An artificial intelligence-based component positioning and assembly system

Abstract

The application provides a kind of parts positioning and assembly system based on artificial intelligence, it is related to artificial intelligence technical field.System includes part positioning module, for different parts are classified to obtain part classification data, part classification data is positioned to obtain part positioning parameter, establishes dynamic positioning system, and part positioning parameter is dynamically compensated to obtain positioning correction data.Path planning module is used for using point cloud registration method to align parts to obtain part matching data, generates robot assembly path according to part matching data, and adjusts the optimal path based on positioning correction data.Assembly traceability module constructs quality traceability mechanism.The application can automatically, intelligently position and assemble parts through dynamic compensation of part positioning parameters and alignment using point cloud registration, improve production efficiency, significantly improve assembly accuracy, reduce errors and reduce production costs.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based component positioning and assembly system. Background Technology

[0002] In the field of modern intelligent manufacturing, high-precision positioning and automated assembly of components are core aspects of ensuring product quality. With the development of high-end manufacturing industries such as aerospace, semiconductors, and new energy vehicles, the requirements for component assembly precision have increased from the traditional millimeter level (±0.1mm) to the micrometer level (±10μm), and even sub-micrometer level (±1μm). For example, semiconductor chip bonding requires ensuring a solder ball spacing error of less than 5μm, and the assembly of aero-engine blade tenons requires a coaxiality error of less than 0.03mm. At the same time, flexible production models with multiple varieties and small batches are gradually becoming mainstream, requiring assembly systems to have the ability to quickly identify, locate, and plan paths for different components.

[0003] Traditional positioning systems often rely on a single visual sensor or laser rangefinder, resulting in significant data noise under complex conditions such as strong light, vibration, and high temperatures (e.g., arc light interference in automotive welding workshops or mechanical vibrations on high-speed production lines). Traditional Iterative Closest Point (ICP) algorithms take over 10 seconds to iterate through millions of point clouds and are prone to getting trapped in local optima (e.g., matching ambiguities in symmetrical parts). Point cloud preprocessing often employs only simple filtering, failing to incorporate multi-dimensional features such as component material properties and surface texture.

[0004] Traditional methods generate paths based on ideal poses. When the positioning error exceeds 0.05mm, the robot is prone to sudden stops and turns (peak acceleration exceeding 15m / s²), leading to part collisions or equipment damage. Furthermore, the lack of force-position hybrid control during the assembly contact stage makes rigid positioning prone to stress concentration.

[0005] To address the aforementioned problems, this invention proposes an artificial intelligence-based component positioning and assembly system. Summary of the Invention

[0006] This invention provides an artificial intelligence-based component positioning and assembly system to address the shortcomings of insufficient real-time positioning accuracy and poor alignment accuracy in existing technologies.

[0007] This invention provides an artificial intelligence-based component positioning and assembly system, comprising:

[0008] The part positioning module is used to classify different parts to obtain part classification data, perform position positioning on the part classification data to obtain part positioning parameters, establish a dynamic positioning system, and perform dynamic compensation on the part positioning parameters to obtain positioning correction data.

[0009] The path planning module is used to align parts using point cloud registration methods to obtain part matching data, generate robot assembly paths based on the part matching data, and adjust the robot assembly paths based on positioning correction data to obtain the optimal path.

[0010] The assembly traceability module assembles parts according to the optimal path, monitors and records assembly parameters to obtain overall process data, and builds a quality traceability mechanism based on the overall process data.

[0011] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the steps for obtaining component classification data include:

[0012] Collect data on the geometric dimensions, material properties, and surface texture of components, and combine this data with business data to form component data.

[0013] Based on the target requirements, determine the classification objectives and formulate classification criteria from geometric features, texture features, and functional features.

[0014] The component data is classified based on the classification criteria, and the classified component data is adjusted according to the similarity to obtain the component classification data.

[0015] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the steps for obtaining component positioning parameters include:

[0016] The positioning data is obtained and preprocessed to obtain the processing data. Based on the product's organizational structure and distribution requirements, the positioning coordinate system is determined.

