Cabinet sheet metal structural part riveting equipment control system and method based on positioning matching

Abstract

The application provides a cabinet sheet metal structural part riveting equipment control system and method based on positioning matching, and belongs to the technical field of intelligent manufacturing. The method comprises the following steps: performing RFID radio frequency identification and machine vision double collection and processing on cabinet sheet metal structural part workpieces transmitted by a production line, performing format verification and noise filtering on collected data, and generating workpiece identity and appearance feature fusion data; performing contour extraction and category labeling processing on the workpiece identity and appearance feature fusion data, and generating a multi-dimensional workpiece feature set; through a three-order closed-loop control architecture composed of visual coarse positioning and inertial navigation fine compensation, when a hole position is offset due to welding thermal deformation or a size deviation of a raw material, a local coordinate system can be reconstructed in real time and a pose deviation can be automatically compensated, the positioning accuracy is controlled within a range of ±0.05mm, the riveting qualified rate is greatly improved, and rivet skewing and plate cracking problems caused by forced riveting of a traditional rigid clamp are avoided.

Description

Technical Field

[0001] This invention proposes a control system and method for a cabinet sheet metal structural component riveting equipment based on positioning matching, belonging to the field of intelligent manufacturing technology. Background Technology

[0002] The press-fitting of sheet metal structural components in equipment cabinets is a critical process in communication equipment manufacturing, and its quality directly affects the structural strength and assembly accuracy of the cabinet. Currently, the control method of press-fitting equipment on the production line mainly relies on PLC preset parameters combined with manual teaching. Operators need to be taught the coordinates of each press-fitting point one by one, and the equipment performs the press-fitting action according to a fixed trajectory.

[0003] In the field of intelligent and precision control of riveting processes, the industry has made some technological explorations. Regarding visual positioning and position compensation, Chinese Patent Publication No. CN109550879B, entitled "A Vision-Based Sheet Metal Riveting Method," uses a photographic device to acquire images of the sheet metal parts to be riveted, calculates the deviation value between the hole position and the reference standard, and then a robot performs position compensation based on the deviation value to complete the riveting, achieving automatic correction of errors in the previous process. However, this method mainly solves the problem of static deviation calibration of the hole position and fails to dynamically adjust core process parameters such as riveting force and holding time during the riveting process. When the mechanical properties of the material batch change, fixed parameters are still used for riveting, making it difficult to ensure the consistency of riveting quality. Regarding visual compensation and cumulative tolerance elimination, Chinese Patent Publication No. CN119281946A, entitled "A High-Precision Riveting Equipment Based on Visual Compensation," obtains the cumulative tolerance between the workpiece to be riveted and the target position through visual recognition and performs position compensation in the X-axis, Y-axis, and angular directions, effectively ensuring riveting accuracy and workpiece pass rate. However, this technical solution focuses on the cumulative tolerance compensation of multi-axis transfer units, lacking the ability to perceive and adaptively adjust the deformation of sheet metal parts themselves (such as hole position displacement caused by welding heat deformation), and also does not involve online closed-loop control of riveting force. Regarding online compensation for stress deformation and data acquisition, Chinese patent publication number CN108856626B, entitled "A Method for Online Compensation of Stress Deformation for Servo Riveting Equipment," collects real-time values ​​of riveting force and push rod displacement data during the riveting process, calculates the stress deformation of the riveting tool, and compensates for the displacement of subsequent rivets, enabling high-precision riveting without pausing processing. However, this solution is mainly suitable for batch production scenarios of large-size thin-walled aerospace parts with a large number of riveting points. When the production line needs to frequently switch between different specifications of cabinet columns or beams, the recalibration and debugging time of process parameters is long, making it difficult to meet the rapid changeover requirements of multi-variety, small-batch customized orders. Furthermore, in Chinese patent publication number CN120030919A, entitled "An Intelligent Recommendation Method and System for Riveting Strength Based on Machine Learning," a machine learning model is used to recommend rivet types, base model numbers, and predict connection strength values ​​by inputting sheet material parameters and the number of overlap layers. This method provides intelligent decision-making tools for the early selection of riveting processes, but it stops at the process recommendation stage. It fails to directly integrate the recommended process parameters with production line equipment to form a closed-loop control, and it does not help solve real-time quality problems caused by sheet metal deformation, incoming material deviations, and other factors during actual riveting processes.

[0004] Although existing technologies have achieved preliminary results in visual positioning, tolerance compensation, and process parameter recommendation, none of the above solutions have effectively addressed the numerous shortcomings exposed by the traditional control method, which relies primarily on PLC preset parameters and manual teaching, in actual production. Firstly, the positioning process relies on rigid fixtures or specialized tooling. When sheet metal parts experience hole displacement due to welding heat deformation or incoming material dimensional deviations, the riveting head still moves along the original taught path, easily causing rivet misalignment or even sheet metal cracking. In some models, the column riveting pass rate is less than 85%. Secondly, process parameters such as riveting force and holding time are programmed during equipment debugging and cannot be adjusted according to the differences in mechanical properties between material batches. The yield strength difference between aluminum alloy 6061-T6 and 5052-H32 can reach over 40%, making it difficult for fixed parameter tables to accommodate the riveting quality requirements of different materials, leading to frequent over-pressure and under-pressure issues. Furthermore, when the production line needs to switch between different specifications of cabinet columns or beams, operators must re-teach the locations, and a single changeover typically takes more than 30 minutes, which cannot meet the rapid delivery requirements of multi-variety, small-batch customized orders. Therefore, there is an urgent need for a method that, on the basis of achieving vision-guided positioning, simultaneously solves deeper problems such as real-time compensation for sheet metal deformation, dynamic control of the riveting process, and rapid changeover for multiple varieties, thereby achieving intelligent perception and closed-loop control of the entire riveting process. Summary of the Invention

[0005] This invention provides a control system and method for a positioning matching-based riveting device for cabinet sheet metal structural components, in order to solve the problems mentioned in the background art above: This invention proposes a control method for a riveting device for cabinet sheet metal structural components based on positioning matching. The method includes: S1. Perform dual acquisition and processing of RFID radio frequency identification and machine vision on the cabinet sheet metal structural parts input from the production line. Perform format verification and noise filtering on the acquired data to generate workpiece identity and appearance feature fusion data. Perform contour extraction and category labeling on the workpiece identity and appearance feature fusion data to generate a multi-dimensional workpiece feature set. S2. Perform pose deviation analysis on the multi-dimensional workpiece feature set, map the deviation to the preset compensation strategy library, and generate pose compensation instruction data; extend the pose compensation instruction data by process association, supplement the riveting process constraints corresponding to the current workpiece material properties and structural part types, and generate extended process constraint data. S3. Perform instruction structure transformation on the extended process constraint data, transforming the riveting requirements in the form of process constraint conditions into structured control instructions that the riveting equipment can execute, and generating structured control instruction data; perform model matching analysis on the structured control instruction data, and select process model types that match the control instructions from the preset riveting process knowledge graph library, and generate process model matching data. S4. Perform parameter domain delimitation processing on the process model matching data. Combine the structural component type, hole distribution characteristics, and material property parameters to limit the boundary of the riveting force, holding time, and mold stroke parameter domains of each matching process model, and generate parameter domain limit data for each process model. Prioritize the parameter domain limit data of each process model to generate process model call priority sequence data. S5. Perform resource pre-allocation processing on the priority sequence data of the process model call, allocate independent control channels and execution timing for each process model, and generate process model resource pre-allocation data; prepare for parallel call of the process model resource pre-allocation data, and generate a multi-process model parallel call ready signal.

[0006] This invention proposes a control system for a cabinet sheet metal structural component riveting equipment based on positioning matching, the system comprising: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0007] The beneficial effects of this invention are as follows: Through a three-order closed-loop control architecture consisting of visual coarse positioning and inertial navigation fine compensation, when hole positions shift due to welding thermal deformation or incoming material dimensional deviations in sheet metal parts, the local coordinate system can be reconstructed in real time and the pose deviation automatically compensated. This keeps the positioning accuracy within ±0.05mm, significantly improving the riveting pass rate and avoiding the problems of rivet misalignment and sheet metal cracking caused by forced riveting with traditional rigid clamps. Simultaneously, the self-evolution mechanism of riveting parameters integrates process knowledge graphs and real-time force-displacement curve characteristics, dynamically matching the optimal riveting force, holding time, and die stroke for different material batches such as aluminum alloy 6061-T6 and 5052-H32. This effectively prevents over-pressure or false pressure caused by parameter rigidity, enhancing the equipment's adaptability to multi-material mixed-pressing conditions. In addition, the RFID-based automatic identification of workpiece models and the retrieval of corresponding positioning templates and compensation strategies have reduced changeover time from more than 30 minutes to less than 30 seconds. This not only meets the rapid delivery needs of multi-variety, small-batch customized orders, but also avoids long downtime caused by manual re-teaching, significantly reducing production line changeover costs. Attached Figure Description

[0008] Figure 1 This is a diagram illustrating the steps of the method described in this invention; Figure 2 This is a detailed flowchart of step S1 in this invention. Detailed Implementation

