Complex casting grinding self-adaptive machining equipment and method

Abstract

The invention relates to the technical field of casting polishing, in particular to complex casting polishing self-adaptive machining equipment and method. The method comprises the steps that a to-be-polished area of a target casting is obtained; all the to-be-polished areas are divided into a plurality of polishing class clusters; the polishing difficulty of each to-be-polished area is obtained, and the total polishing difficulty of each polishing class cluster is determined based on the polishing difficulty of each to-be-polished area; a surface set where each polishing class cluster is located is obtained, and the minimum conversion cost of each polishing class cluster is obtained through the surface set where each polishing class cluster is located; according to the total polishing difficulty and the minimum conversion cost of each polishing class cluster, the comprehensive cost of each polishing class cluster is determined; based on the comprehensive cost of each polishing class cluster, the processing sequence of each to-be-polished area is obtained; a polishing path of each to-be-polished area is obtained; and the to-be-polished areas are polished according to the processing sequence and the polishing path of each to-be-polished area. The polishing quality and efficiency of the complex casting can be improved.

Description

Technical Field

[0001] This invention relates to the field of casting grinding technology, specifically to an adaptive machining equipment and method for grinding complex castings. Background Technology

[0002] Grinding of castings is a key post-casting process that directly affects the final dimensional accuracy, surface quality, and fatigue performance of the product.

[0003] Existing casting grinding methods mainly rely on automated grinding trajectory generation based on 3D scanning and comparison: A point cloud model of the casting is acquired using a 3D scanning device; then, it is precisely compared with the casting's CAD design model to calculate and identify all areas to be ground; next, all identified areas are treated as homogeneous geometry, and a unified trajectory pattern (such as the parallel section method or offset profile method) is used for path planning; finally, the grinding equipment completes the operation according to the planned path. However, when faced with numerous, varied, and dispersed grinding areas (such as risers, shrinkage cavities, edges, and cracks) on large and complex castings, this "one-size-fits-all" method struggles to adapt to the significant differences in material properties, removal allowances, and surface quality requirements across different areas, resulting in unstable grinding quality and low efficiency. Summary of the Invention

[0004] This invention provides an adaptive machining equipment and method for grinding complex castings to solve existing problems.

[0005] The adaptive machining equipment and method for grinding complex castings according to the present invention adopts the following technical solution: One embodiment of the present invention provides an adaptive machining method for grinding complex castings, the method comprising the following steps: Obtain the area of ​​the target casting to be polished; All areas to be polished are clustered and divided into multiple polishing clusters; Obtain the polishing difficulty of each area to be polished, and determine the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished; Obtain the set of surfaces for each polishing cluster, and use the set of surfaces for each polishing cluster to obtain the minimum conversion cost for each polishing cluster; The overall cost of each polishing cluster is determined based on the overall polishing difficulty and minimum switching cost of each polishing cluster. Based on the comprehensive cost of each polishing cluster, the processing order of each area to be polished is obtained; Obtain the polishing path for each area to be polished; Polish the areas to be polished according to the processing order and polishing path of each area.

[0006] Furthermore, the specific steps for obtaining the area to be polished of the target casting are as follows: Acquire the point cloud 3D model of the target casting, the CAD design model, and images of each surface of the target casting; The point cloud 3D model of the target casting is compared with the CAD design model to obtain the allowance area; Defect detection is performed on images of each surface of the target casting to obtain defect areas; The areas with excess material and the areas with defects are identified as the areas to be polished in the target casting.

[0007] Furthermore, the specific steps involved in clustering all areas to be polished into multiple polishing clusters are as follows: Machine learning algorithms are used to determine the polishing type for each area to be polished. Based on the grinding type of each area to be ground, determine the grinding parameters for each area to be ground; Based on the tool parameters in the polishing parameters, all areas to be polished are divided into multiple clusters, resulting in multiple polishing clusters.

[0008] Furthermore, the specific steps for obtaining the polishing difficulty of each area to be polished and determining the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished are as follows: For each area to be polished, calculate the gray-level co-occurrence matrix and obtain the gray level and entropy value of the gray-level co-occurrence matrix; The maximum entropy value is determined by using the gray levels of the gray-level co-occurrence matrix; The ratio of the entropy value of the gray-level co-occurrence matrix to the maximum entropy value is used to determine the polishing difficulty of each area to be polished. Obtain the polishing clusters for each surface of the target casting; The average polishing difficulty of all areas to be polished in each polishing cluster of each surface is determined as the polishing difficulty of each polishing cluster of each surface. The polishing difficulty of the surface containing each polishing cluster is summed, and the sum is used to determine the overall polishing difficulty of each polishing cluster.

