Shaft sleeve riveting automatic hole searching and positioning method and system based on artificial intelligence

Abstract

The invention relates to the technical field of automatic riveting devices, in particular to an automatic hole searching and positioning method and system for shaft sleeve riveting based on artificial intelligence. According to the automatic hole searching and positioning method and system for shaft sleeve riveting based on artificial intelligence, the distance between a pull rod and a workpiece can be accurately measured and controlled by adopting a laser sensor and air cylinder stroke matching technology, and the expansion amount of an air cylinder and the position of the pull rod are corrected in real time; the time sequence relation between historical thrust and displacement data is analyzed through a long-short-term memory network, the synchronism of thrust and offset is dynamically adjusted, the coordination between the pull rod and an air cylinder is optimized, the consistency and stability of equipment are kept within the working range, and the precision of the pull rod displacement is remarkably improved. By detecting the real-time data of the levelness of the supporting surface and the penetration angle of the pull rod, the angle between the supporting surface of the clamp and the pull rod can be quickly judged and adjusted to be within an accurate range, so that the instability in mechanical operation is avoided, and the response speed, the accuracy and the automation degree of a riveting system are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of automatic riveting device technology, and in particular to an automatic hole-finding and positioning method and system for bushing riveting based on artificial intelligence. Background Technology

[0002] The field of automatic riveting devices mainly includes automated equipment and systems that enable automatic riveting, welding, and press-fitting processes. They are commonly used to connect different parts, especially in the manufacturing of industrial products. They have significant advantages in improving production efficiency, reducing operating costs, and enhancing processing accuracy. Automatic riveting devices integrate intelligent technologies, sensors, automatic positioning systems, and artificial intelligence algorithms, enabling the equipment to autonomously identify and locate hole positions and automatically adjust parameters to achieve higher precision riveting and production efficiency.

[0003] The purpose of this artificial intelligence-based automatic hole-finding and positioning method for bushing riveting is to solve the problems of inefficiency, high error rate, and high cost in the traditional bushing riveting process, especially in automated production where it is difficult to maintain accuracy and reliability. The core objective is to use artificial intelligence technology, through machine vision and sensor systems, to achieve automatic positioning and recognition of workpieces, ensure the accuracy of the riveting process, reduce external intervention, lower the error rate, and further improve the efficiency and stability of automated production.

[0004] Existing technologies often rely on semi-automatic equipment for workpiece positioning, parameter adjustment, and position calibration during automatic riveting processes. This results in low efficiency and large errors. Traditional equipment often fails to provide sufficient accuracy and consistency when positioning workpieces, especially when dealing with complex hole positions. This leads to low riveting accuracy, high production costs, and difficult equipment maintenance. Although traditional laser measurement technology can measure position, it cannot adjust the coordination of thrust and displacement in real time under dynamic conditions. Furthermore, it lacks dynamic feedback and optimization of the interaction between the cylinder and the tie rod, which limits the workpiece positioning accuracy and processing efficiency. This results in fixture instability and positional deviation, all of which affect the smooth progress of the riveting process and fail to provide sufficiently high accuracy, automation, and the ability to cope with changes. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and propose an automatic hole finding and positioning method and system for bushing riveting based on artificial intelligence.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an automatic hole-finding and positioning method for bushing riveting based on artificial intelligence, comprising the following steps:

[0007] Step 1: Based on the laser sensor, obtain the distance between the pull rod and the workpiece plane, compare the workpiece height edge boundary and update the pull rod displacement starting point, calculate the pull rod displacement and match the cylinder stroke, and generate the workpiece alignment coordinates;

[0008] Step 2: Based on the workpiece alignment coordinates, read the current extension length of the pull rod and the real-time pressure of the cylinder, measure the hole position reference difference and obtain the offset limit, determine whether the difference exceeds the limit, and then perform synchronous positioning verification to generate a position deviation report.

[0009] Step 3: Based on the position deviation report, adjust the coordination between the tie rod offset and the cylinder thrust, use a long short-term memory network to dynamically adjust the synchronization between the thrust and offset, determine the balance between the horizontal correction of the fixture support surface and the penetration angle of the tie rod, and generate cylinder correction parameters.

[0010] Step 4: Based on the cylinder correction parameters, adjust the fixture movement path and clamping force threshold, compare the pressure between the positioning hook and the hole edge, determine whether the fixture contact is stable, set the fixture stop position, and generate a fixture positioning mark;

[0011] Step 5: Based on the fixture positioning mark, a convolutional neural network is used to monitor the overlap between the positioning pin penetration path and the proximity switch position, confirm the signal triggering state and control the riveting system, determine the triggering validity and call the riveting task sequence to generate a riveting completion mark.