[0017] Key geometric information is extracted from the part classification data and processing basis data, and a contour matching algorithm is used to match it with the theoretical feature information in the processing basis data to determine the positioning result of the part in the positioning coordinate system.

[0018] Based on the positioning results, the rotation matrix and translation vector are extracted and converted into dual quaternions. Assembly constraints are then defined from the angles and positions, expressed by the following formula:

[0019]

[0020]

[0021] In the formula, It is the conjugate of the real part. It is the position error term. It is the unit vector of the rotation axis. It's an angular deviation. It is a translation vector. It is a dual unit. It is the target axis. It is the reference axis of the part. It is an ideal translation vector. It is the real part of the dual quaternion.

[0022] The objective function is defined based on assembly constraints. A nonlinear optimization method is used to minimize the objective function while maintaining the unity constraint of the dual quaternions. The optimized objective function is expressed as follows:

[0023]

[0024] In the formula, It is to optimize the objective function. It is the conjugate of the dual part.

[0025] Based on the optimization objective function and assembly relationship constraints, the positioning parameters of the parts are obtained through positioning calculations.

[0026] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the steps for establishing a dynamic positioning system include:

[0027] Positioning sensors are used to acquire the position and orientation data of the parts in real time, and bilateral filtering is used to eliminate reflections and wavelet transform is used to filter out high-frequency noise.

[0028] The real-time error is obtained by calculating the deviation between the corrected position and orientation of the part and the actual position and orientation.

[0029] The position and orientation data of the part are analyzed from time-varying error, dynamic error and process error to obtain the current characteristic parameters, and the compensation amount is calculated to predict the real-time error.

[0030] The PID control algorithm is used to adjust the compensation amount based on the real-time error to obtain the updated compensation amount, which is expressed by the formula:

[0031]

[0032] In the formula, It is a real-time error. , , These are control parameters. It is the amount of compensation for the update. From 0 The integral of the error at time step, It is the derivative of the error with respect to time.

[0033] If the real-time error exceeds the preset threshold, the part positioning data is adjusted according to the updated compensation amount; otherwise, the position and attitude of the part are monitored.

[0034] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the steps for obtaining positioning correction data include:

[0035] The part positioning data is directly superimposed with the updated compensation amount for translation correction, and the updated compensation amount is converted into an axis-angle vector. The rotation matrix is ​​then updated through exponential mapping for rotation correction.

[0036] The positional error of the part after translation and rotation correction is recalculated. It is then determined whether the positional error is less than the preset value. If so, the positional correction data is output; otherwise, the control parameters for the compensation amount are readjusted and updated.

[0037] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the steps for obtaining component matching data include:

[0038] Multi-source point cloud data of the parts are collected, the data volume is reduced by voxel grid filtering, outlier filtering is used to remove points whose distance from neighboring points exceeds the preset standard deviation, and coordinate unification is performed to obtain point cloud processed data.

[0039] By calculating the point cloud normal vector, the surface features are obtained by reflecting the orientation of the local curved surface, and the curvature abrupt change points and corner points are detected to determine the key points.

[0040] Based on the coarse positioning method and combined with surface features, the key points are initially aligned to obtain initial points in the same coordinate system with similar positions and orientations.

[0041] The iterative nearest point algorithm is used to calculate the transformation relationship between key points, and the initial points are adjusted according to the transformation relationship to obtain part matching data.

[0042] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the steps for obtaining the transformation relationship include:

[0043] Define the identity matrix, which represents the initial state of no transformation between the key points and the initial points.

[0044] For each key point, find the point in the initial point that is closest to that key point to form a corresponding point pair.

[0045] The transformation matrix is ​​calculated based on all corresponding point pairs using the iterative nearest point algorithm.

[0046] The position and orientation of key points are updated based on the transformation matrix. When the error change is less than a preset threshold, the final transformation matrix is ​​output and the least squares problem is solved to obtain the transformation relationship.

[0047] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the step of generating a robot assembly path includes:

[0048] Extract the flip matrix and movement vector of the target part from the part matching data, and set the path planning error threshold according to the matching accuracy.

[0049] The motion primitives and gripping points for robot operations are determined based on the flip matrix and the movement vector, and path nodes are generated based on the motion primitives and gripping points.

