Self-adaptive control method and system for disassembling process parameters of waste household appliances

Abstract

The invention relates to the technical field of waste resource recycling and automatic processing, and discloses a waste household appliance disassembling process parameter adaptive control method and system, and the method comprises the steps: carrying out the registration fusion of multi-view surface data, and obtaining a complete point cloud model; identifying a high-deformation area based on curvature features and fitting cutting reference surface parameters; performing smooth adjustment on the reference plane parameters according to the pose difference of the adjacent areas to obtain a coordinated reference plane set; generating an initial compensation vector in combination with the discrete points of the cutting path, and correcting direction mutation to obtain a preliminary compensation instruction; and real-time cutting feedback data are collected in the disassembling process, if the machining state is abnormal, the compensation vector of the follow-up path is corrected, and a target self-adaptive control instruction is obtained. According to the invention, self-adaptive precise cutting of waste household appliance parts with complex deformation and uneven material characteristics can be realized.

Description

Technical Field

[0001] This invention relates to the field of waste resource recycling and automated processing technology, and in particular to an adaptive control method and system for process parameters of waste household appliance dismantling. Background Technology

[0002] The recycling and reuse of waste household appliances is a crucial link in the resource recycling system, and the quality of its dismantling and processing directly affects the recycling efficiency of valuable components and the purity of materials. With the surge in the amount of discarded household appliances, achieving efficient and precise automated dismantling has become an inevitable requirement for promoting green manufacturing. In practical applications, industrial control systems need to direct mechanical equipment to cut the outer shells and separate the internal components of different types of household appliances to achieve classified recycling of resources.

[0003] In a current technology, the dismantling process relies primarily on pre-set fixed parameters and empirical formulas to perform cutting and separation operations. Typically, industrial control systems plan processing paths based on ideal design models of household appliances (such as regular planes or curved surfaces), generating uniform control commands using traditional Euclidean space plane fitting methods, without considering the physical differences of individual products in actual recycling scenarios. However, actual discarded household appliances often experience complex, non-uniform bending or torsional deformations in their outer shell and internal supports due to factors such as long service life, harsh storage environments, or accidental drops and collisions during transportation. These fixed parameters based on ideal models cannot promptly detect and adapt to these real geometric deformations, resulting in a significant deviation between the cutting path and the actual surface.

[0004] In summary, existing technologies suffer from problems such as low disassembly accuracy and easy damage to components. Summary of the Invention

[0005] This invention provides an adaptive control method and system for dismantling process parameters of waste household appliances to solve the problems of low dismantling accuracy and easily damaged parts.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides an adaptive control method for process parameters of dismantling waste household appliances, comprising:

[0007] Multi-view surface data of waste household appliance parts are acquired, point cloud registration processing is performed on the multi-view surface data to obtain registered point cloud data, and the registered point cloud data is fused to obtain a complete surface point cloud model.

[0008] The curvature geometry features are calculated based on the complete surface point cloud model, and high deformation regions are identified based on the curvature geometry features to obtain a set of deformation regions containing several high deformation regions.

[0009] Based on the set of deformed regions, a chain of boundary points is extracted, and local surface fitting is performed on the chain of boundary points to obtain the cutting reference surface parameters for each of the high deformation regions.

[0010] The spatial pose difference value is calculated based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are weighted and smoothed to obtain a set of coordinated reference surfaces.

[0011] Obtain discrete point data of the cutting path, generate an initial compensation vector sequence based on the set of coordination reference surfaces and the discrete point data of the cutting path, calculate the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and perform abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation command sequence;

[0012] The device is driven to perform a disassembly operation according to the preliminary compensation instruction sequence, and real-time cutting feedback data is collected simultaneously. The processing status characteristics are calculated based on the real-time cutting feedback data. If the processing status characteristics are within a preset abnormal range, the compensation vector to be executed corresponding to the subsequent path in the preliminary compensation instruction sequence is determined, and the magnitude and direction of the compensation vector to be executed are corrected to obtain the target adaptive control instruction.

[0013] Secondly, the present invention provides an adaptive control system for process parameters of dismantling waste household appliances, comprising:

[0014] The data acquisition and modeling module is used to acquire multi-view surface data of waste household appliance parts, perform point cloud registration processing on the multi-view surface data to obtain registered point cloud data, and perform fusion processing on the registered point cloud data to obtain a complete surface point cloud model.

[0015] The deformable region identification module is used to calculate the curvature geometric features based on the complete surface point cloud model, and to identify high-deformation regions based on the curvature geometric features, thereby obtaining a set of deformable regions containing several high-deformation regions.

[0016] The reference surface parameter fitting module is used to extract the boundary point chain based on the set of deformed regions, and to perform local surface fitting on the boundary point chain to obtain the cutting reference surface parameters for each of the high deformation regions.

[0017] The parameter coordination module is used to calculate the spatial pose difference value based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are weighted and smoothed to obtain a set of coordinated reference surfaces.

[0018] The instruction generation module is used to acquire discrete point data of the cutting path, generate an initial compensation vector sequence based on the set of coordination reference surfaces and the discrete point data of the cutting path, calculate the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and perform abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation instruction sequence.

[0019] The adaptive control module is used to drive the device to perform disassembly operations according to the preliminary compensation instruction sequence, and simultaneously collect real-time cutting feedback data. It calculates the processing state characteristics based on the real-time cutting feedback data. If the processing state characteristics are within a preset abnormal range, it determines the compensation vector to be executed in the preliminary compensation instruction sequence corresponding to the subsequent path, and corrects the magnitude and direction of the compensation vector to be executed to obtain the target adaptive control instruction.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] (1) This invention uses three-dimensional scanning to obtain point cloud models, uses curvature features to identify and segment high deformation areas, and then applies local quadratic surface fitting and adjacent reference surface normal vector weighted smoothing adjustment technology. This method can break through the limitations of traditional Euclidean space plane fitting methods in describing complex non-uniform deformations, accurately capture the irregular surface morphology of waste household appliances caused by drop collisions or environmental factors, and establish continuous and coordinated cutting references between adjacent deformation areas. This effectively solves the problem of overcutting, undercutting or cutting deviation caused by fixed parameters not being able to adapt to real deformations, and improves the disassembly accuracy of severely deformed parts and the recycling purity of valuable materials.