[0009] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0010] Example 1 One embodiment of the present invention, such as Figure 1 As shown, a control method for a cabinet sheet metal structural component riveting equipment based on positioning matching is disclosed, the method comprising: S1. Perform dual acquisition and processing of RFID radio frequency identification and machine vision on the cabinet sheet metal structural parts imported from the production line. Perform format verification and noise filtering on the acquired data to generate workpiece identity and appearance feature fusion data. Perform contour extraction and category labeling on the workpiece identity and appearance feature fusion data to extract structural part type labels, hole distribution features and material attribute parameters in the workpiece to generate a multi-dimensional workpiece feature set. S2. Perform pose deviation analysis on the multi-dimensional workpiece feature set, use the visual-inertial fusion navigation algorithm to calculate the deviation between the actual installation pose and the theoretical pose of the workpiece, map the deviation to the preset compensation strategy library, and generate pose compensation command data; extend the pose compensation command data by process association, supplement the riveting process constraints corresponding to the current workpiece material properties and structural component type, and generate extended process constraint data. S3. Perform instruction structure transformation on the extended process constraint data, transforming the riveting requirements in the form of process constraint conditions into structured control instructions that the riveting equipment can execute, and generating structured control instruction data; perform model matching analysis on the structured control instruction data, and select process model types that match the control instructions from the preset riveting process knowledge graph library, and generate process model matching data. S4. Perform parameter domain delimitation processing on the process model matching data. Combine the structural component type, hole distribution characteristics, and material property parameters to limit the boundary of the riveting force, holding time, and mold stroke parameter domains of each matching process model, and generate parameter domain limit data for each process model. Prioritize the parameter domain limit data of each process model to generate process model call priority sequence data. S5. Perform resource pre-allocation processing on the priority sequence data of the process model call, allocate independent control channels and execution timing for each process model, and generate process model resource pre-allocation data; prepare for parallel call of the process model resource pre-allocation data, and generate a multi-process model parallel call ready signal.

[0011] The working principle and effects of the above technical solution are as follows: Through a dual acquisition and processing mode of RFID radio frequency identification and machine vision, the workpiece features of the cabinet sheet metal structural components can be comprehensively collected, improving the comprehensiveness and accuracy of workpiece feature recognition and reducing the feature acquisition omissions caused by a single acquisition method. Through pose deviation analysis and process constraint extension processing, the riveting processing requirements of workpieces with different materials and structures can be accurately adapted, effectively improving the overall accuracy of riveting processing. Relying on the process knowledge graph to complete model matching and parameter domain delineation, the appropriate processing technology can be accurately matched, refining the equipment operating parameter range and reducing processing defects caused by riveting parameter adaptation deviations. Through process model priority sorting and a parallel calling mode of pre-allocated hardware and software resources, the processing flow and equipment resources can be rationally arranged, enhancing the stability and continuity of equipment operation, reducing equipment resource conflicts and process errors, and avoiding processing stalls and parameter errors that occur when multiple processes are running simultaneously. This ensures the consistency of batch sheet metal structural component riveting processing and effectively improves overall processing efficiency.

[0012] One embodiment of the present invention, such as Figure 2 As shown, S1 includes: S11. Receive the original acquisition signal of the cabinet sheet metal structure transmitted from the production line conveyor end, and simultaneously start the RFID radio frequency reading module and the industrial camera vision acquisition module to complete synchronous sampling. Acquire the information stored in the workpiece's built-in radio frequency chip and the full-area image information of the workpiece surface. After the acquisition is completed, perform data bit format verification and image salt-and-pepper noise Gaussian noise layer filtering operation, and output the workpiece identity code data stream and the workpiece appearance grayscale image data stream fused together to form the workpiece identity and appearance feature fusion data. S12. The edge image operator is called to perform edge contour pixel-by-pixel extraction operation on the image components in the fusion data of workpiece identity and appearance features. Simultaneously, the radio frequency analysis program is called to read the preset classification identifier in the identity code to complete the workpiece category marking operation. During the operation, the three types of physical parameters in the image are separated: the hole coordinate arrangement value, the plate thickness and hardness value, and the coating material composition value. All operation outputs are summarized and assembled to form a multi-dimensional workpiece feature set. S13. Read the structural component type label values, hole position distribution coordinate array, and material attribute parameter array stored in the multidimensional workpiece feature set. Call the feature normalization operation program to perform unified dimension scaling conversion on the three types of feature values. After conversion, perform continuous interval interpolation filling on discrete feature points. After filling, remove abnormal feature points whose values ​​fluctuate beyond the reasonable fluctuation range and output the purified standardized workpiece feature array. S14. Perform feature weight assignment operation on the standardized workpiece feature array. According to the influence of different features on the processing accuracy under sheet metal riveting conditions, allocate gradient weight coefficients. After the weight assignment is completed, perform feature vector space mapping operation to convert discrete numerical features into high-dimensional feature vector matrix and output weighted feature vector matrix. S15. Perform feature redundancy compression operation on the weighted feature vector matrix, use principal component decomposition to remove linearly related redundant vector components in the matrix, retain the core feature vector group that has a numerical impact on the riveting process after compression, and output the final simplified and finalized multidimensional workpiece feature set.

[0013] The working principle and effects of the above technical solution are as follows: By using a simultaneous sampling method combining radio frequency identification and machine vision, the identity information and appearance features of the cabinet sheet metal structural components can be completely acquired, reducing the information loss problems caused by a single acquisition method. Multi-level noise filtering can remove invalid interference information from the acquired data, improving the purity of the original data. Feature normalization and interpolation filling can unify the numerical standards of various workpiece features, eliminating calculation deviations caused by different parameter dimensions. The removal of abnormal feature points can avoid the adverse effects of abnormal data on subsequent processing operations. Feature weight assignment and vector space mapping can highlight the influence ratio of key processing features, making feature recognition more closely aligned with the actual working conditions of riveting. Feature redundancy compression processing can simplify the data volume, improve the data processing speed of subsequent processes, ensuring both the accuracy of workpiece feature recognition and the overall efficiency of data processing.

[0014] In one embodiment of the present invention, S14 includes: Read the structural component type features, hole distribution features, and material attribute core feature data from the final multidimensional workpiece feature set. Combine these with the attribute parameters of each candidate process model in the riveting process model matching data to perform multidimensional parameter adaptation comparison calculations. This will complete the preliminary definition of the processing adaptation range of each process model and generate the initial range delineation data for the process model. The initial range definition data of the process model is used to perform processing condition adaptation correction calculation. Combined with the equipment operating condition threshold of sheet metal structural parts riveting processing, the processing adaptation range of each process model is calibrated in the whole domain. Invalid intervals that exceed the operating limits of equipment hardware are eliminated, and the precise range limit data of the process model is generated. Cross-validation calculations are performed on the precise range limit data of the process model to compare the processing coverage of different process models, identify overlapping and blank areas in the processing intervals of multiple models, fill in the gaps in the intervals, segment and optimize the overlapping areas of the intervals, and generate conflict-free search range limit data for each process model. Read the core indicators of interval adaptation accuracy, processing energy consumption, and operational stability of the search range limit data of each process model, carry out multi-dimensional quantitative scoring calculation, perform comprehensive performance quantitative rating of all matching process models, and generate process model quantitative scoring data; The process model quantitative scoring data is sorted hierarchically, and all process models are arranged in order according to their quantitative scoring values. The order of calling each model is marked synchronously, generating an ordered sequence of process model calling priority data.

[0015] The working principle and effects of the above technical solution are as follows: Through multi-dimensional parameter adaptation and comparison, the processing range of the process model can be defined by combining various core features of the workpiece, improving the fit of the process adaptation. Through equipment operating condition threshold calibration, invalid processing ranges that do not meet hardware operating conditions can be screened out, reducing processing failures caused by equipment exceeding limits. The cross-validation of multi-model processing ranges can properly fill in blank areas of the processing range, streamline the processing logic of overlapping ranges, and avoid processing omissions or process disorder during the collaborative operation of multiple process models. Through multi-dimensional index quantification and scoring, the actual operating performance of each process model can be objectively evaluated, making the model selection criteria more closely aligned with actual processing scenarios. The hierarchical sorting method can regulate the process call order, improve the orderliness of the overall processing flow, ensure the overall accuracy of sheet metal riveting processing, stabilize equipment operation, and effectively improve the processing regularity of batch workpieces.

[0016] In one embodiment of the present invention, S15 includes: Read the priority sequence data of the process model call, traverse the processing and operation parameters corresponding to all process model entries in the sequence, call the hardware resource pool and software operation resource pool of the equipment control system, perform full-domain detection and statistics on the idle and occupied status of various resources, and generate equipment resource status statistics. Perform resource matching and allocation calculations on equipment resource status statistics, and combine the computing power consumption, control channel occupation, and cache space occupation requirements of each process model to match suitable system operating resources for process models of different priorities, generating preliminary resource allocation mapping data; The initial resource allocation mapping data is isolated by resource isolation processing. The computing units, control channels, and memory cache areas corresponding to each process model are independently divided to prevent resource cross-occupancy and data interference during the operation of multiple models, and to generate isolated model resource allocation data. The isolated model resource allocation data is processed for runtime environment configuration. Each process model that has completed resource allocation is configured with exclusive runtime parameters, process scheduling rules and data storage paths to generate complete process model resource pre-allocation data. Parallel adaptation verification and startup preprocessing are performed on the pre-allocated resource data of the process model. The process startup timing and running adaptation parameters of each model are uniformly calibrated, the pre-initialization configuration for parallel operation of multiple models is completed, and a multi-process model parallel call ready signal is generated.