[0050] The path nodes are filtered according to the error threshold to obtain qualified path nodes, and the inverse kinematics is solved to obtain the robot assembly path.

[0051] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the step of adjusting to obtain the optimal path includes:

[0052] Positioning correction data is overlaid in real time during robot execution, and a real-time compensation model is established.

[0053] During the assembly contact phase, switch to force-position hybrid control.

[0054] The robot assembly path speed is adjusted using time scaling technology, and the optimal path is obtained by adjusting qualified path nodes.

[0055] According to the artificial intelligence-based component positioning and assembly system provided by the present invention, the step of adjusting the robot assembly path speed includes:

[0056] Define the time scaling factor based on the speed target as needed, and determine the time parameters for qualified path nodes.

[0057] Adjust the time parameters of qualified path nodes according to the time scaling factor, and calculate the timestamp.

[0058] The robot's movement speed at qualified path nodes is recalculated based on the timestamp, thereby generating the robot assembly path speed.

[0059] This invention provides an artificial intelligence-based component positioning and assembly system. By acquiring real-time component position and orientation data, and processing it through bilateral filtering and wavelet transform, the system predicts errors and calculates compensation amounts through time-varying, dynamic, and process error analysis. PID control is used to adjust the compensation amounts, dynamically compensating for component positioning parameters. This solves the problems of traditional positioning methods being unable to adapt to dynamic environments and lacking accuracy, achieving a significant improvement in assembly accuracy and reducing errors. Furthermore, by collecting multi-source point cloud data, using voxel mesh filtering to reduce data volume, outlier filtering to remove abnormal points, calculating point cloud normal vectors to reflect surface features, detecting key points, and combining coarse positioning with iterative nearest-point algorithms to achieve precise alignment, the system obtains component matching data, avoiding assembly problems caused by alignment errors. This results in reduced material waste and rework costs due to positioning and assembly errors, thus lowering production costs. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0061] Figure 1 This is one of the structural flow diagrams of an artificial intelligence-based component positioning and assembly system provided in an embodiment of the present invention.

[0062] Figure 2 This is the second flowchart of an artificial intelligence-based component positioning and assembly system provided in this embodiment of the invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0064] The following is combined Figures 1-2 This invention describes an artificial intelligence-based component positioning and assembly system.

[0065] like Figure 1 As shown in the figure, an embodiment of the present invention provides a component positioning and assembly system based on artificial intelligence. The executing entity can be an artificial intelligence-based component positioning and assembly system, including:

[0066] The part positioning module is used to classify different parts to obtain part classification data, perform position positioning on the part classification data to obtain part positioning parameters, establish a dynamic positioning system, and perform dynamic compensation on the part positioning parameters to obtain positioning correction data.

[0067] The steps to obtain part classification data include:

[0068] Collect component geometric dimensions, material properties, and surface texture data, and combine them with business data to construct component data. Geometric dimensions may include volume, surface area, center of gravity, etc.; material properties may include density, hardness, etc.; and surface texture can be extracted using the LBP algorithm. Business data may include component name, the assembly to which it belongs (e.g., chassis workshop, final assembly workshop), supplier information, and inventory status.

[0069] The classification objectives are determined based on the target requirements, and classification criteria are established based on geometric features, texture features, and functional features. Methods for determining classification objectives can include: In automobile manufacturing, if classification is for subsequent robotic automated assembly, it may be necessary to classify components according to their assembly location on the vehicle (e.g., engine compartment components, chassis components, body components), assembly sequence, shape features, etc. If classification is for ease of inventory management, it may be mainly based on the category of components (e.g., electrical, mechanical, interior, etc.) and their size. Geometric features can include volume, surface area, and center of gravity coordinates. Texture features can include using the Local Binary Pattern (LBP) algorithm to convert surface texture into feature vectors. The functional role of components in the overall product system is analyzed. For example, in the conveyor system of an automated production line, the function of a sprocket is to transmit power and drive the conveyor belt. The function of a sensor is to detect signals and convert them into electrical signals for output. The functional characteristics of components can be determined by reading the product's operating principle manual and communicating with design engineers, and then their functions can be classified, such as power transmission components, control components, and execution components.