[0022] (2) This invention integrates multi-source feedback data such as cutting force, tool vibration, material removal rate and chip morphology in real time during the cutting process. Once an abnormal value is detected, the amplitude of the subsequent path compensation vector is reduced and the direction is finely adjusted. This method can dynamically perceive the local unevenness of material properties or the sudden change of cutting state at the physical processing level, and build a closed-loop control mechanism of "monitoring-judgment-correction". It can eliminate the processing risks caused by model prediction deviation or material hidden dangers in a timely manner, thereby avoiding tool breakage or workpiece surface tearing caused by cutting overload, and ensuring the stability and safety of the entire disassembly process.

[0023] (3) This invention reconstructs the control sequence based on the adjusted compensation instructions and combines online detection of surface roughness and estimation of tool wear increment to continuously and cyclically correct the cutting instructions of subsequent similar parts. This method can transform a single abnormal correction into an adaptive optimization of the entire process. As the tool wear state accumulates and changes, the processing parameters are dynamically adjusted to ensure the consistency of processing quality under long-term continuous operation, thereby extending the life of the core components of the dismantling equipment, improving the equipment utilization efficiency, and providing optimized process data support for the large-scale, intelligent and green recycling of waste household appliances. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the adaptive control method for dismantling process parameters of waste household appliances provided in the first embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of the adaptive control system structure for the dismantling process parameters of waste household appliances provided in the second embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Reference Figure 1 The first embodiment of the present invention provides an adaptive control method for process parameters of dismantling waste household appliances, including the following steps:

[0028] S11, acquire multi-view surface data of waste household appliance parts, perform point cloud registration processing on the multi-view surface data to obtain registered point cloud data, and perform fusion processing on the registered point cloud data to obtain a complete surface point cloud model.

[0029] S12, calculate the curvature geometric features based on the complete surface point cloud model, and identify high deformation regions based on the curvature geometric features to obtain a set of deformation regions containing several high deformation regions.

[0030] S13, extract the boundary point chain according to the set of deformed regions, and perform local surface fitting on the boundary point chain to obtain the cutting reference surface parameters of each of the high deformation regions;

[0031] S14, calculate the spatial pose difference value based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds the preset consistency threshold, then perform weighted smoothing adjustment on the cutting reference surface parameters to obtain a set of coordinated reference surfaces.

[0032] S15, acquire discrete point data of the cutting path, generate an initial compensation vector sequence based on the set of coordination reference surfaces and the discrete point data of the cutting path, calculate the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and perform abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation command sequence;

[0033] S16, drive the device to perform disassembly operation according to the preliminary compensation instruction sequence, and simultaneously collect real-time cutting feedback data. Calculate the processing state characteristics based on the real-time cutting feedback data. If the processing state characteristics are within a preset abnormal range, determine the compensation vector to be executed in the preliminary compensation instruction sequence corresponding to the subsequent path, and correct the magnitude and direction of the compensation vector to be executed to obtain the target adaptive control instruction.

[0034] In step S11, multi-view surface data of discarded household appliance components are acquired, point cloud registration processing is performed on the multi-view surface data to obtain registered point cloud data, and fusion processing is performed on the registered point cloud data to obtain a complete surface point cloud model, including:

[0035] The multi-view surface data is subjected to noise removal processing to obtain clean point cloud data, and the transformation parameters are calculated based on the geometric features of the clean point cloud data to obtain a coarse registration point cloud.

[0036] Construct the projection distance function from the coarsely aligned point cloud to the preset target tangent plane, and iteratively solve the projection distance function until the preset convergence condition is met to obtain the finely aligned point cloud;

[0037] Identify overlapping regions in the precise point cloud, perform a weighted average on the overlapping regions to obtain fused point cloud data, and then smoothly stitch the fused point cloud data to obtain a complete surface point cloud model.

[0038] In practice, multi-view surface data of waste household appliance parts are acquired, and discrete spatial coordinate points on the surface of waste household appliance parts are collected using a 3D scanning device to form multi-view raw scanning data. The reflection intensity value of each point in the multi-view raw scanning data is read, and the reflection intensity value is compared with a preset reflection intensity threshold. Points with reflection intensity values ​​lower than the reflection intensity threshold are removed to obtain clean point cloud data.

[0039] Next, feature calculation and coarse registration are performed on the clean point cloud data. For each data point in the clean point cloud data, its K nearest neighbors in the three-dimensional spatial neighborhood are searched. A covariance matrix is ​​constructed using the three-dimensional coordinates of these K nearest neighbors. Eigenvalue decomposition is performed on the covariance matrix, and the eigenvector corresponding to the smallest eigenvalue is selected as the normal vector of the data point. The proportion of the smallest eigenvalue to the sum of all eigenvalues ​​is calculated as the curvature value of the data point. Based on the normal vector and curvature value, a rotation-invariant feature descriptor is constructed. Matching point pairs with similar feature descriptors are searched among point cloud data from different viewpoints. The rotation matrix and translation vector that minimize the sum of squared Euclidean distances between all matching point pairs are calculated. The coordinates of the clean point cloud data are rotated and translated using the rotation matrix and translation vector to obtain the coarsely registered point cloud.

[0040] Subsequently, the coarsely registered point cloud is refined by establishing a projection distance function that projects the source point in the coarsely registered point cloud onto the corresponding tangent plane of the target point cloud along the normal direction. The partial derivative of this projection distance function with respect to the pose transformation parameters is calculated to obtain the gradient direction. The pose transformation parameters are updated along the gradient direction that reduces the projection distance function value. The projection distance is recalculated using the updated parameters. This gradient calculation and parameter update process is repeated until the difference in projection distance error between two adjacent calculations is less than a preset value, resulting in a finely registered point cloud. Finally, the KD-Tree data structure is used to search for spatially neighboring points of point clouds from different viewpoints in the finely registered point cloud to identify overlapping regions. The Euclidean distance from each point in the overlapping region to the center of its corresponding scanning viewpoint is calculated. The fusion weight is determined based on this Euclidean distance, where a smaller Euclidean distance corresponds to a larger fusion weight. The spatial coordinates of corresponding points in the overlapping region are weighted and averaged according to this fusion weight. A smooth transition is then performed at the connection between the non-overlapping and overlapping regions to obtain a complete surface point cloud model.