[0017] The working principle and effects of the above technical solution are as follows: By detecting the status of the equipment's hardware and software resources across the entire domain, the idle and occupied status of system resources can be accurately grasped, allowing resource allocation to match the actual operating needs of each process model. Differentiated resource allocation methods can match the operating consumption of process models with different priorities, improving the overall utilization rate of system resources and reducing resource waste caused by idle resources. Independent resource isolation processing can prevent resource contention and data interference during the operation of different process models, avoiding program errors and data anomalies when multiple models are running simultaneously. Dedicated operating environment configuration can standardize the operating rules and data storage logic of each model, making the equipment operation process more organized. Parallel preprocessing calibration can unify the startup and operating parameters of multiple models, ensuring the stability of simultaneous operation of multiple process models and significantly improving the overall operating efficiency and processing continuity of the sheet metal riveting process.

[0018] In one embodiment of the present invention, S2 includes: S21. Retrieve the standard theoretical pose coordinate matrix of the same type of structural component stored in the device storage unit. Input the actual spatial coordinate posture matrix of the workpiece obtained by extracting the multi-dimensional workpiece feature set and the theoretical pose coordinate matrix into the visual-inertial fusion navigation calculation unit. The unit performs the calculation of the three-axis translation deviation value and the three-axis rotation angle deviation value axis-by-axis difference value, and summarizes all deviation values ​​to generate a global pose deviation value matrix. S22. Retrieve the complete dataset of the compensation strategy library stored locally on the device, use the deviation amplitude of each axis in the global pose deviation numerical matrix as the search matching factor, traverse all compensation strategy samples in the library to complete the numerical similarity comparison, and generate the initial pose compensation command data by combining the displacement compensation amount and angle correction amount corresponding to the sample with the highest matching degree. S23. Parse the material hardness and thickness parameters and structural component classification label parameters stored in the multi-dimensional workpiece feature set, retrieve the process threshold range array of the corresponding material and structural category in the riveting process constraint database, concatenate the threshold range array to the end field of the initial pose compensation instruction data, complete the process parameter supplementation of the instruction content, and output the supplemented intermediate extended process data. S24. Perform threshold over-limit screening calculation on intermediate extended process data, compare the added process threshold with the equipment hardware limit operating parameters, replace the threshold value that exceeds the hardware bearing range with the hardware critical safe operating value, and output the calibration extended process constraint data after correcting the value. S25. Perform timing field filling operation on the calibration-type extended process constraint data, embed the workpiece processing timing sequence number and equipment station number identifier segment in the data header, and generate complete extended process constraint data with station timing mark after filling.

[0019] The working principle and effects of the above technical solution are as follows: By calculating the three-axis omnidirectional pose deviation, the position and angle offset of the sheet metal workpiece can be accurately captured, improving the detail and accuracy of workpiece pose deviation recognition. Through a deviation value matching compensation strategy, a correction scheme adapted to the current workpiece state can be generated, improving the fit of the workpiece pose correction. By supplementing process parameters with workpiece material and structural properties, the pose compensation operation can be made to fit the riveting processing standards of different workpieces, adapting to diverse workpiece processing scenarios. Through process threshold over-limit screening and correction, situations where process parameters exceed the equipment's carrying capacity can be avoided, reducing equipment wear and processing errors caused by hardware overload. Adding station sequence markers can regulate the workpiece processing sequence, avoiding process errors during batch processing. This ensures both the accuracy of workpiece pose correction and the long-term stable operation of the equipment, effectively improving the overall yield rate of batch cabinet sheet metal riveting processing.

[0020] In one embodiment of the present invention, S3 includes: S31. The extended process constraint data is divided into three categories of segmented data: pose correction value segment, material process threshold segment, and structural component category identification segment. The text-to-machine code compiler is called to convert the natural parameter segment format into binary instruction code segments that the device controller can recognize. Multiple binary code segments are spliced ​​together to generate basic structured control instruction data. S32. Read the riveting processing requirement information carried by the binary code in the basic structured control instruction data, traverse all node entities in the locally deployed riveting process knowledge graph library, calculate the feature cosine similarity between the instruction requirement and each process node entity in the graph, mark the matching level of each node from high to low similarity value, and output the process model matching preliminary screening data with matching level label. S33. Perform graph association node tracing operation on the initial screening data of process model matching, extract upstream and downstream auxiliary process sub-model entities of matching process nodes, include the auxiliary sub-model entities into the matching set, expand the original matching model range, and output the expanded process model matching set data. S34. Perform model conflict verification operation on the expanded process model matching set data, compare the riveting execution axis occupation requirements of different models in the set, mark the pair model combinations with axis resource contention conflicts, and output the process model matching data with marked conflict points. S35. Perform conflict weight assignment calculation on the process model matching data of the marked conflict points, assign priority weight to the conflict model according to the workpiece processing accuracy priority, embed the weight value into the corresponding model field of the matching data, and generate complete process model matching data.

[0021] The working principle and effects of the above technical solution are as follows: By compiling process constraint data into standardized equipment control instructions, equipment identification failures caused by inconsistent parameter formats can be eliminated, improving the equipment's efficiency in identifying processing requirements. Relying on knowledge graph similarity comparison to screen process models, the current workpiece's riveting processing requirements can be accurately matched, improving the model's fit. By tracing and associating process nodes to expand the model set, auxiliary process requirements for the main model's processing stages can be supplemented, reducing the loss of processing details caused by adapting to a single model. Model resource conflict verification can proactively identify execution axis occupancy conflicts, avoiding process stalls and operational errors caused by simultaneous calls to multiple models. Configuring conflict weights through precision priority can reasonably handle model resource competition issues, ensuring the operational stability of core processing stages. The entire process not only refines the process adaptation dimensions of riveting processing but also stabilizes equipment operation, effectively improving the overall accuracy and process standardization of sheet metal structural component riveting processing.

[0022] In one embodiment of the present invention, S34 includes: Read the expanded process model matching set data, extract the equipment execution component occupancy information and process execution timing information corresponding to each process model in the set in batches, and summarize and integrate them to form a model resource occupancy feature dataset. Perform a resource comparison operation on the model resource occupancy feature dataset, and statistically analyze the overlapping occupancy of different process models on the riveting execution shaft, power output component, and sensor acquisition component to generate model resource overlap comparison data. The model resource overlap ratio is used to perform conflict identification calculations on the data, distinguish between compatible parallel resource occupation and mutually exclusive resource occupation, filter out process model combinations that cannot run synchronously in the same period, and generate a model conflict combination dataset. The conflict point marking process is performed on the model conflict combination dataset. The resource conflict type and the process node where the conflict occurred are entered into the corresponding model data field to generate process model verification data with complete conflict information. Redundant model elimination preprocessing is performed on the process model verification data with complete conflict information to screen out invalid matching models with excessive conflict degree and low processing contribution, and process model matching data with marked conflict points is output.

[0023] The working principle and effects of the above technical solution are as follows: By comprehensively collecting component occupancy and process sequence information of each process model, the resource usage of model operation can be fully analyzed, eliminating information omissions during resource investigation. The processing method of comparing resources from the same source can accurately identify overlapping equipment component occupancy during the operation of multiple models, improving the detail of resource conflict investigation. Distinguishing between compatible and exclusive resources allows for precise location of model combinations that cannot run synchronously, reducing computational resource consumption caused by invalid conflict determinations. Completely marking various conflict points and process nodes enables subsequent model scheduling to specifically avoid resource conflicts, preventing process stagnation and processing deviations caused by equipment component contention. Eliminating low-value redundant models simplifies the overall process matching system, ensuring the smooth progress of the riveting process while improving the overall stability of equipment operation and process scheduling efficiency.

[0024] In one embodiment of the present invention, step S4 includes: S41. Read each independent process model entity in the process model matching data line by line, retrieve the three sets of core parameters in the multi-dimensional workpiece feature set: structural part type label, hole position coordinate distribution array, and material hardness and thickness. Use the three sets of parameters as boundary constraint inputs, and substitute them into the riveting force limit solution formula, the holding time adaptation formula, and the mold extension stroke constraint formula to complete the numerical upper and lower limit solution. After the single model solution is completed, output the single model's exclusive parameter upper and lower limit array. S42. Integrate the upper and lower limit arrays of parameters corresponding to all matching process models, summarize them to form the set of original parameter domain data of multiple models, call the numerical interval intersection operation program, calculate the overlapping interval range of three types of parameters, namely riveting force, holding time and mold stroke, between different models, mark the opposite parameter items with no overlapping intervals, and output the set of parameter domain data of multiple models with overlapping marks. S43. Perform a safe interval shrinkage operation on the multi-model parameter domain set data with overlapping markers. Select the maximum value of the lower limit of each model parameter as the unified processing lower limit and the minimum value of the upper limit of each model parameter as the unified processing upper limit. The overlapping interval parameters retain the original interval range. After the shrinkage operation is completed, the parameter domain limit initial version data of each process model is generated. S44. Retrieve the response rate parameters of the hydraulic unit of the equipment and the response lag parameters of the mechanical transmission of the mold. Substitute the two types of hardware response parameters into the parameter domain correction formula, perform lag compensation fine-tuning on all upper and lower limit values ​​within the initial data of the parameter domain, and output the parameter domain limit data of each process model after fine-tuning. S45. Perform a comprehensive scoring operation on the parameter domain limit data of each modified process model. Calculate the comprehensive score of each process model based on three indicators: parameter range fit, hardware load consumption, and estimated processing accuracy. Sort the models from high to low scores and output the process model call priority sequence data.