[0009] Furthermore, the specific steps for obtaining the minimum conversion cost of each polishing cluster using the surface set where each polishing cluster resides include the following: For each set of surfaces containing a polishing cluster, calculate the conversion cost between any two surfaces, and the maximum value among all pairs of conversion costs between surfaces; The fixed conversion cost between two surfaces is determined by the ratio of the conversion cost between any two surfaces to the maximum value among all conversion costs between any two surfaces. Treat each surface as a node, the lines connecting nodes as edges, and the fixed conversion cost between any two surfaces as edge values. Use the shortest path algorithm to obtain the minimum conversion cost for each polishing cluster.

[0010] Furthermore, the specific steps for determining the comprehensive cost of each polishing cluster based on the overall polishing difficulty and minimum conversion cost are as follows: The ratio of the number of elements in the surface set of each grinding cluster to the total number of surfaces of the target casting is determined as the ratio of the number of elements in the surface set of each grinding cluster. The mean of the ratios of all surfaces containing polished clusters is determined as the degree of cluster dispersion, and the difference between 1 and the degree of cluster dispersion is determined as the degree of cluster concentration. The first product is determined by multiplying the cluster concentration degree by the overall polishing difficulty of each polishing cluster. The second product is determined by multiplying the cluster dispersion degree by the minimum conversion cost of each polishing cluster; The sum of the first and second products is used to determine the overall cost for each polishing cluster.

[0011] Furthermore, the specific steps for obtaining the processing order of each area to be polished based on the comprehensive cost of each polishing cluster are as follows: The grinding clusters are sorted according to their overall cost from low to high to obtain the processing order for each grinding cluster. For each grinding cluster, the surfaces containing the grinding cluster are sorted in descending order of the number of areas to be ground, thus obtaining the processing order of the surfaces containing the grinding cluster. For each surface containing a grinding cluster, the areas to be ground are sorted in order of increasing grinding difficulty to obtain the processing order for each area to be ground.

[0012] Furthermore, the specific steps for obtaining the polishing path for each area to be polished are as follows: For each area to be polished, obtain multiple feature points and the corresponding point of each feature point in the CAD design model; The Euclidean distance between each feature point and its corresponding point in the CAD design model is determined as the Euclidean distance between each feature point and its corresponding point. Obtain the maximum value of the Euclidean distance between the feature point and its corresponding point; The first distance for each feature point is determined by the ratio of the Euclidean distance between each feature point and its corresponding point to the maximum value of the Euclidean distance between the feature point and its corresponding point. Obtain multiple corner points in a CAD design model; The minimum Euclidean distance between each feature point and all corner points is determined as the second distance for each feature point; The retention rate of each feature point is determined based on the first and second distances between each feature point. Feature points with a retention rate greater than a preset retention rate threshold are identified as polishing feature points; The TSP algorithm is used to find the shortest path that traverses all the grinding feature points, and the shortest path is determined as the grinding path for each area to be ground.

[0013] Furthermore, the specific steps for determining the retention rate of each feature point based on the first distance and the second distance are as follows: The mean of the first distances of all feature points is determined as the first weight, and the difference between 1 and the first weight is determined as the second weight. The product of the first weight and the first distance of each feature point is determined as the third product; The product of the second weight and the second distance of each feature point is determined as the fourth product; The sum of the third and fourth products is used to determine the retention rate for each feature point.

[0014] One embodiment of the present invention provides an adaptive machining device for grinding complex castings, including a processor and a memory. The processor is used to process instructions stored in the memory to implement the machining process of the following modules: The acquisition module is used to acquire the area to be polished of the target casting; The analysis module is used to cluster all areas to be polished, dividing them into multiple polishing clusters; obtain the polishing difficulty of each area, and determine the overall polishing difficulty of each cluster based on the polishing difficulty of each area; obtain the set of surfaces containing each cluster, and use this set to obtain the minimum conversion cost of each cluster; determine the comprehensive cost of each cluster based on its overall polishing difficulty and minimum conversion cost; obtain the processing order of each area to be polished based on the comprehensive cost of each cluster; and obtain the polishing path for each area. The polishing module is used to polish the areas to be polished according to the processing order and polishing path of each area.