[0041] It should be noted that the preset reflection intensity threshold is determined based on the photosensitive characteristics of the scanning device's sensor and the reflectivity of the surface material of the discarded household appliance parts. In actual parameter setting, those skilled in the art typically select a standard reflectivity plate for testing according to 3D imaging system performance evaluation standards such as ASTM E2544, statistically analyze the intensity distribution of background noise, and set the 95th quantile of the background noise intensity distribution as the threshold, thereby filtering out noise while retaining effective surface data. The preset convergence condition is usually set so that the change in the root mean square error generated by two adjacent iterations is less than a preset small amount (e.g., 1 × 10⁻⁶). -6 (millimeters), or the number of iterations reaches a preset upper limit (e.g., 50 times); this minute setting refers to the definition of measurement resolution in GB / T 12604.11 Nondestructive Testing Terminology Standard, and is selected by those skilled in the art based on the minimum geometric feature size required for subsequent deformation analysis.

[0042] In step S12, curvature geometric features are calculated based on the complete surface point cloud model, and high-deformation regions are identified based on the curvature geometric features to obtain a set of deformation regions containing several high-deformation regions, including:

[0043] By traversing the data points in the complete surface point cloud model, the local principal curvature, Gaussian curvature and curvature gradient direction vector are calculated to obtain the curvature geometric features.

[0044] Based on the curvature geometric features, feature points that meet the preset deformation threshold are selected, and the consistency of the curvature gradient direction vectors in the adjacent regions of the feature points is verified to obtain potential deformation feature points.

[0045] The potential deformable feature points are spatially aggregated using a region growing algorithm to obtain a cluster of highly deformable regions. The edge contours of the clusters of highly deformable regions are then extracted to obtain a set of deformable regions containing several highly deformable regions.

[0046] In practice, the process begins by traversing every data point in the complete surface point cloud model. For each data point, a spatial index structure is used to search for its K-neighborhood point set in 3D space. Based on the spatial coordinates of this neighborhood point set, the least squares method is used to fit the local quadratic surface equation. Next, the second-order partial derivatives of the local quadratic surface equation are calculated with respect to the two orthogonal tangent directions. These second-order partial derivatives are used to construct a second-order Hessian matrix. A characteristic equation containing unknown variables is then constructed. This characteristic equation is obtained by calculating the determinant of the difference between the Hessian matrix and the unknown variables multiplied by the identity matrix. Solving this characteristic equation yields two real roots, which are then identified as the maximum and minimum principal curvatures of the data point, i.e., the local principal curvatures. Simultaneously, the product of these two local principal curvatures is calculated as the Gaussian curvature, and the direction of the maximum rate of change of the local principal curvature in the spatial coordinate system is calculated to obtain the curvature gradient direction vector. Subsequently, a preset deformation judgment threshold is read, and the calculated local principal curvature value is compared with the deformation judgment threshold. Points with a local principal curvature absolute value greater than the threshold are selected as candidate points. The cosine similarity value between the candidate point and the curvature gradient direction vector of other candidate points in its neighborhood is calculated. If the cosine similarity value is greater than the preset direction consistency coefficient, the candidate point is determined to be a potential deformation feature point.

[0047] Finally, a region growing algorithm is executed. The point with the largest curvature modulus is selected from the set of potential deformable feature points as the initial seed point. Points within the spatial neighborhood of this initial seed point are retrieved, and the angle between the normal vector of the neighboring point and the normal vector of the seed point is calculated. If the neighboring points are also potential deformable feature points and the angle is less than a preset growth angle threshold, then the neighboring point is added to the current growth cluster. The search and judgment continue outward from the newly added point as the center until no more new points can be added, forming a connected cluster of high-deformation regions. The shape edge extraction algorithm identifies the outer boundary points of the connected cluster, generates a closed edge contour, and obtains a set of deformable regions containing several high-deformation areas. If no feature points satisfying the preset deformation judgment threshold are found during the traversal search, or if the calculated candidate points fail the orientation consistency verification, it is determined that the surface of the waste household appliance component has not undergone significant plastic deformation. The system will directly skip the subsequent reference surface fitting and coordination steps and directly call the standard cutting path generated based on the standard CAD model as the final control command to improve processing efficiency.

[0048] It should be noted that the preset directional consistency coefficient is usually set to a value between 0.85 and 0.95 (e.g., 0.85). This coefficient is set by collecting a large amount of sample data with real impact dents or bending deformations, calculating the cosine similarity distribution curve of the gradient between adjacent points within the deformation region, and selecting the lower tenths of this distribution curve as the value of the coefficient to distinguish between deformation fields with continuous physical meaning and random discrete measurement noise. The preset growth angle threshold is usually set to a value between 5 degrees and 15 degrees (e.g., 10 degrees). This threshold is set based on the continuous plastic deformation limit of waste household appliance shell materials (such as galvanized steel sheets or ABS plastic). The ultimate bending radius of the material can be obtained by consulting the material mechanical properties handbook, and the maximum allowable normal deflection angle per unit sampling interval can be calculated by combining the point cloud sampling density. Based on the requirements for the integrity of dismantled parts in the dismantling specifications of waste electrical and electronic products such as GB / T 26282, an empirical value that can ensure regional connectivity and avoid over-segmentation is selected as this threshold.

[0049] In step S13, boundary point chains are extracted based on the set of deformable regions, and local surface fitting is performed on the boundary point chains to obtain the cutting reference surface parameters for each of the high-deformation regions, including:

[0050] Calculate the neighborhood distance and tangential vector of discrete points in the deformable region set, and perform continuity filtering based on the neighborhood distance and tangential vector to obtain a chain of highly continuous boundary points;

[0051] Establish a local coordinate system for the highly continuous boundary point chain, and map the highly continuous boundary point chain to the local coordinate system to obtain a local mapped point set;

[0052] The local mapping point set is fitted with a quadratic surface using the least squares method to obtain a locally fitted surface. The geometric properties of the locally fitted surface are then analyzed to obtain the cutting reference surface parameters.