[0025] The working principle and effects of the above technical solution are as follows: By combining the core features of the workpiece to solve the limit range of various processing parameters, the riveting processing parameters can be made to fit the actual processing needs of different sheet metal workpieces, improving the fit of parameter configuration. Comparison calculations of multiple model parameter ranges can clearly distinguish between overlapping and opposing parameter areas, reducing parameter compatibility conflicts that occur in multi-process collaborative processing. Safety range shrinkage processing can unify the processing range of various opposing parameters, eliminating processing over-limit problems caused by unreasonable parameter ranges. Fine-tuning parameter values ​​based on the equipment hardware response characteristics can offset processing errors caused by mechanical lag, making the actual processing parameters more accurate. A multi-index comprehensive scoring and ranking method can select the optimal process operation scheme, avoiding processing defects caused by unreasonable process calls. The entire process can stabilize the riveting processing quality of various workpieces, reduce equipment operating losses, and effectively improve the stability and finished product qualification rate of batch processing of sheet metal structural parts.

[0026] In one embodiment of the present invention, S44 includes: Read the initial data of the parameter domain limit of each process model, extract all extreme values ​​of the riveting force, holding time and mold stroke in the initial data in batches, and summarize and integrate them to form the original corrected dataset of model parameters; The dynamic response rate of the hydraulic drive unit of the equipment and the mechanical hysteresis of the mold transmission structure are collected. The parameter fluctuation offset under continuous operation of the equipment is statistically analyzed to generate hardware response characteristic benchmark data. By substituting the hardware response characteristic baseline data into the parameter domain dynamic correction algorithm, synchronous offset calibration is performed on the interval extreme values ​​in the original correction dataset of model parameters to offset the parameter deviations caused by mechanical transmission and hydraulic response, and intermediate parameter calibration data is generated. Extreme value boundary verification calculations are performed on intermediate parameter calibration data to screen whether the parameter values ​​after calibration exceed the safe operating threshold of the equipment hardware. Critical fall-off adjustments are made for values ​​that exceed the limit to generate compliant parameter calibration data. The calibration data of compliant parameters are normalized to unify the range accuracy standards of the three types of processing parameters of each process model. After normalization, the parameter domain limit data of each process model is output as a correction.

[0027] The working principle and effects of the above technical solution are as follows: By extracting the extreme values ​​of the parameter range of the process model in batches, the processing parameters to be corrected can be completely gathered, ensuring comprehensive coverage of parameter correction work and preventing processing deviations caused by the omission of local parameters. Combining the calibration parameters with the actual response state of the equipment's hydraulic and mechanical transmission can offset the processing errors caused by the inherent lag during equipment operation, ensuring consistency between the set parameters and the actual operating state. Through extreme value boundary verification, the compliance of the calibrated parameters can be checked, preventing parameter values ​​from exceeding the equipment's safe operating standards and reducing malfunctions and processing defects caused by hardware overload. Unifying the accuracy standards of various processing parameters ensures that the parameter outputs of different process models remain consistent and avoids inconsistent parameters in batch processing. This entire processing method not only improves the accuracy of riveting parameter matching but also maintains long-term stable equipment operation, effectively improving the overall quality and consistency of the riveting processing of cabinet sheet metal structural components.

[0028] In one embodiment of the present invention, step S5 includes: S51. Read all process model entries sorted in the priority sequence data of the process model call, split the three independent hardware channel resource pools of the riveting execution axis control channel, pressure sensing acquisition channel and displacement feedback acquisition channel in the equipment main control system, and allocate dedicated control channel and acquisition channel numbers in order of model priority, and write the channel number into the corresponding model storage field. S52. Retrieve the timing scheduling calculation unit, and based on the estimated execution time of each process model's processing action, arrange the timing arrangement arrays for the start time, duration, and end time of the riveting action of different models. Bind the timing arrangement arrays to the corresponding model resource allocation entries, and generate the initial data for the pre-allocation of process model resources that binds the channels and timing. S53. Perform timing gap verification calculation on the initial data of pre-allocation of process model resources, calculate the interval between the end point of adjacent priority model actions and the start point of the next model, and automatically extend the gap to the reset critical time for entries whose interval is less than the minimum time consumption of equipment hardware reset. After the time correction, output the timing calibration type process model resource pre-allocation data. S54. Call the equipment memory partition allocation program, divide each process model in the time-series calibration process model resource pre-allocation data into an independent memory read / write partition, match the partition size with the memory capacity required for storing parameters and caching real-time data during model runtime, write the partition address into the corresponding model resource field, and output the process model resource pre-allocation data matching the memory partition. S55. Perform parallel trigger signal compilation and operation on the pre-allocated data of the process model resources associated with the memory partition, generate an independent level trigger signal code for each process model with allocated resources, and package and integrate all signal codes to generate a unified multi-process model parallel call ready signal.

[0029] The working principle and effects of the above technical solution are as follows: By allocating dedicated hardware channels according to priority, the hardware resource usage permissions of various process models can be standardized, reducing channel contention issues during multi-model operation. Scientifically arranging the processing sequence of each process model can streamline the overall processing flow, making the operation rhythm of each process more regular. By verifying and correcting the interval length through timing gaps, the basic reset requirements of the equipment hardware can be met, avoiding equipment response anomalies and processing disorder caused by excessively rapid process transitions. Independently partitioning dedicated memory can isolate the running data of each model, eliminating data cross-interference and memory resource crowding, and improving the stability of data read and write operations. Unified compilation of parallel trigger signals can achieve synchronous startup and operation of multiple process models, which simplifies equipment scheduling logic, improves overall processing efficiency, maintains process stability and processing consistency in batch sheet metal riveting operations, and reduces operational losses caused by frequent equipment start-ups and shutdowns.

[0030] In one embodiment of the present invention, S54 includes: Read the pre-allocated resource data of the time-series calibration process model, extract the parameter storage volume, real-time computing cache volume, and interactive data read / write volume during the operation of each process model, and summarize and statistically form a dataset of memory resource requirements for each model; The system memory partition allocation program is invoked, the global memory storage space of the device is traversed, the capacity and storage address information of free contiguous memory blocks are counted, and the system free memory resource statistics are generated. The dataset of memory resource requirements for each model is matched with the statistical data of system free memory resources to match independent and continuous memory blocks with appropriate capacity for a single process model, and to generate preliminary matching data for memory partitioning. The initial matching data of the memory partition is partitioned and isolated. Independent memory read and write permissions and data isolation boundaries are set to prevent cross-read and write of data and memory resource contention caused by different process models during operation, and generate isolated memory partition configuration data. Enter the capacity parameters and storage address parameters of the memory partition corresponding to each model into the corresponding model resource information field to complete the binding and adaptation of memory resources and process models, and output the pre-allocation data of process model resources corresponding to the memory partition.

[0031] The working principle and effects of the above technical solution are as follows: By accurately calculating the memory usage requirements of each process model, memory resources can be matched to actual operational consumption, reducing idle and wasted system memory resources. A comprehensive investigation of idle memory blocks in the equipment allows for precise understanding of the system's memory distribution, making memory matching more closely aligned with the actual storage state of the equipment. Refined memory adaptation and matching can allocate appropriately sized contiguous memory space to each process model, avoiding program lag caused by insufficient memory. Strict memory isolation and access control can block data exchange interference between different process models, preventing program malfunctions caused by memory resource contention. Fixed binding of memory parameters to process models can stabilize equipment operation configuration, ensuring data security and operational stability when multiple process models work in parallel, while also improving the overall memory resource utilization efficiency of the equipment, making the multi-process collaborative operation of the sheet metal riveting equipment smoother and more stable.

[0032] One embodiment of the present invention provides a control system for a positioning matching-based riveting device for cabinet sheet metal structural components, the system comprising: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0033] Example 2 This embodiment further refines the pose compensation method for each riveting hole, including: S1. Obtain workpiece identity and hole position feature data; Following the processing method in step S1 of Example 1, the cabinet sheet metal structural components imported from the production line are subjected to dual acquisition processing using RFID radio frequency identification and machine vision. The RFID radio frequency reading module and the industrial camera vision acquisition module are activated simultaneously to read the structural component model, material type, sheet thickness, process batch, and theoretical hole position template number stored in the workpiece's built-in radio frequency chip, and to acquire full-area image information of the workpiece surface.

[0034] The system performs data format verification, salt-and-pepper noise filtering, Gaussian noise filtering, and edge contour extraction on the full-area image information. It identifies the edge contour, center coordinates, diameter, edge integrity, and grayscale variation characteristics of each riveting hole. Based on the theoretical hole position template numbers stored in the RFID chip, each identified riveting hole is sequentially marked, establishing a correspondence between each actual riveting hole and a theoretical riveting hole in the theoretical hole position template.

[0035] For cabinet uprights, beams, or mounting plates with multiple rivet holes, establish the adjacency relationship of each rivet hole according to the hole arrangement direction and hole spacing recorded in the theoretical hole template. For rivet holes arranged along the length direction, rivet holes that are adjacent to each other and have continuous theoretical hole spacing are determined as adjacent hole positions; for rivet holes distributed in a rectangular array, the lateral nearest neighbor hole, longitudinal nearest neighbor hole, and diagonal nearest neighbor hole are respectively recorded into the adjacent hole position set of the corresponding rivet hole.

[0036] Based on the multi-dimensional workpiece feature set formed in Example 1, the theoretical hole position coordinates, actual hole position coordinates, hole edge contour integrity, hole shape matching degree, visual matching confidence degree, and hole position adjacency relationship of each riveting hole are further recorded into the corresponding hole position field to form hole feature data that can be independently called according to the hole position number.

[0037] S2. Determine the overall rigid pose compensation benchmark and the local residuals of each hole position; According to the processing method of step S2 in Embodiment 1, the standard theoretical pose coordinate matrix of the same type of structural component stored in the device storage unit is retrieved, and the actual spatial coordinate pose matrix of the workpiece in the multi-dimensional workpiece feature set and the standard theoretical pose coordinate matrix are input into the visual-inertial fusion navigation calculation unit.

