness topology graph as an example, the spatial gradient G at a point (x, y) is... th (x, y) is calculated as follows: ; When the spatial gradient within a certain segment region When the absolute value of a region exceeds a preset flatness standard deviation, the system identifies it as a region of numerical anomalous change. After extracting the anomalous region, the system further evaluates its physical and geometric properties. The system calculates the continuous two-dimensional area (denoted as S) of the numerical anomalous change region in the topology graph using a polygon approximation algorithm or a connected component labeling algorithm. anomaly Subsequently, the system will transfer the continuous area S anomaly Compared with the preset local warping area threshold (denoted as S) threshold Logical comparison is performed. The threshold for the local warping area is set based on the following: real overlapping of multiple sheets of board (stacked materials) must be a large-area macroscopic surface contact, and the area of ​​the abnormal area usually covers the whole or most of the board; while physical defects such as corners and burrs only manifest as small local point-like or line-like protrusions.

[0110] Therefore, when the continuous area of ​​the region where the system determines the numerical abnormal change is less than or equal to the preset local warping area threshold (i.e. In the case of the current numerical abnormal change area, it is not caused by the double-sheet stacking, but by the local physical deformation of the sheet material during transportation or processing (such as edge bending or protruding burrs).

[0111] The system performs forced masking to remove the data portions corresponding to the abnormal abrupt change regions in the original thickness detection data and original gap detection data in the cache (i.e., those extreme data that are confirmed to be pseudo double-sheet noise), thereby obtaining the remaining data after cleaning and removal.

[0112] Finally, based on the remaining data after cleaning and filtering, the system re-determines the thickness and gap detection data for the current work cycle using statistical methods such as weighted averaging or median filtering. This spatial topology mapping and cleaning step prevents localized physical defects in the board material from contaminating the overall data pool. The resulting thickness and gap detection data accurately reproduce the physical state of the board material, providing a high-quality input source with an extremely high signal-to-noise ratio for subsequent accurate comparison of the double-sheet judgment benchmark values. This eliminates the common industrial phenomenon of system shutdowns caused by minor deformations at the edges of the board material.

[0113] In some embodiments of this application, it also includes: Obtain the periodic average of thickness detection data and the periodic average of smoothing pressure data over multiple consecutive working cycles; If a step jump is detected in the periodic mean of the current thickness detection data and the periodic mean of the current smooth pressure data, and the values ​​after the jump remain stable within the preset observation period, it is determined that a plate specification switching condition has occurred. Thickness detection data, gap detection data, and aerodynamic condition data within a preset observation period are used as reference feature samples; Based on the reference feature samples, determine the thickness standard reference line and the basic load reference line; The double-sheet judgment benchmark value in the system is overwritten and updated based on the thickness standard benchmark and the basic load benchmark.

[0114] In modern flexible manufacturing scenarios, production lines need to frequently switch between processing sheet materials of different specifications (e.g., different thicknesses or densities). When sheet material specifications are legally switched, their physical thickness and the frictional load required for propulsion inevitably change significantly. If the system continues to use the original judgment criteria, it will inevitably lead to continuous false alarms and production line downtime.

[0115] To achieve adaptive type change without manual recalibration, in this embodiment, the system continuously builds a sliding data window during the background monitoring process to obtain the periodic average of thickness detection data over multiple consecutive working cycles (e.g., the previous M historical cycles including the current cycle). and the periodic mean of smoothed pressure data .

[0116] During continuous system operation, the control algorithm compares the current cycle average with the historical moving average in real time. Upon detecting the cycle average of the current thickness detection data... and the periodic mean of the current smoothed pressure data When step transitions occur in all cases, the system triggers the abnormal state tracking mechanism.

[0117] The absolute value of the difference between the current mean and the historical mean exceeds the inherent noise threshold of the equipment's inherent fluctuations. The calculation formula is: ; ; in, and These are the preset thickness background fluctuation threshold and the aerodynamic pressure background fluctuation threshold, respectively.

[0118] Since a simple step change could also be caused by a series of severe material stacking failures, to eliminate fault interference, the system does not immediately determine a changeover. Instead, it initiates a preset observation period (e.g., the subsequent K consecutive work cycles). Within this preset observation period, the system calculates the standard deviation or range of the data after these changes. If the values ​​after the changes are detected to remain stable within the preset observation period (i.e., the statistical variance of the data is lower than the dispersion threshold characterizing a chaotic state), this proves that the material input to the current production line has high consistency and stable physical characteristics, ruling out the possibility of random faults. Based on the above determination, the system ultimately determines that the current equipment has experienced a sheet metal specification switching condition.