[0053] In practice, the set of deformable regions output in step S12 is traversed first, and discrete boundary points of the edges of high-deformation regions are extracted. For each discrete boundary point, the Euclidean distance between it and its nearest neighbor in the spatial neighborhood is calculated as the neighborhood distance, and the unit vector of the coordinate difference between the point and its nearest neighbor is calculated as the tangential vector. These points are filtered according to a preset continuity screening criterion, that is, the neighborhood distance between any two adjacent points must be less than a preset distance threshold, and the angle between the tangential vectors at two adjacent points must be less than a preset boundary continuity angle threshold, so as to exclude unnatural sharp abrupt changes. The discrete boundary points that meet the above criteria are connected in sequence to obtain a chain of high-continuity boundary points.

[0054] Next, the geometric centroid of the high-continuity boundary point chain is calculated. Using this centroid as the origin, the covariance matrix of the point chain coordinate data is decomposed into eigenvalues. A local coordinate system is established using the obtained eigenvectors. A coordinate transformation matrix is ​​used to map all points in the high-continuity boundary point chain from the world coordinate system to this local coordinate system, resulting in a locally mapped point set. Subsequently, a general equation for the local quadratic surface is defined. In this equation, the height coordinate value (i.e., the Z-axis coordinate) of any point on the surface is expressed as a linear superposition of six values. These six values ​​are: the first undetermined coefficient multiplied by the square of the point's x-coordinate (X-axis coordinate); the second undetermined coefficient multiplied by the product of the point's x-coordinate and y-coordinate (Y-axis coordinate); the third undetermined coefficient multiplied by the square of the point's y-coordinate; the fourth undetermined coefficient multiplied by the point's x-coordinate; the fifth undetermined coefficient multiplied by the point's y-coordinate; and the sixth undetermined coefficient (i.e., the constant term).

[0055] To solve for these six undetermined coefficients, a design matrix containing coordinate data of all local mapping point sets is constructed. The number of rows in this design matrix is ​​equal to the total number of data points in the local mapping point sets. Each row corresponds to a data point, and each row contains six elements in sequence: the square of the x-coordinate of the point, the product of the x-coordinate and y-coordinate of the point, the square of the y-coordinate of the point, the x-coordinate of the point, the y-coordinate of the point, and a value of 1. The height coordinates of the corresponding data points are used to form an observation vector. The inverse matrix of the product of the transpose of the design matrix and the design matrix is ​​calculated using the least squares method. This inverse matrix is ​​then multiplied by the transpose of the design matrix and the observation vector to analytically determine the six optimal undetermined coefficients. Based on the analytically determined coefficients, the normal direction of the fitted surface at the geometric center is calculated as the reference plane normal vector, and the distance from the geometric center to the origin is calculated as the reference plane offset from the origin. The cutting reference plane parameters are then output.

[0056] It should be noted that the preset distance threshold is typically set to 1.5 to 2.0 times the average sampling interval of the point cloud (e.g., 0.8 mm). This setting is based on statistical analysis of the resolution and point cloud density of the 3D scanning equipment, ensuring that the extracted boundaries are physically adjacent. The average sampling interval of the point cloud refers to the statistical average of the Euclidean distance between two adjacent sampling points at the standard working distance of the 3D scanning equipment. In this embodiment, the scanning resolution is set to 0.2 mm, and the average point cloud spacing after filtering and smoothing is approximately 0.4 mm. Those skilled in the art can adjust this according to the scanner model and the reflective characteristics of the surface being measured. The preset boundary continuity angle threshold is typically set between 15 and 30 degrees (e.g., 20 degrees). This setting references the distinction between waviness and roughness in the GB / T 1031 surface structure standard, aiming to filter out local normal abrupt changes caused by measurement noise or microscopic burrs, ensuring that the fitted reference surface reflects macroscopic structural deformation.

[0057] It is worth noting that the cutting reference surface parameters (including the reference surface normal vector and offset) obtained by the least squares method are numerical values ​​based on the local coordinate system. Before outputting the cutting reference surface parameters, or before calculating the spatial pose difference value in step S14, the inverse matrix of the coordinate transformation matrix generated when the local coordinate system was established must be used to reverse map the reference surface normal vector and geometric center coordinates back to the original world coordinate system. Only under a unified world coordinate system do the calculations of the normal angle and position offset between subsequent adjacent regions have physical comparative significance.

[0058] In step S14, the spatial pose difference value is calculated based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are weighted and smoothed to obtain a set of coordinated reference surfaces, including:

[0059] Spatial distribution relationship analysis is performed on the high deformation region to determine adjacent high deformation region pairs, and the angle between the normal vectors and the position offset between the adjacent high deformation region pairs are calculated to obtain the spatial pose difference value.

[0060] If the spatial pose difference value exceeds a preset consistency threshold, the surface area of ​​the high deformation region is calculated, and a normalized weighting factor is calculated based on the proportion of the surface area.

[0061] The normalized weighting factor is used to interpolate and smooth the cutting reference surface parameters to obtain updated reference surface parameters, and the results are summarized to obtain a set of coordinated reference surfaces.

[0062] In specific implementation, all high deformation regions are traversed, and the minimum Euclidean distance between any two region edge contour point sets is calculated. If the minimum Euclidean distance is less than the preset adjacency determination distance, the two regions are determined to be adjacent high deformation region pairs. For each pair of adjacent regions, the cutting reference surface parameters fitted in step S13 are extracted, the unit normal vectors of the two reference surfaces are obtained respectively, the dot product value of the two unit normal vectors is calculated, and the inverse cosine function value of the absolute value of the dot product value is obtained to obtain the reference surface normal vector angle. At the same time, the absolute value of the difference between the projected distances of the geometric center points of the two reference surfaces in their respective normal directions is calculated as the position offset. The reference surface normal vector angle and the position offset are combined to define the spatial pose difference value.