[0070] The component data is classified based on the classification criteria, and the classified component data is adjusted according to the similarity to obtain the component classification data.

[0071] like Figure 2 As shown, the steps to obtain the part positioning parameters include:

[0072] The positioning data is acquired and preprocessed to obtain processing data, and the positioning coordinate system is determined based on the product's structure and distribution requirements. Methods for acquiring positioning data may include:

[0073] Design drawings: Detailed design drawings contain information such as the theoretical assembly positions, dimensions, shapes, and positional relationships of each component of the product. This is the basis for location determination. By comparing and matching the actual part classification data with the design drawings, the precise position of each part in the overall product structure can be determined.

[0074] Coordinate measurement data: Using high-precision measuring equipment such as a coordinate measuring machine (CMM), precise coordinate measurements are performed on the parts to obtain a set of coordinate points in three-dimensional space. This coordinate data accurately reflects the shape and positional characteristics of the parts and serves as an important reference for positioning.

[0075] Laser scanning data: Laser scanning technology can quickly acquire three-dimensional point cloud data of the surface of parts. By processing and analyzing the point cloud data, a three-dimensional model of the part can be reconstructed and its position in space can be determined.

[0076] Visual recognition system data: Computer vision-based recognition systems can capture images of parts using cameras, and then use image processing algorithms to identify the shape, feature points, and other information of the parts, thereby determining their position. For example, by using an industrial camera to capture images of parts, the coordinates of the parts in the field of view can be calculated through methods such as feature point matching or contour recognition in the image.

[0077] Key geometric information is extracted from the part classification data and processing basis data, and a contour matching algorithm is used to match it with the theoretical feature information in the processing basis data to determine the positioning result of the part in the positioning coordinate system.

[0078] Based on the positioning results, the rotation matrix and translation vector are extracted and converted into dual quaternions. Assembly constraints are then defined from the angles and positions, expressed by the following formula:

[0079]

[0080]

[0081] In the formula, It is the conjugate of the real part. It is the position error term. It is the unit vector of the rotation axis. It's an angular deviation. It is a translation vector. It is a dual unit. It is the target axis. It is the reference axis of the part. It is an ideal translation vector. It is the real part of the dual quaternion.

[0082] The objective function is defined based on assembly constraints. A nonlinear optimization method is used to minimize the objective function while maintaining the unity constraint of the dual quaternions. The optimized objective function is expressed as follows:

[0083]

[0084] In the formula, It is to optimize the objective function. It is the conjugate of the dual part.

[0085] Based on the optimization objective function and assembly relationship constraints, the positioning parameters of the parts are obtained through positioning calculations.

[0086] The steps to establish a dynamic positioning system include:

[0087] Positioning sensors are used to acquire the position and orientation data of the parts in real time, and bilateral filtering is used to eliminate reflections and wavelet transform is used to filter out high-frequency noise.

[0088] The real-time error is obtained by calculating the deviation between the corrected position and orientation of the part and the actual position and orientation, expressed by the formula:

[0089]

[0090] In the formula, This is the corrected pose error. These are the corrected position parameters. These are the actual pose parameters.

[0091] The position and orientation data of the part are analyzed from time-varying errors, dynamic errors, and process errors to obtain the current characteristic parameters, and the compensation amount is calculated to predict the real-time error. The formula is expressed as:

[0092]

[0093] In the formula, It is the amount of compensation. Based on the current feature parameters and time The predicted error value.

[0094] The PID control algorithm is used to adjust the compensation amount based on the real-time error to obtain the updated compensation amount, which is expressed by the formula:

[0095]

[0096] In the formula, It is a real-time error. , , These are control parameters. It is the amount of compensation for the update. From 0 The integral of the error at time step, It is the derivative of the error with respect to time.

[0097] If the real-time error exceeds the preset threshold, the part positioning data is adjusted according to the updated compensation amount; otherwise, the position and attitude of the part are monitored.