[0012] As a further aspect of the present invention, the workpiece alignment coordinates are specifically the three-dimensional spatial distance between the current position of the pull rod and the center point of the workpiece, covering the coordinate values ​​of the X-axis, Y-axis and Z-axis. The position deviation report includes the displacement difference between the pull rod and the target hole, the deviation of the cylinder thrust from the preset value, and the compliance judgment within the tolerance range. The cylinder correction parameters include the adjusted cylinder thrust range, the synchronization correction value of the pull rod displacement and thrust, and the triggering timing setting of the synchronous linkage. The fixture positioning mark is specifically the final positioning coordinate when the fixture stops, the threshold setting of the clamping force, and the contact pressure value between the positioning hook and the edge of the workpiece. The riveting completion mark includes the actual overlap of the positioning pin penetration path, the triggering state of the proximity switch, and the task execution confirmation signal of the riveting system operation.

[0013] As a further aspect of the present invention, the specific steps for generating the workpiece alignment coordinates are as follows:

[0014] Based on a laser sensor, the vertical distance between the workpiece surface and the tie rod is measured. The distance between the current tie rod position and the target hole position is compared to calculate the actual displacement. The extension distance of the tie rod is adjusted and compared with the workpiece surface to generate the target path of the tie rod displacement.

[0015] Based on the target displacement path of the tie rod, the extension and retraction of the cylinder are adjusted, the working stroke of the cylinder is set, the tie rod is accurately aligned, the cylinder is corrected and the tie rod is extended and retracted synchronously, the final positions of the cylinder and the tie rod are compared, and the workpiece alignment coordinates are generated.

[0016] As a further aspect of the present invention, the specific steps for generating the position deviation report are as follows:

[0017] Based on the workpiece alignment coordinates, the extension length of the pull rod and the cylinder pressure value are read, the difference between the hole position and the target hole position is measured, the deviation between the pull rod position and the predetermined position is calculated, the measured deviation is compared with the standard tolerance value, and the position deviation value is generated.

[0018] Based on the position deviation value, it is determined whether the error exceeds the preset tolerance range. If it does, the pull rod displacement and cylinder pressure are adjusted to calibrate the fixture position and align the workpiece, complete the synchronous positioning verification, and generate a position deviation report.

[0019] As a further aspect of the present invention, the specific steps for generating the cylinder correction parameters are as follows:

[0020] Based on the position deviation report, the displacement of the tie rod and the cylinder thrust value are read, the offset of the tie rod is adjusted, and the temporal relationship between historical thrust and displacement data is analyzed through a long short-term memory network to make dynamic adjustments, optimize the synchronization of thrust and offset, and make the two match each other within the working range to generate a coordinated adjustment result.

[0021] Based on the coordination adjustment results, the levelness of the fixture support surface and the penetration angle of the tie rod are detected, the actual levelness of the support surface is compared with the standard value, the angle of the fixture support surface and the tie rod are adjusted, it is determined whether the preset correction threshold is reached, and the fixture correction result is generated.

[0022] Based on the fixture correction results, the current displacement value of the pull rod and the current output pressure of the cylinder are locked synchronously, the corresponding action parameters of the two are frozen, and the signal of the pull rod stop position and the contact state of the positioning pin is compared to generate cylinder correction parameters.

[0023] As a further aspect of the present invention, the specific execution process of the long short-term memory network includes: based on the position deviation report, reading historical thrust and displacement data, analyzing the temporal relationship and making dynamic adjustments; firstly, taking the historical thrust data and the rod displacement as input, learning the long-term dependency relationship between thrust and displacement; and optimizing the rod offset according to the output adjustment parameters, so that the thrust and offset are synchronized and matched within the working range, generating a coordinated adjustment result.

[0024] As a further aspect of the present invention, the specific steps for generating the fixture positioning mark are as follows:

[0025] Based on the cylinder correction parameters, the fixture movement path and clamping force parameters are read, the starting point and target position of the fixture movement are adjusted, the workpiece surface and the fixture clearance are compared, the fixture movement trajectory and clamping force output are calculated, and the fixture movement path is generated.

[0026] Based on the movement path of the clamp, the pressure changes between the positioning hook and the edge of the hole are compared, the pressure feedback of the clamp at the time of contact is monitored, the deviation between the clamp pressure value and the target value is analyzed, the displacement when the clamp stops is confirmed, and a clamp positioning mark is generated.

[0027] As a further aspect of the present invention, the specific steps for generating the riveting completion identifier are as follows:

[0028] Based on the fixture positioning mark, a convolutional neural network is used to compare the penetration path of the positioning pin with the trigger position of the proximity switch in real time, determine the degree of overlap between the two, and ensure that the positioning pin completely penetrates the hole, thereby generating the degree of overlap of the penetration path.

[0029] Based on the overlap of the penetration path, monitor the state of the proximity switch, compare the change amplitude of the signal, determine whether the signal is triggered, and generate a signal trigger state.

[0030] Based on the signal triggering state, the validity of the signal is determined. If the validity condition is met, the riveting system is started, and the operation is carried out according to the riveting sequence to generate a riveting completion mark.