[0038] In this embodiment, multiple riveting holes with complete edge contours, high visual matching confidence, and complete adjacent hole relationships are preferentially selected as reference holes for overall pose calculation. The visual-inertial fusion navigation calculation unit calculates the overall translation and overall rotation of the current cabinet sheet metal structure based on the actual spatial coordinates and corresponding theoretical spatial coordinates of the reference holes, combined with the translational and rotational changes reflected by the inertial navigation data during the workpiece's entry into the riveting station.

[0039] In calculating the overall translation and rotation, holes that significantly deviate from the trend of most reference holes are temporarily excluded from the final solution to avoid interference from locally warped or abnormally identified holes with the global rigid pose results. Based on the calculated overall translation and rotation, a global rigid pose transformation is performed on all theoretical holes to obtain the global predicted coordinates of each theoretical hole in the current riveting position.

[0040] Subsequently, the actual center coordinates of each riveting hole are compared with the corresponding global predicted coordinates to obtain the local residuals remaining after eliminating the overall translational and angular deviations of each riveting hole. The local residuals include the residual deviation direction and the residual deviation magnitude. The residual deviation direction points from the global predicted coordinates to the actual hole center coordinates, and the residual deviation magnitude indicates the degree of positional deviation that cannot be eliminated by overall rigid pose compensation.

[0041] The displacement compensation amount and angle correction amount matched from the compensation strategy library in Example 1 are used as the initial global compensation amount for each riveting hole, and the local residual of each riveting hole is used as the basis data for subsequent judgment of local non-rigid deformation and abnormal identification points.

[0042] S3. Match the riveting process model and limit the allowable compensation range for each hole position; Following the processing method in step S3 of Embodiment 1, the extended process constraint data is processed by instruction structure conversion, converting the current workpiece's pose correction data, material process threshold data, and structural component category identification data into structured control instructions that the equipment controller can recognize.

[0043] Using structured control instructions, the knowledge graph of the riveting process is traversed to select process models that match the structural type, material hardness, plate thickness, hole distribution, and riveting specifications of the current cabinet sheet metal structural components. The process model matching data determined in Example 1 is then associated with the hole feature data formed in this example according to the hole position number.

[0044] Following the processing method of step S4 in Example 1, the riveting force, holding time, and mold stroke parameter range of each matching process model are limited by combining the structural component type, hole distribution characteristics, and material property parameters. The parameter boundary is then corrected based on the dynamic response rate of the hydraulic drive unit and the mechanical lag of the mold transmission structure.

[0045] Based on the existing parameter domain limits, the allowable position compensation range for each riveting hole is further determined according to the structural space around each riveting hole. For riveting holes near bending edges, welds, reinforcing ribs, or adjacent protruding structures, the allowable lateral position compensation range is reduced to avoid interference between the riveting head or mold and adjacent structures after the riveting execution axis completes local compensation; for riveting holes located in the middle of a flat plate surface with sufficient surrounding space, the position compensation range determined in Example 1 based on the equipment hardware limits is maintained.

[0046] Therefore, a riveting parameter domain and a position compensation safety boundary corresponding to the hole position number are formed for each riveting hole, which are called in step S6 when generating the final compensation amount.

[0047] S4. Complete the pre-allocation of resources for the hole splitting process model; According to the processing method of step S5 in Example 1, read the priority sequence data of the process model call, allocate the riveting execution axis control channel, pressure sensing acquisition channel and displacement feedback acquisition channel to each process model, and arrange the start time, duration and end time according to the expected action execution time of each process model.

[0048] Based on the pre-allocated resource data of the process model formed in Example 1, a corresponding data storage field is divided for each riveting hole to store the theoretical hole position coordinates, actual hole position coordinates, global predicted coordinates, local residuals, visual matching confidence level, degree of consistency of deformation of adjacent holes, degree of continuity of inertial navigation pose, and actual feedback deviation after riveting.

[0049] To avoid the added hole compensation data in this embodiment affecting the resource isolation relationship between the process models in Embodiment 1, the hole feature data of each riveting hole is still stored in the independent memory read / write partition allocated to the corresponding process model, and is distinguished by hole position number, without changing the original allocation method of computing power unit, control channel and memory cache area in Embodiment 1.

[0050] After configuring the control channel, execution timing, and orifice data fields, a multi-process model parallel call ready signal is generated. This ready signal serves as the trigger condition for starting orifice compensation decision-making in step S6.

[0051] S5. Collect the actual feedback deviation of the completed riveting points; After completing the initial classification and compensation priority determination of each riveting hole in step S6, the riveting holes with higher visual matching confidence, smaller local residuals, and those judged as overall rigid pose deviations are selected as the reference riveting holes.

[0052] The riveting actuator moves to the target riveting position according to the initial global compensation of the reference riveting hole, and completes the riveting according to the riveting force, holding time, and die stroke determined in Example 1. During the riveting process, the actual contact position of the riveting actuator, the actual downward endpoint of the die, and the reset position after unloading are recorded using the displacement feedback acquisition channel. The pressure sensor acquisition channel records the pressure establishment point and pressure change state when the riveting head contacts the workpiece.

[0053] The actual contact position of the riveting actuator shaft is compared with the target contact position of the reference riveting hole to obtain the actual feedback deviation of the completed riveting point. This actual feedback deviation represents the positional difference that remains after the initial global compensation amount formed in Embodiment 1 is applied to the actual workpiece.

[0054] For multiple riveting points that have been completed within the same adjacent group of holes, the direction and amplitude of each actual feedback deviation are further compared. When the actual feedback deviation directions of multiple adjacent completed riveting points are similar, and the deviation amplitude continuously increases, decreases, or fluctuates along the hole arrangement direction, it indicates that the deviation in this area has continuous deformation characteristics. When the actual feedback deviation direction of a completed riveting point is significantly opposite to that of its adjacent riveting points, or when its deviation amplitude undergoes an isolated jump, it indicates that the visual coordinates formed by the hole before riveting may be affected by hole edge defects, reflections, or obstructions.

[0055] The actual feedback deviation of the completed riveting points is written back to the corresponding sub-hole feature data according to the hole position number, which is used to update the local non-rigid deformation judgment results of adjacent holes that have not yet been riveted.

[0056] S6. Classify each riveting hole and generate compensation decisions; Step S6 includes steps S61 to S63.

[0057] S61. Determine the degree of anomaly identification for each riveting hole to be pressed; For any riveting hole to be pressed, read the local residual of the hole, the local residual of the adjacent holes, and the visual matching confidence level of the hole.

[0058] First, the degree of spatial residual is determined based on the local residual amplitude of the rivet hole to be pressed. To avoid the influence of dimensional differences between sheet metal structural parts of different specifications on the judgment results, the local residual amplitude of the hole position is compared with the maximum allowable position compensation amount determined in step S3, so that the degree of spatial residual for workpieces of different specifications is converted to the same evaluation range. The closer the local residual is to or exceeds the maximum allowable position compensation amount, the higher the degree of spatial residual; when the local residual is within the normal positioning error range of the equipment, the lower the degree of spatial residual.

[0059] Secondly, the degree of deformation consistency between adjacent holes is determined based on the changes in the direction and amplitude of the local residuals between the hole to be riveted and its adjacent holes. When the direction of the local residuals of the hole to be riveted and multiple adjacent holes is similar, and the amplitude of the local residuals continuously increases, continuously decreases, or remains within the same fluctuation range along the hole arrangement direction, the degree of deformation consistency between adjacent holes is relatively high. When the direction of the local residuals of the hole to be riveted is opposite to that of adjacent holes, or when its local residual amplitude undergoes an isolated jump relative to adjacent holes, the degree of deformation consistency between adjacent holes is relatively low.

[0060] The consistency of deformation between adjacent holes is also constrained by the amplitude of spatial residuals. When the local residuals of multiple adjacent holes after overall rigid pose compensation are all within the normal positioning error range of the equipment, even if their residual directions are similar, the area will not be identified as a local warping area, so as to avoid misjudging normal measurement fluctuations as local non-rigid deformation of the workpiece.

[0061] Next, the confidence level of visual matching is determined based on the completeness of the hole edge outline, the grayscale gradient of the hole edge, the roundness of the hole shape, the occlusion ratio of the hole edge, and the degree of matching between the actual hole position and the theoretical hole position template. A high confidence level of visual matching is achieved when the hole edge outline is complete, the grayscale change between the hole edge and the board surface is clear, the hole shape matches the theoretical template, and there is no obvious occlusion. A low confidence level of visual matching is achieved when there are gaps in the hole edge, solder slag occlusion, strong reflection of the plating, outline adhesion, or obvious distortion of the hole shape.

[0062] The degree of anomaly identification for the riveting hole is obtained by comprehensively considering the degree of spatial residual, the degree of inconsistency in deformation of adjacent holes, and the degree of visual matching uncertainty. Among these, the degree of spatial residual and the degree of inconsistency in deformation of adjacent holes are used as the main judgment criteria, while the degree of visual matching uncertainty is used as the basis for correcting the image quality of the hole edge.

[0063] In one specific processing method, the influence ratio of spatial residual degree in the hole position anomaly identification result is set to 40%, the influence ratio of the degree of inconsistency in deformation of adjacent hole positions is set to 35%, and the influence ratio of visual matching uncertainty degree is set to 25%. For cabinet sheet metal structural components with strong plating reflectivity or holes that are easily obscured by welding slag, the influence ratio of visual matching uncertainty degree can be increased; for cabinet columns with holes densely arranged along a straight line, the influence ratio of the degree of inconsistency in deformation of adjacent hole positions can be increased.