[0119] Upon confirming a change in sheet material specifications, the system immediately exits the standard dual-sheet alarm mode and enters a baseline self-learning mode. The system's control and storage module selectively captures and acquires thickness detection data, gap detection data, and aerodynamic condition data within a preset observation period. Because this data has undergone local warping cleanup and steady-state verification, it can purely and accurately reflect the physical parameters and dynamic drag characteristics of the new sheet material specification. Therefore, the system packages this data as a whole as a baseline feature sample for reconstructing the new baseline.

[0120] After acquiring the baseline feature samples, the system algorithm scheduling core begins to recalibrate and calculate the baseline parameters: First, the system extracts all thickness detection data from the baseline feature samples, and determines the thickness standard baseline (Base) representing the standard thickness of the new specification sheet by using the arithmetic mean and combining it with the sensor's own systematic error. thickness Secondly, the system extracts aerodynamic operating condition data (i.e., equivalent air pressure for propulsion resistance) from the baseline feature samples, and determines the baseline of the foundation load required to overcome the dynamic friction of the new specification plate foundation through dynamic load mean calculation. load ).

[0121] Finally, the system integrates the double-sheet determination logic with the two newly calculated physical benchmarks mentioned above. Based on the thickness standard benchmark and the basic load benchmark, the system recalculates the physical threshold for double-sheet triggering according to a preset redundancy coefficient (such as 1.5 times the plate thickness feature or a specific air pressure tolerance coefficient), and overwrites and updates the double-sheet determination benchmark value in the system memory with an overwrite write instruction.

[0122] Using the method in this embodiment, when the system detects a change in material specifications, it does not require manual intervention to stop production, nor does it require operators to re-enter thickness parameters or calibrate air pressure. Instead, it automatically senses the change through steady-state analysis of time-series fluctuations and uses self-collected reliable samples to smoothly evolve and overwrite the core judgment criteria. The method in this embodiment endows the equipment with a high degree of flexibility and adaptability, maximizing the continuous uptime and error-proofing accuracy of the production line under complex operating conditions.

[0123] Secondly, embodiments of this application provide a posture correction system, including: The first acquisition module is used to acquire thickness detection data, gap detection data, aerodynamic condition data and attitude image data during the process of the main propulsion equipment pushing the plate. The first determination module is used to perform feature fusion analysis on thickness detection data, gap detection data and aerodynamic condition data to determine the stacking state of the sheet metal. The stacking state includes single sheet metal state and multiple sheet metal state. The first module is used to extract features from the posture image data to obtain the positional deviation value and the angle deviation value of the plate. The first generation module is used to generate target control commands based on the stacking status, position deviation value, and angle deviation value, and to control the main propulsion equipment and auxiliary fine-tuning equipment to perform corresponding correction actions through the target control commands.

[0124] The functions of each module in each device in the embodiments of this application can be found in the corresponding descriptions in the above methods, and will not be repeated here.

[0125] Figure 2 A structural block diagram of an electronic device according to an embodiment of this application is shown. Figure 2As shown, the electronic device includes a memory 410 and a processor 420. The memory 410 stores instructions that can be executed on the processor 420. When the processor 420 executes the instructions, it implements the posture correction method in the above embodiments. The number of memories 410 and processors 420 can be one or more. This electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0126] The electronic device may also include a communication interface 430 for communicating with external devices and exchanging data. The devices are interconnected using different buses and can be mounted on a common motherboard or otherwise as needed. The processor 420 can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). The bus can be divided into address buses, data buses, control buses, etc. For ease of illustration, Figure 2 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0127] Optionally, in a specific implementation, if the memory 410, processor 420 and communication interface 430 are integrated on a single chip, the memory 410, processor 420 and communication interface 430 can communicate with each other through an internal interface.

[0128] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be microprocessors or any conventional processor. It is worth noting that the processor can be a processor supporting Advanced Reduced Instruction Set Machines (ARM) architecture.

[0129] This application provides a computer-readable storage medium (such as the memory 410 described above) that stores computer instructions, which, when executed by a processor, implement the method provided in this application.