[0031] As a further aspect of the present invention, the specific execution process of the convolutional neural network includes: based on the fixture positioning mark, comparing the penetration path of the positioning pin with the trigger position of the proximity switch in real time; inputting the penetration path image of the positioning pin and the trigger position image of the proximity switch into the convolutional neural network; the network extracts image features through multi-layer convolution operations and calculates the overlap between the path and the trigger position; and then judging the overlap between the two based on the feature extraction results to ensure that the positioning pin completely penetrates the hole and generate the penetration path overlap.

[0032] An AI-based automatic hole-finding and positioning system for bushing riveting, wherein the AI-based automatic hole-finding and positioning system for bushing riveting is used to execute the aforementioned AI-based automatic hole-finding and positioning method for bushing riveting, the system comprising:

[0033] Workpiece alignment module: Based on a laser sensor, it obtains the distance between the pull rod and the workpiece plane, compares the boundary information of the workpiece height edge, updates the starting point of the pull rod displacement, calculates the pull rod displacement, and matches it with the cylinder stroke to generate workpiece alignment coordinates;

[0034] Position deviation verification module: Based on the workpiece alignment coordinates, read the current extension length of the pull rod and the real-time pressure of the cylinder, measure the hole position reference difference and obtain the offset limit, determine whether the difference exceeds the limit range, if it exceeds the limit, perform synchronous positioning verification and generate a position deviation report;

[0035] Cylinder correction module: Based on the position deviation report, adjust the coordination between the offset of the tie rod and the cylinder thrust. Use a long short-term memory network to analyze the temporal relationship between historical thrust and displacement data, make dynamic adjustments, optimize the synchronization of thrust and offset, so that the two match each other within the working range, and generate cylinder correction parameters.

[0036] Fixture positioning module: Based on the cylinder correction parameters, adjust the movement path and clamping force threshold of the fixture, compare the pressure between the positioning hook and the edge of the hole, determine whether the fixture contact is stable, set the fixture stop position, and generate a fixture positioning mark;

[0037] Riveting completion module: Based on the fixture positioning mark, a convolutional neural network is used to monitor the overlap between the positioning pin penetration path and the proximity switch position, confirm the signal triggering state and control the riveting system, determine whether the signal is valid and call the riveting task sequence, and generate a riveting completion mark.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] 1. In this invention, by employing laser sensor and cylinder stroke matching technology, the distance between the pull rod and the workpiece can be accurately measured and controlled, and the cylinder extension and retraction amount and pull rod position can be corrected in real time, thereby achieving precise alignment of the workpiece and significantly improving the accuracy of the pull rod displacement;

[0040] 2. In this invention, the temporal relationship between historical thrust and displacement data is analyzed by long short-term memory network, the synchronization of thrust and offset is dynamically adjusted, the coordination between the tie rod and the cylinder is optimized, the lag and error between thrust and displacement in the traditional method are eliminated, and the equipment maintains consistency and stability within the working range.

[0041] 3. In this invention, by detecting real-time data of the levelness of the support surface and the penetration angle of the tie rod, it is possible to quickly determine and adjust the angle between the clamp support surface and the tie rod within a precise range, thereby avoiding instability in mechanical operation and significantly improving the response speed, accuracy and automation of the riveting system. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0043] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0046] Example 1

[0047] Please see Figure 1 This invention provides a technical solution: an automatic hole-finding and positioning method for bushing riveting based on artificial intelligence, comprising the following steps:

[0048] Step 1: Based on the laser sensor, obtain the distance between the pull rod and the workpiece plane, compare the workpiece height edge boundary and update the pull rod displacement starting point, calculate the pull rod displacement and match the cylinder stroke, and generate the workpiece alignment coordinates;

[0049] Step 2: Based on the workpiece alignment coordinates, read the current extension length of the pull rod and the real-time pressure of the cylinder, measure the difference between the hole position reference and obtain the offset limit, determine whether the difference exceeds the limit, and then perform synchronous positioning verification to generate a position deviation report.

[0050] Step 3: Based on the position deviation report, adjust the coordination between the tie rod offset and the cylinder thrust, use a long short-term memory network to dynamically adjust the synchronization between the thrust and offset, determine the balance between the horizontal correction of the fixture support surface and the penetration angle of the tie rod, and generate cylinder correction parameters.

[0051] Step 4: Based on the cylinder correction parameters, adjust the fixture movement path and clamping force threshold, compare the pressure between the positioning hook and the edge of the hole, determine whether the fixture contact is stable, set the fixture stop position, and generate a fixture positioning mark;

[0052] Step 5: Based on the fixture positioning mark, a convolutional neural network is used to monitor the overlap between the positioning pin penetration path and the proximity switch position, confirm the signal triggering status and control the riveting system, determine the triggering validity and call the riveting task sequence to generate a riveting completion mark.