[0064] The higher the degree of hole position anomaly identification, the more difficult it is to explain the local residual of the hole position by continuous workpiece deformation. At the same time, the lower the visual recognition reliability of the hole position, the more likely it is to be an anomaly identification point.

[0065] S62. Determine the degree of local non-rigid deformation of each riveting hole to be pressed; For any riveting hole to be pressed, continue to read the degree of spatial residual and the degree of consistency of deformation of adjacent holes used in step S61, and simultaneously read the degree of continuity of inertial navigation pose and the degree of consistency of feedback of completed riveting points.

[0066] The continuity of inertial navigation pose is determined based on the continuous sampling data collected by the visual-inertial fusion navigation computing unit during the workpiece's entry into the riveting station and the riveting process. When the overall translation and overall rotation reflected by the inertial navigation data remain stable during the continuous sampling period and there are no sudden changes exceeding the allowable range of equipment vibration, the continuity of inertial navigation pose is high, indicating that the local residual of the current hole position is not easily caused by overall workpiece shaking, loose clamping, or instantaneous vibration of the industrial camera.

[0067] When inertial navigation data shows significant jumps within a short period, or when the overall translation and rotation amounts change repeatedly, the continuity of inertial navigation pose is low. In this case, even if the visual residuals of adjacent holes have a certain degree of continuity, they are not directly identified as actual local warping of the workpiece.

[0068] The consistency of feedback for completed riveting points is determined based on the actual feedback deviation collected in step S5. When the direction of the actual feedback deviation of completed riveting points within the same adjacent hole group is similar to the direction of the local residual of the current hole to be riveted, and the amplitude of the actual feedback deviation shows a continuous change along the hole arrangement direction, the consistency of feedback for completed riveting points is high, indicating that the local residual obtained by visual acquisition can be verified by the actual position feedback during the riveting process.

[0069] When the actual feedback deviation of the completed riveting point is close to zero, or the direction of the actual feedback deviation is significantly opposite to the direction of the local residual of the current riveting hole, the consistency of the feedback of the completed riveting point is low, indicating that the visual residual of the current riveting hole may be affected by hole edge reflection, occlusion, or contour recognition error.

[0070] When there are no completed riveting points around the current riveting hole, the consistency of feedback from completed riveting points is set to an intermediate state, so that the judgment result of local non-rigid deformation is not increased or decreased temporarily. After at least one reference riveting hole in the same adjacent hole group has been riveted, this result is updated according to its actual feedback deviation.

[0071] The degree of local non-rigid deformation of the current riveting hole is obtained by comprehensively considering the degree of spatial residual, the degree of consistency of deformation of adjacent holes, the degree of continuity of inertial navigation pose, and the degree of consistency of feedback from completed riveting points. Among these, the degree of spatial residual and the degree of consistency of deformation of adjacent holes are the main criteria for judgment, the degree of continuity of inertial navigation pose is used to exclude overall workpiece shaking or instantaneous disturbances of the acquisition equipment, and the degree of consistency of feedback from completed riveting points is used to verify the visual residual using the actual riveting results.

[0072] In one specific processing method, the influence ratios of the degree of spatial residual and the degree of consistency of deformation of adjacent holes in the judgment result of local non-rigid deformation are set to 30%, and the influence ratios of the degree of continuity of inertial navigation pose and the degree of consistency of feedback of completed riveting points are set to 20%.

[0073] Steps S61 and S62 both utilize the degree of spatial residual and the degree of consistency of deformation between adjacent holes, but they function differently. In step S61, inconsistent deformation between adjacent holes improves the degree of hole anomaly identification; in step S62, consistent deformation between adjacent holes improves the degree of local non-rigid deformation.

[0074] Therefore, even if a hole position has a large spatial residual, it will not be directly identified as a local non-rigid deformation based solely on the residual amplitude. Instead, it is necessary to further confirm whether the residual is continuous with adjacent holes, whether the inertial navigation pose is stable, and whether the actual feedback from the completed riveting point supports the residual.

[0075] The higher the degree of local non-rigid deformation, the more spatial continuity, positional stability and actual feedback consistency the local residual of the hole position has, and therefore the more likely it is to originate from welding thermal deformation, bending springback or local warping of the plate surface.

[0076] S63. Determine the hole location type, local coordinate system, compensation direction, compensation amount, and compensation priority. Pre-set thresholds for hole position anomalies and local non-rigid deformation. In one specific embodiment, an anomaly determination condition is set when the degree of hole position anomaly identification reaches 65% of the evaluation range, and a local non-rigid deformation determination condition is set when the degree of local non-rigid deformation reaches 55% of the evaluation range. The determination thresholds can be modified based on historical riveting data of structural components of the same model, but the same initial determination criteria are used for sheet metal structural components of the same batch of cabinets.

[0077] When the degree of hole position anomaly identification of a certain hole to be pressed reaches or exceeds the hole position anomaly judgment limit, the hole position is determined as an anomaly identification point. Instead of directly generating compensation amount based on the actual hole center coordinates obtained from this visual acquisition, the industrial camera visual acquisition module is controlled to perform local secondary acquisition of the hole position.

[0078] During local secondary acquisition, the exposure time and image acquisition parameters of the corresponding hole area are adjusted, and the hole edge contour extraction and hole center coordinate recognition are re-executed. If the degree of hole position anomaly recognition after secondary acquisition is reduced to below the anomaly judgment limit, step S62 is re-executed according to the updated hole position features; if the degree of hole position anomaly recognition after secondary acquisition still reaches or exceeds the anomaly judgment limit, the riveting action of the hole position is paused and a verification mark is generated.

[0079] When the degree of hole position abnormality identification of a certain rivet hole is lower than the abnormality judgment threshold, and the degree of local non-rigid deformation is lower than the local non-rigid deformation judgment threshold, the hole position is determined as an overall rigid posture deviation hole.

[0080] For holes with overall rigid pose deviation, the global coordinate system corrected by overall translation and overall rotation in Example 1 is used as the coordinate reference for the hole position. The globally predicted coordinates of the hole position after overall rigid pose transformation are used as the target position, and the initial global compensation amount obtained in step S2 is used as the compensation amount for the hole position. No additional local residuals are superimposed on the hole position, avoiding the introduction of normal visual measurement fluctuations into the compensation action of the riveting actuator.

[0081] When the degree of abnormality identification of a certain hole to be pressed is lower than the abnormality judgment threshold, and the degree of local non-rigid deformation reaches or exceeds the local non-rigid deformation judgment threshold, the hole is identified as a local non-rigid deformation hole.

[0082] For locally non-rigid deformation holes, a local coordinate system is reconstructed using at least three effective adjacent holes around the hole. The global predicted coordinates of the hole to be riveted are set as the origin of the local coordinate system, the direction pointing to the nearest effective adjacent hole in the same hole column is set as the first coordinate axis direction, the local plate surface formed by fitting multiple effective adjacent holes is used as the local coordinate plane, and the normal direction of the local plate surface is set as the third coordinate axis direction.

[0083] If there are fewer than three valid adjacent holes in the initial neighborhood, the neighborhood is expanded step by step according to the theoretical hole spacing until valid adjacent holes that meet the local plate surface fitting requirements are obtained. Holes identified as anomaly points are not included in the local coordinate system reconstruction to prevent the coordinates of anomaly holes from changing the fitting direction of the local plate surface.

[0084] The compensation direction for locally non-rigid deformation holes is determined by the local residual direction of that hole location. The final compensation amount is composed of the initial global compensation amount formed in Example 1 and the local compensation amount determined in this example. The local compensation amount is obtained proportionally from the corresponding local residual based on the degree of local non-rigid deformation of that hole location.

[0085] The higher the degree of local non-rigid deformation, the more likely the local residual is to originate from the actual deformation of the workpiece, and the higher the proportion of local residual participating in the final compensation. When the degree of local non-rigid deformation is only slightly higher than the judgment limit, only a part of the local residual is added to the initial global compensation amount to avoid over-compensation of the riveting hole caused by compensating for all visual residuals at once.

[0086] The final compensation amount is compared with the allowable position compensation range for the hole determined in step S3. When the final compensation amount is within the allowable position compensation range, it is written into the riveting execution axis control field of the corresponding process model; when the final compensation amount exceeds the allowable position compensation range, the compensation amount is limited within the equipment safety boundary, and a compensation over-limit mark is generated for the hole.

[0087] After determining the hole type and compensation amount, the compensation priority for each hole is further determined. Abnormal identification points are not included in the riveting execution sequence until secondary acquisition or verification is completed. Holes with overall rigid pose deviations have a higher compensation priority than holes with local non-rigid deformations, allowing the riveting equipment to prioritize completing reference riveting holes with reliable visual recognition and small local residuals. The actual feedback deviation of the reference riveting holes is then used to update adjacent holes that have not yet been riveted.

[0088] Within holes with overall rigid pose deviations, the lower the degree of hole position anomaly recognition and the higher the visual matching confidence level, the earlier the corresponding hole position is processed. Within holes with local non-rigid deformation, the lower the degree of hole position anomaly recognition and the higher the degree of local non-rigid deformation, the earlier the corresponding hole position is processed, so as to prioritize the verification of areas with clear local deformation features.

[0089] For each riveting hole completed, its actual feedback deviation is written back to the corresponding hole feature data, and the consistency of feedback of completed riveting points, the degree of local non-rigid deformation, and the final compensation amount of adjacent holes that have not yet been riveted are re-determined. This forms a compensation update process that gradually advances from the reference riveting hole to the local deformation area.