[0130] Optionally, memory 410 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device, etc. Furthermore, memory 410 may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 410 may optionally include memory remotely located relative to processor 420, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0131] Furthermore, the functional devices in the various embodiments of this application can be integrated into a processing module, or each device can exist physically separately, or two or more devices can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0132] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A posture correction method, characterized in that, include: Acquire thickness detection data, gap detection data, aerodynamic condition data, and attitude image data during the process of the main propulsion equipment pushing the plate; The thickness detection data, the gap detection data, and the aerodynamic condition data are subjected to feature fusion analysis to determine the stacking state of the sheet metal, which includes single sheet metal state and multiple sheet metal state. Feature extraction is performed on the posture image data to obtain the positional deviation value and angle deviation value of the plate. Based on the stacking state, the position deviation value, and the angle deviation value, a target control command is generated, and the main propulsion device and the auxiliary fine-tuning device are controlled to perform corresponding correction actions through the target control command.

2. The method according to claim 1, characterized in that, The thickness detection data, the gap detection data, and the aerodynamic operating condition data are subjected to feature fusion analysis to determine the stacking state of the sheet metal. The stacking state includes single sheet metal state and multiple sheet metal state, including: Based on the thickness detection data, determine the thickness fluctuation variance within the current sampling period; Based on the gap detection data, determine the gap fluctuation variance within the current sampling period; Based on the thickness fluctuation variance and the gap fluctuation variance, the real-time reliability weight of the thickness data and the real-time reliability weight of the gap data are obtained, wherein the fluctuation variance is inversely proportional to the corresponding real-time reliability weight. Based on the aforementioned pneumatic operating data, determine the continuous pressure variation curve; The initial peak data in the start-up phase of the continuous pressure change curve is removed to obtain smooth pressure data in the steady-state feeding phase. The load confidence weight of the aerodynamic operating condition data is determined based on the slope of the change in the smoothed pressure data. Based on the real-time confidence weight and the load confidence weight, the thickness detection data, the gap detection data, and the smoothing pressure data are dynamically weighted and summed to obtain the weighted summation result; If the weighted summation result is greater than the preset double-sheet determination benchmark value, the stacking state of the board is determined to be the multi-sheet state.

3. The method according to claim 2, characterized in that, The step of extracting features from the posture image data to obtain the positional deviation value and angle deviation value of the board includes: In the attitude image data, a first projection axis and a second projection axis that are perpendicular to the direction of travel of the plate are determined, and the pixel gradient distribution of the attitude image data on the first projection axis and the second projection axis is calculated to obtain the gradient peak region. The gradient peak region is extracted to obtain discrete edge pixel coordinates. Based on all the edge pixel coordinates, determine the distance residuals from all the edge pixel coordinates to the fitted center line; Edge pixel coordinates with distance residuals greater than a preset noise threshold are removed to obtain an effective set of edge feature points. Based on the set of effective edge feature points, the actual outline line of the plate is obtained; Based on the geometric relationship between the actual contour line and the preset reference coordinate system, the position deviation value and the angle deviation value are obtained. If feature extraction of the attitude image data fails, extract multiple historical position deviation values ​​and historical angle deviation values ​​that the main propulsion device has successfully calculated consecutively within a preset historical time window. Based on the multiple historical position deviation values ​​and the multiple historical angle deviation values, a trend evolution trajectory is generated; Based on the trend evolution trajectory, the estimated position deviation and estimated angle deviation values ​​for the current working cycle are extrapolated and calculated. Based on the estimated position deviation value and the estimated angle deviation value, the position deviation value and angle deviation value are determined to trigger the generation of a specific degraded control command for limiting the maximum permissible propulsion speed of the main propulsion device and coordinating the extension of the physical guide limit device.