[0053] The workpiece alignment coordinates are specifically the three-dimensional spatial distance between the current position of the pull rod and the center point of the workpiece, covering the coordinate values ​​of the X-axis, Y-axis and Z-axis. The position deviation report includes the displacement difference between the pull rod and the target hole, the deviation of the cylinder thrust from the preset value, and the compliance judgment within the tolerance range. The cylinder correction parameters include the adjusted cylinder thrust range, the synchronization correction value of the pull rod displacement and thrust, and the triggering timing setting of the synchronous linkage. The fixture positioning mark is specifically the final positioning coordinate when the fixture stops, the threshold setting of the clamping force, and the contact pressure value between the positioning hook and the edge of the workpiece. The riveting completion mark includes the actual overlap of the positioning pin penetration path, the triggering status of the proximity switch, and the task execution confirmation signal of the riveting system operation.

[0054] The specific steps for generating workpiece alignment coordinates are as follows:

[0055] Based on a laser sensor, the vertical distance between the workpiece surface and the tie rod is measured. The distance between the current tie rod position and the target hole position is compared to calculate the actual displacement. The extension distance of the tie rod is adjusted and compared with the workpiece surface to generate the target path of the tie rod displacement.

[0056] Based on the target path of the tie rod displacement, the extension and retraction of the cylinder are adjusted, the working stroke of the cylinder is set, the tie rod is accurately aligned, the cylinder is corrected and the tie rod is extended and retracted synchronously, the final positions of the cylinder and the tie rod are compared, and the workpiece alignment coordinates are generated.

[0057] Based on a laser sensor, the vertical distance between the workpiece surface and the tie rod is measured. The distance between the current tie rod position and the target hole position is compared to calculate the actual displacement. The three-dimensional coordinates of the workpiece surface and the tie rod are input, and the actual displacement between the tie rod and the workpiece is calculated using the Euclidean distance formula. Based on the actual displacement, the extension distance of the tie rod is adjusted, the displacement of the tie rod is adjusted, the target extension distance is set and compared with the workpiece surface, the extension accuracy of the tie rod is adjusted, and the target path of the tie rod displacement is generated.

[0058] Based on the target displacement path of the tie rod, the extension and retraction of the cylinder are adjusted, the working stroke of the cylinder is set, the current stroke of the cylinder and the preset target stroke are input, the extension and retraction of the cylinder are adjusted to achieve the precise working stroke, and the displacement of the tie rod is synchronized. The adjustment of the cylinder extension and retraction depends on the current cylinder pressure value and the extension length of the tie rod. The final position of the tie rod and the cylinder is compared in real time, and the difference between the target value and the current value is compared using an error correction method, and adjustments are made to generate workpiece alignment coordinates.

[0059] The specific steps for generating a position deviation report are as follows:

[0060] Based on the workpiece alignment coordinates, the extension length of the pull rod and the cylinder pressure value are read, the difference between the hole position and the target hole position is measured, the deviation between the pull rod position and the predetermined position is calculated, the measured deviation is compared with the standard tolerance value, and the position deviation value is generated.

[0061] Based on the position deviation value, determine whether the error exceeds the preset tolerance range. If it does, adjust the pull rod displacement and cylinder pressure, calibrate the fixture position and align the workpiece, complete the synchronous positioning verification, and generate a position deviation report.

[0062] Based on the workpiece alignment coordinates, the extension length of the pull rod and the cylinder pressure value are read, the difference between the hole position and the target hole position is calculated, the deviation between the pull rod position and the predetermined position is calculated using the Euclidean distance formula, and compared with the preset tolerance range. The current position coordinates of the pull rod and the target hole position coordinates are input, and the spatial difference between the two is obtained through the three-dimensional spatial calculation formula. The error between the actual displacement of the pull rod and the target position is obtained, and the error value is compared with the standard tolerance value to generate the position deviation value.

[0063] Based on the position deviation value, it is determined whether the error exceeds the preset tolerance range. If it exceeds the tolerance range, the displacement of the adjustment rod and the cylinder pressure are adjusted. The current cylinder pressure and the extension length of the adjustment rod are read. The cylinder pressure is adjusted through closed-loop control to make the adjustment rod accurately aligned with the target hole. Then, according to the adjusted deviation, synchronous positioning verification is performed to bring the error value back to the allowable range. The position of the adjustment rod and the cylinder are gradually adjusted through iterative optimization to generate a position deviation report.

[0064] The specific steps for generating cylinder correction parameters are as follows:

[0065] Based on the position deviation report, the displacement of the tie rod and the cylinder thrust value are read, the offset of the tie rod is adjusted, and the temporal relationship between historical thrust and displacement data is analyzed through a long short-term memory network to make dynamic adjustments, optimize the synchronization of thrust and offset, and make the two match each other within the working range to generate a coordinated adjustment result.

[0066] Based on the coordination adjustment results, the levelness of the fixture support surface and the penetration angle of the tie rod are detected. The actual levelness of the support surface is compared with the standard value. The angles of the fixture support surface and the tie rod are adjusted to determine whether the preset correction threshold has been reached and the fixture correction result is generated.

[0067] Based on the fixture correction results, the current displacement value of the pull rod and the current output pressure of the cylinder are locked simultaneously, the corresponding action parameters of the two are frozen, and the signal of the pull rod stop position and the contact state of the positioning pin are compared to generate the cylinder correction parameters.