[0063] Next, a preset consistency threshold is read, which includes a preset reference plane coordination angle threshold and a preset distance threshold. The calculated reference plane normal vector angle is compared with the preset reference plane coordination angle threshold, and the position offset is compared with the preset distance threshold. If either value exceeds the corresponding threshold, it indicates that there is a sudden change in the processing reference of adjacent areas, and smoothing processing is required. At this time, the physical surface area of ​​two adjacent high deformation areas is calculated by accumulating the area values ​​of all triangular mesh faces within the high deformation area. The normalized weight factor is constructed by dividing the surface area value of the first high deformation area by the sum of the surface areas of the two adjacent high deformation areas to obtain the weight factor of the first area. The weight factor of the second area is calculated similarly.

[0064] Subsequently, the normalized weighting factor is used to perform weighted interpolation smoothing on the cutting reference surface parameters. For the reference surface normal vector, the normal vector of the first region is multiplied by its corresponding weighting factor, and the normal vector of the second region is multiplied by its corresponding weighting factor to obtain the intermediate vector. The intermediate vector is then normalized by calculating its magnitude, which is the square root of the sum of the squares of the horizontal, vertical, and triangular coordinates of the intermediate vector. The horizontal, vertical, and triangular coordinates of the intermediate vector are then divided by this magnitude to obtain the updated normalized coordination normal vector. For the offset of the reference surface from the origin, the same linear weighted logic is used to calculate the updated coordination offset. The updated parameters are used to replace the original parameters, and after traversing all adjacent pairs to complete the adjustment, the set of coordination reference surfaces for continuous transition of spatial attitude is obtained.

[0065] It should be noted that the preset consistency threshold is determined based on the maximum allowable cutting angle of the cutting tool of the disassembly equipment and the dynamic response capability of the servo system. Those skilled in the art typically set the preset reference plane coordination angle threshold to 5 to 10 degrees and the preset distance threshold to 1 / 10 to 1 / 5 of the cutting tool diameter (e.g., 0.5 mm) based on the ISO 10791 machining center inspection standard, combined with the maximum depth of cut and lateral force limit parameters provided by the cutting tool manufacturer. This prevents tool interference or servo overload due to drastic changes in the reference plane. The preset adjacency determination distance is typically set between 2 and 5 millimeters (e.g., 3 millimeters). This threshold is based on the segmentation gap of the previous region growing algorithm. When setting this value, those skilled in the art typically set it to 4 to 6 times the average sampling interval of the point cloud data (e.g., 0.5 mm), or the maximum allowable breakage distance set in the region growing algorithm. This ensures that physically adjacent regions with small gaps during data segmentation can be correctly identified, preventing missed adjacency determination due to missing data.

[0066] In step S15, discrete point data of the cutting path is acquired. An initial compensation vector sequence is generated based on the set of coordination reference surfaces and the discrete point data of the cutting path. The direction angle between adjacent initial compensation vectors in the initial compensation vector sequence is calculated. The initial compensation vector sequence is then corrected for abrupt changes based on the direction angle to obtain a preliminary compensation command sequence, including:

[0067] Obtain discrete point data of the cutting path, calculate the normal projection distance and direction of the discrete point data of the cutting path to the set of coordination reference surfaces, and obtain the initial compensation vector sequence;

[0068] Calculate the spatial angle between adjacent vectors in the initial compensation vector sequence, identify mutation locations that exceed the mutation determination threshold, and obtain the mutation vector index;

[0069] Extract the non-mutated vectors in the neighborhood of the mutation vector index as stable vectors, calculate the smoothed substitution vectors based on the stable vectors, and use the smoothed substitution vectors to correct the initial compensation vector sequence corresponding to the mutation vector index, thereby obtaining a preliminary compensation instruction sequence.

[0070] In practice, the data of discrete points of the pre-planned cutting path is first retrieved. This data is obtained by discretizing and sampling the standard trajectory data generated by the standard CAD model. Each discrete point of the cutting path is traversed, and the reference surface with the closest spatial distance to the discrete point is searched in the set of coordination reference surfaces output in step S14 as the reference surface. The vertical projection distance from the discrete point to the reference surface is calculated, and the normal direction of the reference surface is obtained. The vertical projection distance is multiplied by the normal direction to generate a spatial vector pointing from the cutting path point to the actual deformed surface. The vectors are arranged in the path order to form an initial compensation vector sequence.

[0071] Next, a smoothness check is performed on the initial compensation vector sequence. This process iterates through each vector in the sequence arranged in path order as the current vector and its immediately following vector as the comparison vector. For each pair of current and comparison vectors, the horizontal, vertical, and triangular coordinate components in the spatial coordinate system are read, and the dot product of the two vectors is calculated, which is the sum of the products of the corresponding coordinate components. At the same time, the magnitude of the two vectors is calculated using the Pythagorean theorem, which is the arithmetic square root of the sum of the squares of each coordinate component. The dot product is divided by the product of the magnitudes of the two vectors to obtain the cosine of the included angle. The inverse cosine of the included angle is then performed to obtain the spatial angle representing the degree of deviation between the directions of the two vectors. The spatial angle is compared with a preset mutation judgment threshold. If the spatial angle is greater than the mutation judgment threshold, it is determined that there is a command mutation at the current position, and the index number of the position is recorded as the mutation vector index.

[0072] Subsequently, for each mutation vector index, a front and back search window is set (e.g., 3 points before and after), all vectors within the window are extracted, vectors that are also marked as mutation vector indices are removed, and the remaining vectors are defined as stable vectors; the vector sum of all stable vectors is calculated, that is, their x-coordinate, y-coordinate, and y-coordinate are added together, and then divided by the total number of stable vectors to obtain the smooth replacement vector; the smooth replacement vector is used to replace the mutation vector at the corresponding index in the original sequence. After all mutations are corrected, the corrected vector sequence is decomposed into incremental values ​​of each axis (X, Y, Z and rotation axis) of the disassembly device to generate a preliminary compensation command sequence.

[0073] It should be noted that the preset sudden change judgment threshold is usually set between 15 and 25 degrees (e.g., 15 degrees). The threshold is set based on the dynamic response characteristics and acceleration / deceleration capabilities of the disassembly equipment servo system. Those skilled in the art usually calculate the maximum allowable attitude change rate within a unit interpolation cycle based on the maximum allowable angular acceleration parameter in the equipment manual and the cutting feed rate, thereby setting the threshold to avoid motor overload alarms or high-frequency vibrations in the mechanical structure caused by sudden command changes.