4. The method according to claim 3, characterized in that, The step of generating a target control command based on the stacking state, the position deviation value, and the angle deviation value, and controlling the main propulsion device and the auxiliary fine-tuning device to perform corresponding correction actions through the target control command includes: When the stacked state is the single sheet state, and the position deviation value and the angle deviation value are within the preset fine-tuning correction range, a flexible correction command is generated. Based on the flexible correction command, the main propulsion device is controlled to reduce the air supply pressure to a preset maintenance pressure threshold so that the plate is in a low-friction pushing state. While the plate is in the low-friction pushed state, the auxiliary fine-tuning device is controlled to apply a lateral thrust to the side of the plate according to the position deviation value and the angle deviation value until the plate moves to the target posture position. After detecting that the plate has reached the target posture position, the main propulsion device is controlled to restore the normal pushing air supply pressure; When the stacked state is the state of multiple sheets of material, an emergency waste discharge command is generated; Based on the emergency exhaust command, the main air intake channel of the main propulsion device is controlled to exhaust air urgently, and the reverse return air path is controlled to open to the maximum degree of ventilation so that the main propulsion device can quickly return to the initial position. When the main propulsion equipment retracts, the pneumatic flap located at the bottom of the material channel is opened, allowing multiple overlapping sheets of the material to fall into the waste collection area. After the waste is discharged, the locking system is reset. Acquire the pneumatic operating condition data for multiple consecutive working cycles within a preset time window, and acquire the smoothed pressure data corresponding to each working cycle; The slope of the mean change trend of all smoothed pressure data is obtained by calculating the time sliding window. If the slope of the average change trend indicates that the overall pressure is slowly increasing and the increase does not exceed the preset mechanical limit threshold, then the main propulsion equipment is determined to be in a state of natural wear and aging. Based on the slope of the mean change trend, the pneumatic pressure judgment baseline of the double sheet judgment benchmark value in subsequent working cycles is updated proportionally with positive compensation.

5. The method according to claim 4, characterized in that, The step of controlling the auxiliary fine-tuning device to apply a lateral thrust to the side of the plate according to the position deviation value and the angle deviation value until the plate moves to the target posture position includes: The rate of reduction of the positional deviation of the plate during the application of the lateral thrust is obtained; If the current position deviation is greater than the preset pulse switching threshold, the auxiliary fine-tuning device is controlled to start the continuous air supply mode in order to perform a continuous approaching action toward the target attitude position. If the current position deviation value decreases to be equal to or less than the pulse switching threshold, and the target attitude position has not yet been reached, the continuous air supply mode is switched to the pulse jog air supply mode. In the pulse-driven air supply mode, the transient residual air pressure in the auxiliary fine-tuning device is released by utilizing the air interruption interval between pulse air supplies, and the plate is controlled to move in a step-by-step manner until it reaches the target posture position.

6. The method according to claim 5, characterized in that, The acquisition of thickness detection data and gap detection data during the process of the main propulsion equipment pushing the plate includes: Acquire raw thickness detection data, raw gap detection data, and real-time physical displacement data of the main propulsion device during the pushing process; Based on the real-time physical displacement data, the original thickness detection data and the original gap detection data collected based on the time series are mapped to a preset physical surface coordinate system to generate a surface thickness spatial distribution topology map and a surface gap spatial distribution topology map of the plate. Detect regions of abrupt numerical changes in the surface thickness spatial distribution topology map or the surface gap spatial distribution topology map; If the continuous area of ​​the numerically abnormal change region is less than or equal to a preset local warping area threshold, the numerically abnormal change region is determined to be a local physical deformation of the board material. The data portions corresponding to the numerical abnormal change regions in the original thickness detection data and the original gap detection data are removed to obtain the remaining data after cleaning and removal. Based on the remaining data after cleaning and removal, thickness detection data and gap detection data are determined.

7. The method according to claim 6, characterized in that, Also includes: Obtain the periodic average of the thickness detection data and the periodic average of the smoothing pressure data over multiple consecutive working cycles; If a step jump is detected in the current average value of the thickness detection data and the current average value of the smoothing pressure data, and the values ​​after the jump remain stable within a preset observation period, it is determined that a plate specification switching condition has occurred. The thickness detection data, the gap detection data, and the aerodynamic condition data within the preset observation period are obtained as reference feature samples; Based on the aforementioned reference feature samples, determine the thickness standard reference line and the basic load reference line; The double-sheet determination benchmark value in the system is overwritten and updated based on the thickness standard benchmark and the basic load benchmark.

8. A posture correction system, characterized in that, include: The first acquisition module is used to acquire thickness detection data, gap detection data, aerodynamic condition data and attitude image data during the process of the main propulsion equipment pushing the plate. The first determining module is used to perform feature fusion analysis on the thickness detection data, the gap detection data and the aerodynamic condition data to determine the stacking state of the sheet metal, which includes single sheet metal state and multiple sheet metal state. The first obtaining module is used to extract features from the posture image data to obtain the position deviation value and angle deviation value of the plate. The first generation module is used to generate target control commands based on the stacking state, the position deviation value, and the angle deviation value, and to control the main propulsion device and the auxiliary fine-tuning device to perform corresponding correction actions through the target control commands.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method as described in any one of claims 1-7.

Images