[0068] Based on the position deviation report, the displacement of the connecting rod and the cylinder thrust value are read. By collecting the temporal relationship of historical thrust and displacement data, the data is input into the long short-term memory network. The long short-term memory network analyzes the time dependence in the data through the recurrent neuron structure, and then calculates the optimization parameters of thrust and displacement through feedforward propagation. The network weights are updated using the backpropagation algorithm to gradually optimize the synchronization of thrust and displacement and generate a coordinated adjustment result.

[0069] Based on the coordination adjustment results, the horizontality of the fixture support surface and the penetration angle of the tie rod are detected. Data from the laser sensor is read, and the error between the actual horizontality of the fixture support surface and the standard value is compared. The fixture support surface is adjusted in real time to make the contact surface between the fixture and the workpiece parallel. Then, the tie rod angle is adjusted through geometric angle correction to make the penetration angle consistent with the target hole position. It is determined whether the preset correction threshold has been reached after the adjustment, and the fixture correction result is generated.

[0070] Based on the fixture correction results, the current displacement value of the synchronous pull rod and the current output pressure of the cylinder are monitored in real time. The corresponding action parameters of the two are frozen to keep the system in a stable state. The signals of the pull rod stop position and the contact state of the positioning pin are compared. The error between the contact signal and the positioning data is calculated and adjusted to make the signal valid, and the cylinder correction parameters are generated.

[0071] The specific execution process of the Long Short-Term Memory Network includes: based on the position deviation report, reading historical thrust and displacement data, analyzing the temporal relationship through the Long Short-Term Memory Network, and then making dynamic adjustments. First, the historical thrust data and the displacement of the connecting rod are used as inputs. After processing by the Long Short-Term Memory Network, the long-term dependency relationship between thrust and displacement is learned. According to the adjustment parameters output by the Long Short-Term Memory Network, the offset of the connecting rod is optimized to make the thrust and offset synchronous and match each other within the working range, generating a coordinated adjustment result.

[0072] Long Short-Term Memory (LSTM) networks, according to the formula:

[0073] ;

[0074] in: Indicates the first The result value of the synchronization adjustment at any moment. Indicates the first The input vector at time t, Indicates the first The hidden state at all times This represents the input weight matrix, used to weight the input vector. Mapped to the hidden layer, This represents the cyclic weight matrix, used to store historical hidden states. Mapped to the current computation, This represents the time step weight matrix, used to correct for the dynamic response delay between the rod displacement and cylinder thrust caused by the sampling interval. Indicates the first The time step parameter at each moment, This represents the clamp angle correction weight matrix, used to reflect the influence of the clamp support surface levelness and the tie rod penetration angle on synchronization. Indicates the first The amount of adjustment for the angle of the clamp support surface at any given time. Indicates the bias term. This represents the weighting coefficient, initially set to 0.8; Represents the hyperbolic tangent activation function;

[0075] Execution process: First, read the... Input vector at time step The input vector includes the rod displacement and cylinder thrust value, and is processed by the input weight matrix. Map the vector to the hidden layer space, and simultaneously call the first... Hidden state of time The state is determined by the cyclic weight matrix. The transmission is used to represent the temporal characteristics of historical thrust and displacement, followed by the introduction of the time step parameter. And through the time step weight matrix Weighting is applied to correct the asynchrony between the pull rod and cylinder caused by dynamic sampling delay. Then, a laser sensor is introduced to measure the angle difference between the levelness of the support surface and the standard value. The weight matrix is ​​corrected by adjusting the clamp angle. Weighting and correcting the impact of angular deviation on synchronization results involves weighting and summing the inputs and adding a bias term. Then through weighting coefficients Adjusting the overall contribution using the hyperbolic tangent function Activate and generate the first Synchronization adjustment result value at time .

[0076] The specific steps for generating fixture positioning marks are as follows:

[0077] Based on the cylinder correction parameters, the fixture movement path and clamping force parameters are read, the starting point and target position of the fixture movement are adjusted, the workpiece surface and the fixture clearance are compared, the fixture movement trajectory and clamping force output are calculated, and the fixture movement path is generated.

[0078] Based on the movement path of the fixture, the pressure changes between the positioning hook and the edge of the hole are compared, the pressure feedback of the fixture when in contact is monitored, the deviation between the fixture pressure value and the target value is analyzed, the displacement when the fixture stops is confirmed, and the fixture positioning mark is generated.

[0079] Based on cylinder correction parameters, the fixture movement path and clamping force parameters are read, the starting point and target position of the fixture are calculated, and then input into the path planning model to dynamically adjust the target path. The trajectory of the fixture movement path is optimized through backpropagation. By comparing the sensor data of the gap between the fixture and the workpiece during the fixture movement, the movement trajectory and clamping force output of the fixture are calculated, so as to achieve precise control of the fixture in each movement stage and generate the fixture movement path.