[0098] The steps to obtain positioning correction data include:

[0099] The updated compensation amount is directly superimposed on the part positioning data for translation correction, and then converted into an axis-angle vector. Rotation correction is performed by updating the rotation matrix through exponential mapping. The formula is expressed as:

[0100]

[0101] In the formula, It is an antisymmetric matrix. It is the corrected rotation matrix. It is the rotation matrix before correction.

[0102] The positional error of the part after translation and rotation correction is recalculated. It is then determined whether the positional error is less than the preset value. If so, the positional correction data is output; otherwise, the control parameters for the compensation amount are readjusted and updated.

[0103] The path planning module is used to align parts using point cloud registration methods to obtain part matching data, generate robot assembly paths based on the part matching data, and adjust the robot assembly paths based on positioning correction data to obtain the optimal path.

[0104] The steps to obtain part matching data include:

[0105] Multi-source point cloud data of the parts are collected, the data volume is reduced by voxel grid filtering, outlier filtering is used to remove points whose distance from neighboring points exceeds the preset standard deviation, and coordinate unification is performed to obtain point cloud processed data.

[0106] By calculating the point cloud normal vector, the surface features are obtained by reflecting the orientation of the local curved surface, and the curvature abrupt change points and corner points are detected to determine the key points.

[0107] Based on the coarse positioning method and combined with surface features, the key points are initially aligned to obtain initial points in the same coordinate system with similar positions and orientations.

[0108] The iterative nearest point algorithm is used to calculate the transformation relationship between key points, and the initial points are adjusted according to the transformation relationship to obtain part matching data.

[0109] The steps to obtain the transformation relationship include:

[0110] Define the identity matrix, which represents the initial state of no transformation between the key points and the initial points.

[0111] For each key point, find the point in the initial point that is closest to that key point to form a corresponding point pair.

[0112] The transformation matrix is ​​calculated based on all corresponding point pairs using the iterative nearest point algorithm, expressed by the following formula:

[0113]

[0114] In the formula, It is a flipped matrix. It is a movement vector. , They are corresponding point pairs.

[0115] The position and orientation of keypoints are updated based on the transformation matrix. Once the error change is less than a preset threshold, the final transformation matrix is ​​output, and a least-squares problem is solved to obtain the transformation relationship. The transformation relationship consists of the optimal flip matrix and the optimal movement vector.

[0116] The steps for generating a robot assembly path include:

[0117] Extract the flip matrix and movement vector of the target part from the part matching data, and set the path planning error threshold according to the matching accuracy.

[0118] The motion primitives and gripping points for robot operations are determined based on the flip matrix and the movement vector, and path nodes are generated based on the motion primitives and gripping points.

[0119] The path nodes are filtered according to the error threshold to obtain qualified path nodes, and the inverse kinematics is solved to obtain the robot assembly path.

[0120] The steps to adjust and obtain the optimal path include:

[0121] Positioning correction data is overlaid in real time during robot execution, and a real-time compensation model is established, expressed by the formula:

[0122]

[0123] In the formula, It is the compensation gain matrix. It is a real-time correction amount. It is the compensated joint angle vector. It is the planned joint angle vector.

[0124] During the assembly contact phase, the system switches to force-position hybrid control, expressed by the following formula:

[0125]

[0126] In the formula, It is contact force. It is the desired location. , It refers to the stiffness and damping coefficients. It is the actual location. It is the expected speed. That's the actual speed.

[0127] The robot assembly path speed is adjusted using time scaling technology, and the optimal path is obtained by adjusting qualified path nodes.

[0128] The steps for adjusting the robot assembly path speed include:

[0129] Define the time scaling factor based on the speed target as needed, and determine the time parameters for qualified path nodes.

[0130] Adjust the time parameters of qualified path nodes according to the time scaling factor, and calculate the timestamp. The formula is as follows:

[0131]

[0132] In the formula, It's a timestamp. It is the time scaling factor. It is a time parameter.

[0133] The robot's movement speed at qualified path nodes is recalculated based on the timestamp, thereby generating the robot assembly path speed.

[0134] The assembly traceability module is used to assemble parts according to the optimal path, monitor and record assembly parameters to obtain overall process data, and build a quality traceability mechanism based on the overall process data.