[0074] In step S16, the device is driven to perform a disassembly operation according to the preliminary compensation instruction sequence, and real-time cutting feedback data is collected simultaneously. The machining state characteristics are calculated based on the real-time cutting feedback data. If the machining state characteristics are within a preset abnormal range, the compensation vector to be executed corresponding to the subsequent path in the preliminary compensation instruction sequence is determined, and the magnitude and direction of the compensation vector to be executed are corrected to obtain the target adaptive control instruction, including:

[0075] Real-time cutting feedback data is collected during the disassembly process, and time-frequency domain analysis is performed on the real-time cutting feedback data to obtain the processing state characteristics;

[0076] If the machining state feature indicates a cutting abnormality, the difference between the machining state feature and the preset normal reference value is calculated to obtain the degree of abnormal deviation, and the amplitude adjustment coefficient and the direction fine-tuning angle are calculated based on the degree of abnormal deviation.

[0077] Extract the compensation vector to be executed corresponding to the subsequent path from the preliminary compensation instruction sequence, and use the amplitude adjustment coefficient and the direction fine-tuning angle to superimpose and correct the compensation vector to be executed to obtain the target adaptive control instruction.

[0078] In specific implementation, the preliminary compensation instruction sequence generated in step S15 is first loaded into the CNC system of the dismantling equipment, driving the servo motor to control the cutter to cut the waste household appliance parts along a predetermined trajectory. During the cutting process, high-frequency sensors are used to collect cutting force signals and vibration acceleration signals as real-time cutting feedback data. Next, feature extraction is performed on the collected signals to calculate the high-frequency energy ratio, average cutting force, and material removal rate in the vibration spectrum. To eliminate the influence of different physical dimensions, a maximum-minimum normalization method is preferred, linearly mapping the energy ratio, average cutting force, and material removal rate to a dimensionless interval of zero to one. Then, a weighted summation method (e.g., each component has a weight of 1 / 3) is used to combine the normalized values ​​to obtain the current processing state characteristics. Read the preset normal reference value (which has also been normalized), subtract the currently calculated machining state feature from the normal reference value, and take the absolute value of the result as the degree of abnormal deviation; if the degree of abnormal deviation is greater than the preset safety tolerance, the correction mechanism is triggered; if the degree of abnormal deviation is less than or equal to the preset safety tolerance, the current cutting state is determined to be stable, and the subsequent operations are continued according to the current preliminary compensation instruction sequence without additional correction.

[0079] It should be noted that the chip morphology feature value is obtained by acquiring chip images in real time through a machine vision imaging unit installed next to the tool. The curling radius, breakage length and serration degree of the chip are extracted by the edge detection algorithm. The acquired chip morphology parameters are compared with the chip morphology template library of the same material under normal cutting conditions. The morphology deviation index between 0 and 1 is output as the chip morphology feature value. The morphology deviation index and the material removal rate together constitute the input dimension of the processing state feature.

[0080] At this point, a negative exponential function is selected as the attenuation model to calculate the amplitude adjustment coefficient. Specifically, a sensitivity constant greater than zero is set, and the product of this sensitivity constant and the degree of abnormal deviation is calculated. The negative of this product is taken as the exponent, and the natural constant (e) is raised to the power of this exponent to obtain a value between zero and one, which is used as the amplitude adjustment coefficient. Simultaneously, the angular deviation between the resultant force direction and the feed direction is calculated based on the cutting force components. This angular deviation is multiplied by a preset gain coefficient to obtain the directional fine-tuning angle, and the rotation axis vector is determined. Specifically, when path correction within the cutting plane is required, the normal vector of the cutting plane is selected as the rotation axis; when lateral avoidance in three-dimensional space is required, the cross product vector of the compensation vector to be executed and the preset avoidance direction vector is selected as the rotation axis.

[0081] It should be noted that the sensitivity constant is used to adjust the sensitivity of amplitude attenuation to the degree of abnormal deviation. The larger the value, the greater the tool retraction amplitude under the same degree of abnormal deviation. In this embodiment, the typical range of the sensitivity constant is 0.5~2.0. The specific value can be determined by the following calibration test: a cutting test is carried out on typical waste household appliance sheet material, such as 0.6~1.0mm galvanized steel plate, and 10%, 20%, and 30% feed force overload is artificially applied. The maximum safe amplitude attenuation rate that does not cause tool chipping and surface tearing is recorded, and the sensitivity constant value is deduced. Those skilled in the art can adjust it within this range according to the toughness of the material to be disassembled, the tool diameter, and the rigidity of the equipment, or automatically optimize it through the self-tuning function during the system debugging stage.

[0082] Subsequently, the direction is corrected using the Rodriguez rotation formula. Specifically, the first component is obtained by calculating the product of the compensation vector to be executed and the cosine of the direction adjustment angle; the second component is obtained by calculating the cross product of the rotation axis vector and the compensation vector to be executed, and multiplying this cross product by the sine of the direction adjustment angle; the third component is obtained by calculating the dot product of the rotation axis vector and the compensation vector to be executed, multiplying this dot product by the rotation axis vector, and then multiplying by the difference between the numerical value and the cosine of the direction adjustment angle; the first, second, and third components are then vector-sumped to obtain the direction-corrected vector.

[0083] Finally, the magnitude of the directionally corrected vector is multiplied by the amplitude adjustment coefficient to obtain the final target adaptive control command, which replaces the original command in the buffer.

[0084] It should be noted that the preset safety margin is typically set to 15% to 20% of the normal baseline value (e.g., 15%). This setting is based on the definition of cutting stability in the ISO 10791 machining center inspection conditions standard and the tool life decay curve provided by the tool manufacturer. Those skilled in the art typically select 80% of the critical value of the characteristic mutation before the tool enters the rapid wear stage as this safety margin to ensure early intervention before destructive failure occurs. The preset gain coefficient is typically set to 0.05 to 0.15 radians per Newton (e.g., 0.1 rad / N). This setting depends on the end effector stiffness of the disassembly equipment's robotic arm and the position loop gain of the servo system. Those skilled in the art obtain the dynamic stiffness data of the system through hammer impact modal experiments and, based on stability criteria in control theory (such as the Nyquist criterion), select 60% of the maximum feedback gain that ensures the closed-loop system does not diverge and oscillate as the set value of this coefficient.