[0080] Based on the movement path of the fixture, the pressure changes between the positioning hook and the edge of the hole are compared. Pressure data of the contact surface of the fixture is obtained through a pressure sensor and compared with the target value in real time. The pressure signal is noise filtered to eliminate interference. Standard deviation analysis is used to analyze the pressure change amplitude to determine whether the deviation from the target value exceeds the preset range. The displacement when the fixture stops is confirmed and a fixture positioning mark is generated.

[0081] The specific steps for generating the riveting completion mark are as follows:

[0082] Based on the fixture positioning mark, a convolutional neural network is used to compare the penetration path of the positioning pin with the trigger position of the proximity switch in real time, determine the degree of overlap between the two, and ensure that the positioning pin completely penetrates the hole, thereby generating the degree of overlap of the penetration path.

[0083] Based on the overlap of the penetration path, the state of the proximity switch is monitored, the change amplitude of the signal is compared, it is determined whether the signal is triggered, and a signal trigger state is generated.

[0084] Based on the signal triggering state, the validity of the signal is determined. If the validity condition is met, the riveting system is started, and the operation is carried out according to the riveting sequence to generate a riveting completion mark.

[0085] Based on the fixture positioning mark, a convolutional neural network is used to process the penetration path image of the positioning pin and the trigger position image of the proximity switch in real time. The module with feature extraction capability in the convolutional structure is selected for the initial encoding of the image input. The image input resolution is set to 640×480, and the image channel order is defined as RGB format. The image edge and structural features are extracted by layer-by-layer scanning. Parallel convolution calculation is performed on the two types of images using three image windows of different scales. After calculation, the spatial position mapping overlap of the two images is output through the classification structure, and matching judgment is performed to generate the penetration path overlap degree.

[0086] Based on the overlap of the penetration path, the status of the proximity switch is monitored, and the numerical fluctuation of the signal voltage during the change process is read. Voltage sampling is performed within a set detection period, with the sampling period set to 1,000 times per second. The signal amplitude difference between adjacent time points is compared in each period. If the difference value exceeds the preset amplitude, it is judged that the signal change is significant when the amplitude range is preset to be above 0.2 volts. The trigger time is marked on the sampling time axis, and the number of consecutive triggers and the distribution range are recorded to determine whether the proximity switch is in working state and generate a signal trigger state.

[0087] Based on the signal trigger status, the validity of the trigger is determined by the signal verification logic. The effective trigger count range is set to five or more consecutive stable trigger records within a specified operation cycle. The judgment logic is embedded in the status check node before the riveting process is executed. If the specified number and duration are met, the action control unit of the riveting system is activated to control the actuator to complete the task nodes in the riveting operation instruction list, including the start, pressurization, displacement, depressurization and retraction stages. The system feedback instructions confirm the process loop and generate a riveting completion mark.

[0088] The specific execution process of the convolutional neural network includes: based on the positioning mark of the fixture, comparing the penetration path of the positioning pin with the trigger position of the proximity switch in real time; inputting the image of the penetration path of the positioning pin and the image of the trigger position of the proximity switch into the convolutional neural network; the network extracts image features through multi-layer convolution operations and calculates the overlap between the path and the trigger position; and then judges the overlap between the two based on the feature extraction results to ensure that the positioning pin completely penetrates the hole and generates the penetration path overlap.

[0089] Convolutional neural networks, according to the formula:

[0090] ;

[0091] in: Indicates the first The overlap value of the penetration path at the location. This indicates the index of the output feature map along the row direction; This indicates the index of the output feature map in the column direction. Indicates the input image in The pixel value of the location, Indicates the convolution kernel at the th... The weight parameters of the location, where For row direction index, For column direction index, This indicates the size of the convolution kernel in the row direction. This indicates the size of the convolution kernel in the column direction. Indicates the first Voltage difference at location, Indicates the first The color channel intensity difference at a location is calculated from the RGB three-channel intensity. This represents the voltage differential weighting coefficient. Indicates the color channel weighting coefficient. This represents the bias term, used to correct the baseline value of the convolution result. This represents the adjustment factor, used to balance the overall contribution of the convolution calculation and the correction term. The initial value is set to 0.9. Represents the hyperbolic tangent function;

[0092] Execution process: First, read the pixel values ​​of the input image. The pixel value is formed by superimposing the image of the positioning pin's penetration path and the proximity switch's position, and then weighted by the convolution kernel. Local features of the image are extracted, and the instantaneous voltage fluctuation difference of the proximity switch voltage within the sampling period is calculated. Through weighting coefficients Weighting is applied to supplement the dynamic electrical signal characteristics, and the intensity difference of the RGB three channels is calculated to introduce color channel difference. Through weighting coefficients Weighting is used to enhance the accuracy of image edge detection. Then, the convolution calculation result is summed with the correction term and a bias term is added. The overall result is then adjusted by the coefficient. The correction and adjustment coefficients are constructed using an error function and iteratively updated using the gradient descent method to ensure that signal and image features remain consistent under dynamic conditions, through the hyperbolic tangent function. Activate the generation of penetration path overlap value .