[0135] When assembling parts according to the optimal path, the robot control system must first convert the path instructions into real-time motion parameters for each joint. Simultaneously, multi-sensor collaborative monitoring is activated to collect dynamic parameters such as force-position data, temperature changes, and vibration amplitude during the assembly process. For example, in the assembly of an automotive engine cylinder head, a six-dimensional force sensor records the axial force curve during bolt tightening, a vision system tracks the fit gap between the cylinder head and the cylinder block, and synchronously records the execution time of each assembly action, robot joint angles, and other trajectory data. After assembly, all monitoring data (such as real-time feedback values ​​from force control sensors, compensation amounts for positioning corrections, and speed fluctuation curves during path execution) are structurally integrated to form a complete process data chain containing timestamps, parameter thresholds, and execution status. When building a quality traceability mechanism based on this data chain, a multi-dimensional index database is first established, linking assembly processes, part batches, equipment numbers, and other information with real-time parameters, using blockchain technology to ensure data immutability. Secondly, a visual traceability platform is developed, supporting queries by time, workstation, quality indicators, etc. For example, by entering a batch part number, one can trace the path execution accuracy during assembly, compensation adjustment records, and fluctuations in key process parameters. Finally, AI algorithms are introduced to mine historical data and establish a quality prediction model. When real-time data deviates from the historical acceptable range, an automatic warning is issued, realizing full-process digital management from assembly execution to quality traceability, and providing data support for process optimization and defect analysis.

[0136] This embodiment provides an AI-based component positioning and assembly system. By combining geometric dimensions, material properties, surface texture, and business data, it classifies components from multiple dimensions, improving classification accuracy. It also acquires component matching data through a dynamic positioning system, real-time error compensation, and point cloud registration. Based on this matching data, it generates an initial robot assembly path, adjusts the path in real time based on positioning correction data, optimizes path speed using time scaling technology and force-position hybrid control, and introduces an error threshold to improve path accuracy. This automated and intelligent component positioning and assembly process reduces manual intervention, improves production efficiency, significantly enhances assembly accuracy, reduces errors, minimizes material waste and rework costs caused by positioning and assembly errors, lowers production costs, and can quickly adapt to different component and assembly requirements, improving the flexibility and responsiveness of the production system.

[0137] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0138] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An artificial intelligence based parts positioning and assembly system, characterized by, The method comprises the following steps: A part positioning module is used to classify different parts to obtain part classification data, position the part classification data to obtain part positioning parameters, establish a dynamic positioning system, and dynamically compensate the part positioning parameters to obtain positioning correction data. The step of classifying to obtain the part classification data comprises: Geometric dimensions, material properties, and surface texture data of the parts are collected, and combined with business data to form part data; Classification targets are determined according to target requirements, and classification standards are formulated from geometric features, texture features, and functional features; The part data is classified based on the classification standards, and the classified part data is adjusted according to similarity to obtain the part classification data; The step of establishing the dynamic positioning system comprises: Position and attitude data of the parts are obtained in real time using a positioning sensor, and the reflected light is eliminated using bilateral filtering, and high-frequency noise is filtered out through wavelet transform; The deviation between the position and attitude of the corrected parts and the real position and attitude is calculated to obtain real-time error; The position and attitude data of the parts are analyzed to obtain current characteristic parameters from time-varying error, dynamic error, and process error, and the compensation amount is predicted based on real-time error calculation; A PID control algorithm is adopted, the compensation amount is adjusted according to the real-time error to obtain an updated compensation amount, and the formula is expressed as: In the formula, It is a real-time error. , , These are control parameters. It is the amount of compensation for the update. From 0 The integral of the error at time step, It is the derivative of the error with respect to time; It is judged whether the real-time error exceeds a preset threshold, if yes, the part positioning parameters are adjusted according to the updated compensation amount, otherwise the position and attitude of the parts are continuously monitored; The step of obtaining the positioning correction data comprises: The part positioning parameters after translation correction and rotation correction are recalculated for pose error, and it is judged whether the pose error is less than a preset value, if yes, the positioning correction data is output, otherwise the control parameters of the updated compensation amount are adjusted; A path planning module is used to align the parts using a point cloud registration method to obtain part matching data, generate a robot assembly path based on the part matching data, and adjust the robot assembly path based on the positioning correction data to obtain an optimal path; An assembly traceability module is used to assemble the parts according to the optimal path, monitor and record assembly parameters to obtain overall process data, and construct a quality traceability mechanism based on the overall process data. The step of obtaining the part positioning parameters comprises:

2. The artificial intelligence based component positioning and assembly system as claimed in claim 1, wherein, Positioning basis data is obtained and preprocessed to obtain processing basis data, and a positioning coordinate system is determined according to product arrangement structure and allocation requirements; Key geometric information is extracted from the part classification data and the processing basis data, and a contour matching algorithm is used to match the key geometric information with theoretical feature information in the processing basis data to determine the positioning result of the parts in the positioning coordinate system; The rotation matrix and translation vector are converted into dual quaternions according to the positioning result, and assembly constraints are formulated from angles and positions, and the formula is expressed as: ​ wherein is the real part of the conjugate, is the positional error term, is the unit vector of the rotation axis, is the angular deviation, is the translation vector, is the dual unit, is the target axis, is the part reference axis, is the ideal translation vector, is the real part of the dual quaternion. According to the assembly constraint definition target function, the optimization method is used to minimize the target function, and the optimization target function is obtained by keeping the unit constraint of the dual quaternion, which is expressed as: wherein is an optimization objective function, is a dual part conjugate; According to the optimization target function and the assembly relationship constraint condition, positioning calculation is performed to obtain the part positioning parameters.

3. The artificial intelligence based component positioning and assembly system as claimed in claim 1, wherein, The step of obtaining the part matching data includes: Collecting multi-source point cloud data of the part, reducing data volume through voxel grid filtering, removing points with a distance from the neighborhood points exceeding a preset standard deviation through outlier filtering, and performing coordinate unification to obtain point cloud processing data; By calculating the point cloud normal vector, the surface features are obtained by reflecting the local surface orientation, and the key points are determined by detecting curvature discontinuity points and identifying corner points; According to the coarse positioning method, the key points are initially aligned in combination with the surface features, and the initial points are obtained in the same coordinate system and with similar positions and postures; The iterative closest point algorithm is used to calculate the transformation relationship between the key points, and the initial points are adjusted according to the transformation relationship to obtain the part matching data.

4. The artificial intelligence based component positioning and assembly system as claimed in claim 3, wherein, The step of obtaining the transformation relationship includes: Determine the unit matrix, which represents the initial transformation between the key points and the initial points; In each of the key points, find the closest point in the initial points to form a corresponding point pair; According to the transformation matrix calculated by all corresponding point pairs of the iterative closest point algorithm; According to the transformation matrix, the position and posture of the key points are updated, and when the error change is less than a preset threshold, the final transformation matrix is output to solve the least squares problem to obtain the transformation relationship.

5. The artificial intelligence based component positioning and assembly system as claimed in claim 1, wherein, The step of generating the robot assembly path includes: Extracting the flip matrix and movement vector of the target part from the part matching data, and setting a path planning error threshold according to the matching accuracy; According to the flip matrix and the movement vector, determine the action primitive and the grasping point of the robot operation, and generate the path node according to the action primitive and the grasping point; According to the error threshold, the path node is screened to obtain a qualified path node, and inverse kinematics is solved to obtain the robot assembly path.

6. The artificial intelligence based component positioning and assembly system as claimed in claim 5, wherein, The step of adjusting the optimal path includes: Real-time superposition of the positioning correction data during robot execution, and establishment of a real-time compensation model; Switch to force-position hybrid control during assembly contact stage; Adjust the speed of the robot assembly path using time scaling technology, and adjust the qualified path node to obtain the optimal path.

7. The artificial intelligence based component positioning and assembly system as claimed in claim 6, wherein, The step of adjusting the speed of the robot assembly path includes: According to the speed target to be adjusted, define the time scaling factor, and determine the time parameter of the qualified path node; According to the time scaling factor, adjust the time parameter of the qualified path node, and calculate the time stamp; According to the time stamp, the motion speed of the robot at the qualified path node is recalculated, thereby generating the speed of the robot assembly path.

Images