[0085] It is worth noting that the mathematical mechanism of using a negative exponential function to calculate the amplitude adjustment coefficient can achieve nonlinear adaptive adjustment. When the degree of abnormal deviation is small, the function value is close to 1, and the cutting parameters are only fine-tuned to ensure machining efficiency. When the degree of abnormal deviation increases sharply (such as when encountering hard inclusions), the function value decays rapidly to close to 0, forcing the tool to decelerate or retreat significantly, thereby achieving the optimal balance between protecting equipment safety and maintaining process cycle time.

[0086] Furthermore, to achieve continuous optimization of the disassembly process, this system also includes a self-evolving update mechanism. After completing a disassembly task, the system automatically records the data chain of this operation: "part model - deformation characteristics - final corrected control instructions - processing quality score". The processing quality score is calculated based on the flatness of the cut after disassembly and the recovery rate of valuable parts. If the processing quality score of this operation is higher than the historical average, the data chain is stored in the "adaptive case library". When processing the same type of appliance parts subsequently, if the similarity between the identified deformation area features (such as curvature geometry features) and a record in the case library exceeds 90%, the system will directly call the validated and optimized compensation vector from that record as the initial value, thereby reducing online computation and improving convergence speed.

[0087] In summary, this invention constructs a cutting benchmark based on local quadratic surface fitting and adjacent smooth coordination, and integrates multi-source feedback data such as cutting force, vibration, and chip morphology to perform closed-loop dynamic correction of compensation commands, thereby achieving adaptive and precise cutting and disassembly of waste household appliance parts with complex deformation and material inhomogeneity.

[0088] Reference Figure 2 The second embodiment of the present invention provides an adaptive control system for process parameters of waste household appliance dismantling, comprising:

[0089] The data acquisition and modeling module is used to acquire multi-view surface data of waste household appliance parts, perform point cloud registration processing on the multi-view surface data to obtain registered point cloud data, and perform fusion processing on the registered point cloud data to obtain a complete surface point cloud model.

[0090] The deformable region identification module is used to calculate the curvature geometric features based on the complete surface point cloud model, and to identify high-deformation regions based on the curvature geometric features, thereby obtaining a set of deformable regions containing several high-deformation regions.

[0091] The reference surface parameter fitting module is used to extract the boundary point chain based on the set of deformed regions, and to perform local surface fitting on the boundary point chain to obtain the cutting reference surface parameters for each of the high deformation regions.

[0092] The parameter coordination module is used to calculate the spatial pose difference value based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are weighted and smoothed to obtain a set of coordinated reference surfaces.

[0093] The instruction generation module is used to acquire discrete point data of the cutting path, generate an initial compensation vector sequence based on the set of coordination reference surfaces and the discrete point data of the cutting path, calculate the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and perform abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation instruction sequence.

[0094] The adaptive control module is used to drive the device to perform disassembly operations according to the preliminary compensation instruction sequence, and simultaneously collect real-time cutting feedback data. It calculates the processing state characteristics based on the real-time cutting feedback data. If the processing state characteristics are within a preset abnormal range, it determines the compensation vector to be executed in the preliminary compensation instruction sequence corresponding to the subsequent path, and corrects the magnitude and direction of the compensation vector to be executed to obtain the target adaptive control instruction.

[0095] It should be noted that the adaptive control system for dismantling waste household appliances provided in this embodiment of the invention is used to execute all the process steps of the adaptive control method for dismantling waste household appliances in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0096] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0097] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An adaptive control method for process parameters in the dismantling of waste household appliances, characterized in that, include: Multi-view surface data of waste household appliance parts are acquired, point cloud registration processing is performed on the multi-view surface data to obtain registered point cloud data, and the registered point cloud data is fused to obtain a complete surface point cloud model. The curvature geometry features are calculated based on the complete surface point cloud model, and high deformation regions are identified based on the curvature geometry features to obtain a set of deformation regions containing several high deformation regions. Based on the set of deformed regions, a chain of boundary points is extracted, and local surface fitting is performed on the chain of boundary points to obtain the cutting reference surface parameters for each of the high deformation regions. The spatial pose difference value is calculated based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are weighted and smoothed to obtain a set of coordinated reference surfaces. Obtain discrete point data of the cutting path, generate an initial compensation vector sequence based on the set of coordination reference surfaces and the discrete point data of the cutting path, calculate the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and perform abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation command sequence; The device is driven to perform a disassembly operation according to the preliminary compensation instruction sequence, and real-time cutting feedback data is collected simultaneously. The processing status characteristics are calculated based on the real-time cutting feedback data. If the processing status characteristics are within a preset abnormal range, the compensation vector to be executed corresponding to the subsequent path in the preliminary compensation instruction sequence is determined, and the magnitude and direction of the compensation vector to be executed are corrected to obtain the target adaptive control instruction.

2. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, The process of acquiring multi-view surface data of discarded household appliance components, performing point cloud registration processing on the multi-view surface data to obtain registered point cloud data, and fusing the registered point cloud data to obtain a complete surface point cloud model includes: The multi-view surface data is subjected to noise removal processing to obtain clean point cloud data, and the transformation parameters are calculated based on the geometric features of the clean point cloud data to obtain a coarse registration point cloud. Construct the projection distance function from the coarsely aligned point cloud to the preset target tangent plane, and iteratively solve the projection distance function until the preset convergence condition is met to obtain the finely aligned point cloud; Identify overlapping regions in the precise point cloud, perform a weighted average on the overlapping regions to obtain fused point cloud data, and then smoothly stitch the fused point cloud data to obtain a complete surface point cloud model.

3. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, The curvature geometry features are calculated based on the complete surface point cloud model, and high-deformation regions are identified based on the curvature geometry features to obtain a set of deformation regions containing several high-deformation regions, including: By traversing the data points in the complete surface point cloud model, the local principal curvature, Gaussian curvature and curvature gradient direction vector are calculated to obtain the curvature geometric features. Based on the curvature geometric features, feature points that meet the preset deformation threshold are selected, and the consistency of the curvature gradient direction vectors in the adjacent regions of the feature points is verified to obtain potential deformation feature points. The potential deformable feature points are spatially aggregated using a region growing algorithm to obtain a cluster of highly deformable regions. The edge contours of the clusters of highly deformable regions are then extracted to obtain a set of deformable regions containing several highly deformable regions.

4. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, The step of extracting boundary point chains based on the set of deformed regions and performing local surface fitting on the boundary point chains to obtain the cutting reference surface parameters for each of the high-deformation regions includes: Calculate the neighborhood distance and tangential vector of discrete points in the deformable region set, and perform continuity filtering based on the neighborhood distance and tangential vector to obtain a chain of highly continuous boundary points; Establish a local coordinate system for the highly continuous boundary point chain, and map the highly continuous boundary point chain to the local coordinate system to obtain a local mapped point set; The local mapping point set is fitted with a quadratic surface using the least squares method to obtain a locally fitted surface. The geometric properties of the locally fitted surface are then analyzed to obtain the cutting reference surface parameters.

5. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, The step involves calculating the spatial pose difference value based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are then weighted and smoothed to obtain a set of coordinated reference surfaces, including: Spatial distribution relationship analysis is performed on the high deformation region to determine adjacent high deformation region pairs, and the angle between the normal vectors and the position offset between the adjacent high deformation region pairs are calculated to obtain the spatial pose difference value. If the spatial pose difference value exceeds a preset consistency threshold, the surface area of ​​the high deformation region is calculated, and a normalized weighting factor is calculated based on the proportion of the surface area. The normalized weighting factor is used to interpolate and smooth the cutting reference surface parameters to obtain updated reference surface parameters, and the results are summarized to obtain a set of coordinated reference surfaces.

6. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, The process involves acquiring discrete point data of the cutting path, generating an initial compensation vector sequence based on the set of coordinated reference surfaces and the discrete point data of the cutting path, calculating the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and performing abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation command sequence, including: Obtain discrete point data of the cutting path, calculate the normal projection distance and direction of the discrete point data of the cutting path to the set of coordination reference surfaces, and obtain the initial compensation vector sequence; Calculate the spatial angle between adjacent vectors in the initial compensation vector sequence, identify mutation locations that exceed the mutation determination threshold, and obtain the mutation vector index; Extract the non-mutated vectors in the neighborhood of the mutation vector index as stable vectors, calculate the smoothed substitution vectors based on the stable vectors, and use the smoothed substitution vectors to correct the initial compensation vector sequence corresponding to the mutation vector index, thereby obtaining a preliminary compensation instruction sequence.

7. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, The process involves driving the device to perform a disassembly operation according to the preliminary compensation instruction sequence, simultaneously collecting real-time cutting feedback data, calculating machining state characteristics based on the real-time cutting feedback data, and determining the compensation vector to be executed corresponding to the subsequent path in the preliminary compensation instruction sequence if the machining state characteristics are within a preset abnormal range. The amplitude and direction of the compensation vector to be executed are then corrected to obtain the target adaptive control instruction, including: Real-time cutting feedback data is collected during the disassembly process, and time-frequency domain analysis is performed on the real-time cutting feedback data to obtain the processing state characteristics; If the machining state feature indicates a cutting abnormality, the difference between the machining state feature and the preset normal reference value is calculated to obtain the degree of abnormal deviation, and the amplitude adjustment coefficient and the direction fine-tuning angle are calculated based on the degree of abnormal deviation. Extract the compensation vector to be executed corresponding to the subsequent path from the preliminary compensation instruction sequence, and use the amplitude adjustment coefficient and the direction fine-tuning angle to superimpose and correct the compensation vector to be executed to obtain the target adaptive control instruction.

8. The adaptive control method for process parameters of waste household appliance dismantling according to claim 1, characterized in that, Before acquiring the discrete point data of the cutting path, the method further includes: Obtain a standard 3D model of the waste household appliance, and plan the standard cutting trajectory under ideal conditions based on the standard 3D model to obtain standard trajectory data; The standard trajectory data is discretized and sampled to obtain discrete point data of the cutting path containing a spatial coordinate sequence.

9. The adaptive control method for process parameters of waste household appliance dismantling according to claim 7, characterized in that, The processing state characteristics include at least the material removal rate and chip morphology characteristics.

10. An adaptive control system for process parameters of dismantling waste household appliances, characterized in that, include: The data acquisition and modeling module is used to acquire multi-view surface data of waste household appliance parts, perform point cloud registration processing on the multi-view surface data to obtain registered point cloud data, and perform fusion processing on the registered point cloud data to obtain a complete surface point cloud model. The deformable region identification module is used to calculate the curvature geometric features based on the complete surface point cloud model, and to identify high-deformation regions based on the curvature geometric features, thereby obtaining a set of deformable regions containing several high-deformation regions. The reference surface parameter fitting module is used to extract the boundary point chain based on the set of deformed regions, and to perform local surface fitting on the boundary point chain to obtain the cutting reference surface parameters for each of the high deformation regions. The parameter coordination module is used to calculate the spatial pose difference value based on the cutting reference surface parameters of adjacent high deformation regions. If the spatial pose difference value exceeds a preset consistency threshold, the cutting reference surface parameters are weighted and smoothed to obtain a set of coordinated reference surfaces. The instruction generation module is used to acquire discrete point data of the cutting path, generate an initial compensation vector sequence based on the set of coordination reference surfaces and the discrete point data of the cutting path, calculate the direction angle between adjacent initial compensation vectors in the initial compensation vector sequence, and perform abrupt correction on the initial compensation vector sequence based on the direction angle to obtain a preliminary compensation instruction sequence. The adaptive control module is used to drive the device to perform disassembly operations according to the preliminary compensation instruction sequence, and simultaneously collect real-time cutting feedback data. It calculates the processing state characteristics based on the real-time cutting feedback data. If the processing state characteristics are within a preset abnormal range, it determines the compensation vector to be executed in the preliminary compensation instruction sequence corresponding to the subsequent path, and corrects the magnitude and direction of the compensation vector to be executed to obtain the target adaptive control instruction.

Images