[0093] An AI-based automatic hole-finding and positioning system for bushing riveting, the system comprising:

[0094] Workpiece alignment module: Based on a laser sensor, it obtains the distance between the pull rod and the workpiece plane, compares the boundary information of the workpiece height edge, updates the starting point of the pull rod displacement, calculates the pull rod displacement, and matches it with the cylinder stroke to generate workpiece alignment coordinates;

[0095] Position deviation verification module: Based on the workpiece alignment coordinates, read the current extension length of the pull rod and the real-time pressure of the cylinder, measure the hole position reference difference and obtain the offset limit, determine whether the difference exceeds the limit range, if it exceeds the limit, perform synchronous positioning verification and generate a position deviation report;

[0096] Cylinder correction module: Based on the position deviation report, adjust the coordination between the offset of the tie rod and the cylinder thrust. Use long short-term memory network to analyze the temporal relationship between historical thrust and displacement data, make dynamic adjustments, optimize the synchronization of thrust and offset, so that the two match each other within the working range, and generate cylinder correction parameters.

[0097] Fixture positioning module: Based on cylinder correction parameters, adjust the movement path and clamping force threshold of the fixture, compare the pressure between the positioning hook and the edge of the hole to determine whether the fixture contact is stable, set the fixture stop position, and generate a fixture positioning mark;

[0098] Riveting completion module: Based on the fixture positioning mark, a convolutional neural network is used to monitor the overlap between the positioning pin penetration path and the proximity switch position, confirm the signal triggering state and control the riveting system, determine whether the signal is valid and call the riveting task sequence, and generate a riveting completion mark.

[0099] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An automatic hole-finding and positioning method for bushing riveting based on artificial intelligence, characterized in that, Includes the following steps: Step 1: Based on the laser sensor, obtain the distance between the pull rod and the workpiece plane, compare the workpiece height edge boundary and update the pull rod displacement starting point, calculate the pull rod displacement and match the cylinder stroke, and generate the workpiece alignment coordinates; Step 2: Based on the workpiece alignment coordinates, read the current extension length of the pull rod and the real-time pressure of the cylinder, measure the hole position reference difference and obtain the offset limit, determine whether the difference exceeds the limit, and then perform synchronous positioning verification to generate a position deviation report. Step 3: Based on the position deviation report, adjust the coordination between the tie rod offset and the cylinder thrust, use a long short-term memory network to dynamically adjust the synchronization between the thrust and offset, determine the balance between the horizontal correction of the fixture support surface and the penetration angle of the tie rod, and generate cylinder correction parameters. Step 4: Based on the cylinder correction parameters, adjust the fixture movement path and clamping force threshold, compare the pressure between the positioning hook and the hole edge, determine whether the fixture contact is stable, set the fixture stop position, and generate a fixture positioning mark; Step 5: Based on the fixture positioning mark, a convolutional neural network is used to monitor the overlap between the positioning pin penetration path and the proximity switch position, confirm the signal triggering state and control the riveting system, determine the triggering validity and call the riveting task sequence to generate a riveting completion mark.

2. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The workpiece alignment coordinates are specifically the three-dimensional spatial distance between the current position of the pull rod and the center point of the workpiece, covering the X-axis, Y-axis and Z-axis coordinate values. The position deviation report includes the displacement difference between the pull rod and the target hole, the deviation of the cylinder thrust from the preset value, and the compliance judgment within the tolerance range. The cylinder correction parameters include the adjusted cylinder thrust range, the synchronization correction value of the pull rod displacement and thrust, and the triggering timing setting of the synchronous linkage. The fixture positioning mark is specifically the final positioning coordinate when the fixture stops, the threshold setting of the clamping force, and the contact pressure value between the positioning hook and the edge of the workpiece. The riveting completion mark includes the actual overlap of the positioning pin penetration path, the triggering status of the proximity switch, and the task execution confirmation signal of the riveting system operation.

3. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific steps for generating the workpiece alignment coordinates are as follows: Based on a laser sensor, the vertical distance between the workpiece surface and the tie rod is measured. The distance between the current tie rod position and the target hole position is compared to calculate the actual displacement. The extension distance of the tie rod is adjusted and compared with the workpiece surface to generate the target path of the tie rod displacement. Based on the target displacement path of the tie rod, the extension and retraction of the cylinder are adjusted, the working stroke of the cylinder is set, the tie rod is accurately aligned, the cylinder is corrected and the tie rod is extended and retracted synchronously, the final positions of the cylinder and the tie rod are compared, and the workpiece alignment coordinates are generated.

4. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific steps for generating the position deviation report are as follows: Based on the workpiece alignment coordinates, the extension length of the pull rod and the cylinder pressure value are read, the difference between the hole position and the target hole position is measured, the deviation between the pull rod position and the predetermined position is calculated, the measured deviation is compared with the standard tolerance value, and the position deviation value is generated. Based on the position deviation value, it is determined whether the error exceeds the preset tolerance range. If it does, the pull rod displacement and cylinder pressure are adjusted to calibrate the fixture position and align the workpiece, complete the synchronous positioning verification, and generate a position deviation report.

5. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific steps for generating the cylinder correction parameters are as follows: Based on the position deviation report, the displacement of the tie rod and the cylinder thrust value are read, the offset of the tie rod is adjusted, and the temporal relationship between historical thrust and displacement data is analyzed through a long short-term memory network to make dynamic adjustments, optimize the synchronization of thrust and offset, and make the two match each other within the working range to generate a coordinated adjustment result. Based on the coordination adjustment results, the levelness of the fixture support surface and the penetration angle of the tie rod are detected, the actual levelness of the support surface is compared with the standard value, the angle of the fixture support surface and the tie rod are adjusted, it is determined whether the preset correction threshold is reached, and the fixture correction result is generated. Based on the fixture correction results, the current displacement value of the pull rod and the current output pressure of the cylinder are locked simultaneously, the corresponding action parameters of the two are frozen, and the signal of the pull rod stop position and the contact state of the positioning pin is compared to generate cylinder correction parameters.

6. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific execution process of the Long Short-Term Memory Network includes: based on the position deviation report, reading historical thrust and displacement data, analyzing the temporal relationship and making dynamic adjustments. First, the historical thrust data and the displacement of the tie rod are used as inputs to learn the long-term dependency relationship between thrust and displacement. According to the output adjustment parameters, the tie rod offset is optimized to make the thrust and offset synchronous and match each other within the working range, generating a coordinated adjustment result.

7. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific steps for generating the fixture positioning mark are as follows: Based on the cylinder correction parameters, the fixture movement path and clamping force parameters are read, the starting point and target position of the fixture movement are adjusted, the workpiece surface and the fixture clearance are compared, the fixture movement trajectory and clamping force output are calculated, and the fixture movement path is generated. Based on the movement path of the fixture, the pressure changes between the positioning hook and the edge of the hole are compared, the pressure feedback of the fixture when in contact is monitored, the deviation between the fixture pressure value and the target value is analyzed, the displacement when the fixture stops is confirmed, and a fixture positioning mark is generated.

8. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific steps for generating the riveting completion identifier are as follows: Based on the fixture positioning mark, a convolutional neural network is used to compare the penetration path of the positioning pin with the trigger position of the proximity switch in real time, determine the degree of overlap between the two, and ensure that the positioning pin completely penetrates the hole, thereby generating the degree of overlap of the penetration path. Based on the overlap of the penetration path, monitor the state of the proximity switch, compare the change amplitude of the signal, determine whether the signal is triggered, and generate a signal trigger state. Based on the signal triggering state, the validity of the signal is determined. If the validity condition is met, the riveting system is started, and the operation is carried out according to the riveting sequence to generate a riveting completion mark.

9. The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to claim 1, characterized in that, The specific execution process of the convolutional neural network includes: based on the fixture positioning mark, comparing the penetration path of the positioning pin with the trigger position of the proximity switch in real time; inputting the penetration path image of the positioning pin and the trigger position image of the proximity switch into the convolutional neural network; the network extracts image features through multi-layer convolution operations and calculates the overlap between the path and the trigger position; and then judging the overlap between the two based on the feature extraction results to ensure that the positioning pin completely penetrates the hole and generate the penetration path overlap.

10. An automatic hole-finding and positioning system for bushing riveting based on artificial intelligence, characterized in that, The automatic hole-finding and positioning method for bushing riveting based on artificial intelligence according to any one of claims 1-9, the system comprising: Workpiece alignment module: Based on a laser sensor, it obtains the distance between the pull rod and the workpiece plane, compares the boundary information of the workpiece height edge, updates the starting point of the pull rod displacement, calculates the pull rod displacement, and matches it with the cylinder stroke to generate workpiece alignment coordinates; Position deviation verification module: Based on the workpiece alignment coordinates, read the current extension length of the pull rod and the real-time pressure of the cylinder, measure the hole position reference difference and obtain the offset limit, determine whether the difference exceeds the limit range, if it exceeds the limit, perform synchronous positioning verification and generate a position deviation report; Cylinder correction module: Based on the position deviation report, adjust the coordination between the offset of the tie rod and the cylinder thrust. Use a long short-term memory network to analyze the temporal relationship between historical thrust and displacement data, make dynamic adjustments, optimize the synchronization of thrust and offset, so that the two match each other within the working range, and generate cylinder correction parameters. Fixture positioning module: Based on the cylinder correction parameters, adjust the movement path and clamping force threshold of the fixture, compare the pressure between the positioning hook and the edge of the hole, determine whether the fixture contact is stable, set the fixture stop position, and generate a fixture positioning mark; Riveting completion module: Based on the fixture positioning mark, a convolutional neural network is used to monitor the overlap between the positioning pin penetration path and the proximity switch position, confirm the signal triggering state and control the riveting system, determine whether the signal is valid and call the riveting task sequence, and generate a riveting completion mark.

Images