[0090] In one specific embodiment of this example, the cabinet column to be processed is provided with eighteen riveting holes spaced apart along its length. After the industrial camera vision acquisition module completes the full-domain image acquisition, the visual-inertial fusion navigation calculation unit determines, based on multiple high-confidence reference holes, that the cabinet column has an overall translational deviation of approximately 1.18 mm in the first horizontal direction, an overall translational deviation of approximately -0.64 mm in the second horizontal direction, and an overall rotational deviation of approximately 0.36 degrees relative to the standard theoretical pose.

[0091] After the overall rigid pose transformation, the local residuals of the first to sixth riveting holes and the fifteenth to eighteenth riveting holes are small, the hole edge contours are complete, and there is no obvious continuous local offset between adjacent holes. Therefore, they are identified as holes with overall rigid pose deviation.

[0092] The local residuals of the seventh to tenth crimping holes ranged from approximately 0.28 mm to 0.42 mm. The directions of the local residuals for multiple hole positions were basically consistent, and the amplitude of the local residuals varied continuously along the arrangement direction of the holes. Meanwhile, inertial navigation data indicated that the cabinet uprights did not experience overall shaking or loosening of the clamps during the current sampling period. Therefore, the seventh to tenth crimping holes were identified as locally non-rigid deformation holes.

[0093] The fourteenth riveting hole is affected by the welding slag and plating reflection around the hole, resulting in only a partial closure of its edge contour. This leads to a low confidence level in visual matching, and its local residual direction differs from that of the thirteenth and fifteenth riveting holes. Therefore, it is identified as an anomaly. No compensation will be generated for the fourteenth riveting hole initially; instead, a secondary local acquisition will be performed by the industrial camera's vision acquisition module.

[0094] The riveting equipment first riveted the first, third, and fifth riveting holes according to the initial global compensation amount formed in Example 1, and then obtained the corresponding actual feedback deviations through the displacement feedback acquisition channel. The actual feedback deviations of the above-mentioned riveting holes were all within the allowable range of the equipment, indicating that the overall rigid pose compensation result formed in Example 1 is reliable.

[0095] Subsequently, the riveting equipment completes the local compensation riveting of the seventh riveting hole according to the reconstructed local coordinate system. The actual feedback deviation direction after riveting the seventh riveting hole is consistent with the local residual direction of the eighth to tenth riveting holes, thereby improving the consistency of feedback of the completed riveting points of the eighth to tenth riveting holes, and re-determining the degree of local non-rigid deformation and the final compensation amount of the eighth to tenth riveting holes.

[0096] The fourteenth riveting hole was re-acquired after the exposure time was adjusted. If the re-extracted hole edge contour is restored to completeness and the degree of hole position anomaly identification is reduced to below the anomaly judgment threshold, it is added to the subsequent riveting execution sequence according to the updated judgment result; if reliable hole center coordinates cannot be obtained after the second acquisition, the process is paused to avoid transmitting abnormal visual coordinates to the riveting execution axis.

[0097] The degree of hole position anomaly identification is determined by the degree of spatial residual, the degree of consistency of deformation between adjacent holes, and the degree of confidence in visual matching. When the local residual of a hole position is large, but the residual remains continuous with adjacent holes and the hole edge contour is reliably identified, it will not be identified as an anomaly point solely because of the large local residual. When the local residual of a hole position is large, the changes of adjacent holes are inconsistent, and the degree of confidence in visual matching is low, it is preferentially identified as an anomaly point to prevent the coordinates of abnormal holes from entering the local coordinate system reconstruction and compensation calculation process.

[0098] The degree of local non-rigid deformation is further determined by the degree of spatial residual, the degree of consistency of deformation between adjacent holes, the degree of continuity of inertial navigation pose, and the degree of consistency of feedback from completed riveting points. This ensures that the judgment result of local warping is simultaneously constrained by visual data before riveting, inertial navigation data, and actual position feedback after riveting. As the number of completed riveting points increases, adjacent holes that have not yet been riveted can use the newly added actual feedback data to correct the judgment result of local non-rigid deformation and the final compensation amount.

[0099] Finally, the overall rigid pose deviation holes continue to use the global coordinate system and initial global compensation amount formed in Example 1; the local non-rigid deformation holes use the local coordinate system reconstructed based on the adjacent effective hole positions, and the local compensation amount adapted to the degree of local non-rigid deformation is superimposed on the initial global compensation amount; the abnormal identification points are suspended and secondary acquisition or verification is performed.

[0100] This enables individual hole compensation for different rivet holes, reducing local overcompensation and undercompensation caused by uniform pose compensation, and minimizing problems such as rivet head offset, rivet misalignment, and hole edge cracking caused by abnormal visual coordinates.

[0101] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A control method for a cabinet sheet metal structural component riveting equipment based on positioning matching, characterized in that, The method includes: S1. Perform dual acquisition and processing of RFID radio frequency identification and machine vision on the cabinet sheet metal structural parts input from the production line. Perform format verification and noise filtering on the acquired data to generate workpiece identity and appearance feature fusion data. Perform contour extraction and category labeling on the workpiece identity and appearance feature fusion data to generate a multi-dimensional workpiece feature set. S2. Perform pose deviation analysis on the multi-dimensional workpiece feature set, map the deviation to the preset compensation strategy library, and generate pose compensation instruction data; extend the pose compensation instruction data by process association, supplement the riveting process constraints corresponding to the current workpiece material properties and structural part types, and generate extended process constraint data. S3. Perform instruction structure transformation on the extended process constraint data, transforming the riveting requirements in the form of process constraint conditions into structured control instructions that the riveting equipment can execute, and generating structured control instruction data; perform model matching analysis on the structured control instruction data, and select process model types that match the control instructions from the preset riveting process knowledge graph library, and generate process model matching data. S4. Perform parameter domain delimitation processing on the process model matching data. Combine the structural component type, hole distribution characteristics, and material property parameters to limit the boundary of the riveting force, holding time, and mold stroke parameter domains of each matching process model, and generate parameter domain limit data for each process model. Prioritize the parameter domain limit data of each process model to generate process model call priority sequence data. S5. Perform resource pre-allocation processing on the priority sequence data of the process model call, allocate independent control channels and execution timing for each process model, and generate process model resource pre-allocation data; prepare for parallel call of the process model resource pre-allocation data, and generate a multi-process model parallel call ready signal.

2. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 1, characterized in that, S1 includes: S11. Receive the original acquisition signal of the cabinet sheet metal structure transmitted from the production line conveyor end, and simultaneously start the RFID radio frequency reading module and the industrial camera vision acquisition module to complete synchronous sampling. Acquire the information stored in the workpiece's built-in radio frequency chip and the full-area image information of the workpiece surface. After the acquisition is completed, perform data bit format verification and image salt-and-pepper noise Gaussian noise layer filtering operation, and output the workpiece identity code data stream and the workpiece appearance grayscale image data stream fused together to form the workpiece identity and appearance feature fusion data. S12. The edge image operator is called to perform edge contour pixel-by-pixel extraction operation on the image components in the fusion data of workpiece identity and appearance features. Simultaneously, the radio frequency analysis program is called to read the preset classification identifier in the identity code to complete the workpiece category marking operation. During the operation, the three types of physical parameters in the image are separated: the hole coordinate arrangement value, the plate thickness and hardness value, and the coating material composition value. All operation outputs are summarized and assembled to form a multi-dimensional workpiece feature set. S13. Read the structural component type label values, hole position distribution coordinate array, and material attribute parameter array stored in the multidimensional workpiece feature set. Call the feature normalization operation program to perform unified dimension scaling conversion on the three types of feature values. After conversion, perform continuous interval interpolation filling on discrete feature points. After filling, remove abnormal feature points whose values ​​fluctuate beyond the reasonable fluctuation range and output the purified standardized workpiece feature array. S14. Perform feature weight assignment operation on the standardized workpiece feature array. According to the influence of different features on the processing accuracy under sheet metal riveting conditions, allocate gradient weight coefficients. After the weight assignment is completed, perform feature vector space mapping operation to convert discrete numerical features into high-dimensional feature vector matrix and output weighted feature vector matrix. S15. Perform feature redundancy compression operation on the weighted feature vector matrix, use principal component decomposition to remove linearly related redundant vector components in the matrix, retain the core feature vector group that has a numerical impact on the riveting process after compression, and output the final simplified and finalized multidimensional workpiece feature set.

3. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 2, characterized in that, S14 includes: Read the structural component type features, hole distribution features, and material attribute core feature data from the final multidimensional workpiece feature set. Combine these with the attribute parameters of each candidate process model in the riveting process model matching data to perform multidimensional parameter adaptation comparison calculations. This will complete the preliminary definition of the processing adaptation range of each process model and generate the initial range delineation data for the process model. The initial range definition data of the process model is used to perform processing condition adaptation correction calculation. Combined with the equipment operating condition threshold of sheet metal structural parts riveting processing, the processing adaptation range of each process model is calibrated in the whole domain. Invalid intervals that exceed the operating limits of equipment hardware are eliminated, and the precise range limit data of the process model is generated. Cross-validation calculations are performed on the precise range limit data of the process model to compare the processing coverage of different process models, identify overlapping and blank areas in the processing intervals of multiple models, fill in the gaps in the intervals, segment and optimize the overlapping areas of the intervals, and generate conflict-free search range limit data for each process model. Read the core indicators of interval adaptation accuracy, processing energy consumption, and operational stability of the search range limit data of each process model, carry out multi-dimensional quantitative scoring calculation, perform comprehensive performance quantitative rating of all matching process models, and generate process model quantitative scoring data; The process model quantitative scoring data is sorted hierarchically, and all process models are arranged in order according to their quantitative scoring values. The order of calling each model is marked synchronously, generating an ordered sequence of process model calling priority data.

4. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 2, characterized in that, S15 includes: Read the priority sequence data of the process model call, traverse the processing and operation parameters corresponding to all process model entries in the sequence, call the hardware resource pool and software operation resource pool of the equipment control system, perform full-domain detection and statistics on the idle and occupied status of various resources, and generate equipment resource status statistics. Perform resource matching and allocation calculations on equipment resource status statistics, and combine the computing power consumption, control channel occupation, and cache space occupation requirements of each process model to match suitable system operating resources for process models of different priorities, generating preliminary resource allocation mapping data; The initial resource allocation mapping data is isolated by resource isolation processing. The computing units, control channels, and memory cache areas corresponding to each process model are independently divided to prevent resource cross-occupancy and data interference during the operation of multiple models, and to generate isolated model resource allocation data. The isolated model resource allocation data is processed for runtime environment configuration. Each process model that has completed resource allocation is configured with exclusive runtime parameters, process scheduling rules and data storage paths to generate complete process model resource pre-allocation data. Parallel adaptation verification and startup preprocessing are performed on the pre-allocated resource data of the process model. The process startup timing and running adaptation parameters of each model are uniformly calibrated, the pre-initialization configuration for parallel operation of multiple models is completed, and a multi-process model parallel call ready signal is generated.

5. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 1, characterized in that, S2 includes: S21. Retrieve the standard theoretical pose coordinate matrix of the same type of structural component stored in the device storage unit. Input the actual spatial coordinate posture matrix of the workpiece obtained by extracting the multi-dimensional workpiece feature set and the theoretical pose coordinate matrix into the visual-inertial fusion navigation calculation unit. The unit performs the calculation of the three-axis translation deviation value and the three-axis rotation angle deviation value axis-by-axis difference value, and summarizes all deviation values ​​to generate a global pose deviation value matrix. S22. Retrieve the complete dataset of the compensation strategy library stored locally on the device, use the deviation amplitude of each axis in the global pose deviation numerical matrix as the search matching factor, traverse all compensation strategy samples in the library to complete the numerical similarity comparison, and generate the initial pose compensation command data by combining the displacement compensation amount and angle correction amount corresponding to the sample with the highest matching degree. S23. Parse the material hardness and thickness parameters and structural component classification label parameters stored in the multi-dimensional workpiece feature set, retrieve the process threshold range array of the corresponding material and structural category in the riveting process constraint database, concatenate the threshold range array to the end field of the initial pose compensation instruction data, complete the process parameter supplementation of the instruction content, and output the supplemented intermediate extended process data. S24. Perform threshold over-limit screening calculation on intermediate extended process data, compare the added process threshold with the equipment hardware limit operating parameters, replace the threshold value that exceeds the hardware bearing range with the hardware critical safe operating value, and output the calibration extended process constraint data after correcting the value. S25. Perform timing field filling operation on the calibration-type extended process constraint data, embed the workpiece processing timing sequence number and equipment station number identifier segment in the data header, and generate complete extended process constraint data with station timing mark after filling.

6. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 1, characterized in that, The S3 includes: S31. The extended process constraint data is divided into three categories of segmented data: pose correction value segment, material process threshold segment, and structural component category identification segment. The text-to-machine code compiler is called to convert the natural parameter segment format into binary instruction code segments that the device controller can recognize. Multiple binary code segments are spliced ​​together to generate basic structured control instruction data. S32. Read the riveting processing requirement information carried by the binary code in the basic structured control instruction data, traverse all node entities in the locally deployed riveting process knowledge graph library, calculate the feature cosine similarity between the instruction requirement and each process node entity in the graph, mark the matching level of each node from high to low similarity value, and output the process model matching preliminary screening data with matching level label. S33. Perform graph association node tracing operation on the initial screening data of process model matching, extract upstream and downstream auxiliary process sub-model entities of matching process nodes, include the auxiliary sub-model entities into the matching set, expand the original matching model range, and output the expanded process model matching set data. S34. Perform model conflict verification operation on the expanded process model matching set data, compare the riveting execution axis occupation requirements of different models in the set, mark the pair model combinations with axis resource contention conflicts, and output the process model matching data with marked conflict points. S35. Perform conflict weight assignment calculation on the process model matching data of the marked conflict points, assign priority weight to the conflict model according to the workpiece processing accuracy priority, embed the weight value into the corresponding model field of the matching data, and generate complete process model matching data.

7. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 1, characterized in that, The S4 includes: S41. Read each independent process model entity in the process model matching data line by line, retrieve the three sets of core parameters in the multi-dimensional workpiece feature set: structural part type label, hole position coordinate distribution array, and material hardness and thickness. Use the three sets of parameters as boundary constraint inputs, and substitute them into the riveting force limit solution formula, the holding time adaptation formula, and the mold extension stroke constraint formula to complete the numerical upper and lower limit solution. After the single model solution is completed, output the single model's exclusive parameter upper and lower limit array. S42. Integrate the upper and lower limit arrays of parameters corresponding to all matching process models, summarize them to form the set of original parameter domain data of multiple models, call the numerical interval intersection operation program, calculate the overlapping interval range of three types of parameters, namely riveting force, holding time and mold stroke, between different models, mark the opposite parameter items with no overlapping intervals, and output the set of parameter domain data of multiple models with overlapping marks. S43. Perform a safe interval shrinkage operation on the multi-model parameter domain set data with overlapping markers. Select the maximum value of the lower limit of each model parameter as the unified processing lower limit and the minimum value of the upper limit of each model parameter as the unified processing upper limit. The overlapping interval parameters retain the original interval range. After the shrinkage operation is completed, the parameter domain limit initial version data of each process model is generated. S44. Retrieve the response rate parameters of the hydraulic unit of the equipment and the response lag parameters of the mechanical transmission of the mold. Substitute the two types of hardware response parameters into the parameter domain correction formula, perform lag compensation fine-tuning on all upper and lower limit values ​​within the initial data of the parameter domain, and output the parameter domain limit data of each process model after fine-tuning. S45. Perform a comprehensive scoring operation on the parameter domain limit data of each modified process model. Calculate the comprehensive score of each process model based on three indicators: parameter range fit, hardware load consumption, and estimated processing accuracy. Sort the models from high to low scores and output the process model call priority sequence data.

8. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 7, characterized in that, S44 includes: Read the initial data of the parameter domain limit of each process model, extract all extreme values ​​of the riveting force, holding time and mold stroke in the initial data in batches, and summarize and integrate them to form the original corrected dataset of model parameters; The dynamic response rate of the hydraulic drive unit of the equipment and the mechanical hysteresis of the mold transmission structure are collected. The parameter fluctuation offset under continuous operation of the equipment is statistically analyzed to generate hardware response characteristic benchmark data. By substituting the hardware response characteristic baseline data into the parameter domain dynamic correction algorithm, synchronous offset calibration is performed on the interval extreme values ​​in the original correction dataset of model parameters to offset the parameter deviations caused by mechanical transmission and hydraulic response, and intermediate parameter calibration data is generated. Extreme value boundary verification calculations are performed on intermediate parameter calibration data to screen whether the parameter values ​​after calibration exceed the safe operating threshold of the equipment hardware. Critical fall-off adjustments are made for values ​​that exceed the limit to generate compliant parameter calibration data. The calibration data of compliant parameters are normalized to unify the range accuracy standards of the three types of processing parameters of each process model. After normalization, the parameter domain limit data of each process model is output as a correction.

9. The control method for a cabinet sheet metal structural component riveting equipment based on positioning matching according to claim 1, characterized in that, The S5 includes: S51. Read all process model entries sorted in the priority sequence data of the process model call, split the three independent hardware channel resource pools of the riveting execution axis control channel, pressure sensing acquisition channel and displacement feedback acquisition channel in the equipment main control system, and allocate dedicated control channel and acquisition channel numbers in order of model priority, and write the channel number into the corresponding model storage field. S52. Retrieve the timing scheduling calculation unit, and based on the estimated execution time of each process model's processing action, arrange the timing arrangement arrays for the start time, duration, and end time of the riveting action of different models. Bind the timing arrangement arrays to the corresponding model resource allocation entries, and generate the initial data for the pre-allocation of process model resources that binds the channels and timing. S53. Perform timing gap verification calculation on the initial data of pre-allocation of process model resources, calculate the interval between the end point of adjacent priority model actions and the start point of the next model, and automatically extend the gap to the reset critical time for entries whose interval is less than the minimum time consumption of equipment hardware reset. After the time correction, output the timing calibration type process model resource pre-allocation data. S54. Call the equipment memory partition allocation program, divide each process model in the time-series calibration process model resource pre-allocation data into an independent memory read / write partition, match the partition size with the memory capacity required for storing parameters and caching real-time data during model runtime, write the partition address into the corresponding model resource field, and output the process model resource pre-allocation data matching the memory partition. S55. Perform parallel trigger signal compilation and operation on the pre-allocated data of the process model resources associated with the memory partition, generate an independent level trigger signal code for each process model with allocated resources, and package and integrate all signal codes to generate a unified multi-process model parallel call ready signal.

10. A control system for a cabinet sheet metal structural component riveting equipment based on positioning matching, characterized in that, The system includes: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images