[0015] The beneficial effects of the technical solution of this invention are as follows: This invention proposes an adaptive machining equipment and method for grinding complex castings. It obtains the actual model of the casting through 3D scanning and compares it with the design model to identify all areas to be ground. Next, each area undergoes multi-dimensional feature quantification and process classification, and based on material properties and geometric morphology, it is clustered into several "process families" that can share the same set of grinding methods, tool components, and process parameters. Then, collaborative path planning is performed within each process family: on the one hand, grinding paths that satisfy the geometric characteristics of specific areas are generated; on the other hand, the processing paths for multiple areas within the family are optimized to minimize idle travel and processing time. Finally, cross-process family scheduling optimization is performed at the global level, comprehensively planning the processing sequence, tool change points, and robot posture switching strategies between different process families, thereby achieving the comprehensive goals of minimizing tool changes, smoothing process switching, and shortening the total processing path, thus improving the grinding quality and efficiency of complex castings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of an adaptive machining method for grinding complex castings according to the present invention. Figure 2 This is a schematic diagram of a module of an adaptive machining equipment for grinding complex castings according to the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an adaptive machining equipment and method for grinding complex castings according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details the specific scheme of the adaptive machining equipment and method for grinding complex castings provided by this invention.

[0021] The grinding equipment includes: A robotic arm (articular arm): Composed of multiple joints (rotation axes) connecting arm segments (such as the upper arm and forearm), used to simulate the flexible movement of a human arm, driving the end-effector grinding tool to move precisely in three-dimensional space, achieving grinding operations at different angles and positions. An end-effector (grinding tool assembly): Consists of a drive unit and a grinding head (grinding wheel / sand disc). The drive unit provides rotational power to the grinding head and controls the grinding speed; the grinding head directly contacts the workpiece, achieving deburring, polishing, and weld treatment functions through high-speed rotation. A cabling package: Cables, air pipes, and other pipelines arranged along the robotic arm, transmitting power, signals, or compressed air to the drive motors of the robotic arm joints and the power system of the end-effector. A base: A fixed base at the bottom of the robotic arm, used to support the entire robotic arm and ensure stability during operation. A control system: Usually integrated within the robotic arm or base, acting as the "brain," receiving commands and controlling the robotic arm's movement trajectory, grinding force, speed, etc., while combining sensor data to achieve precise force-controlled grinding.

[0022] For large castings, there are numerous and dispersed areas that need to be ground. At the same time, the shapes of the areas to be ground are complex and varied, with risers and many surface defects (such as shrinkage cavities, sharp edges, cracks, etc.), which increases the time required for grinding trajectory planning and is not conducive to efficient grinding. Different grinding processes have significant differences in the planning of grinding trajectories and grinding intensity. Based on this, this invention classifies defects and uses different grinding tools and methods for different defects. Combined with feature point selection and grinding path optimization, the grinding process is adaptively optimized to improve grinding efficiency.

[0023] Please see Figure 1 The diagram illustrates a flowchart of an adaptive machining method for grinding complex castings according to an embodiment of the present invention. The method includes the following steps: Step S001: Obtain the area to be polished of the target casting.

[0024] It should be noted that: the target casting refers to a large casting that requires grinding.

[0025] Step S001 further includes steps S0011-S0014: Step S0011: Obtain the point cloud 3D model of the target casting, the CAD design model, and images of each surface of the target casting.

[0026] Specifically, a point cloud 3D model of the target casting is obtained through point cloud scanning: a laser scanner is used to scan the target casting from multiple angles to acquire point cloud data of its surface; the point cloud data is then denoised, registered, and triangulated to generate a point cloud 3D model of the target casting. Point cloud scanning is a well-known technique and will not be elaborated upon here.

[0027] The CAD design model is a standard triangular mesh model. Images of each surface of the target casting are obtained using a visible light camera.

[0028] Step S0012: Compare the point cloud 3D model of the target casting with the CAD design model to obtain the allowance area.

[0029] Specifically: By comparing the actual model (point cloud 3D model) and the CAD design model, excess material is obtained and recorded as the excess area; the actual model and the design model are precisely aligned in the spatial coordinate system; then, the distance from each point on the surface of the actual model to the surface of the design model is calculated, and areas where the distance is greater than a preset threshold (such as 0.5mm) are determined to have excess material; finally, all spatially adjacent areas that meet the conditions are automatically clustered and marked as an independent excess area. Step S0013: Perform defect detection on the surface images of the target casting to obtain the defect area.

[0030] Specifically, defect detection is performed on each surface image using a neural network (such as a CNN deep neural network) to obtain the defect region: each surface image is used as input to the CNN deep neural network, which analyzes each surface image and outputs pixel-level defect segmentation results, i.e., the defect region.

[0031] Step S0014: Determine the areas of excess material and defects as the areas to be polished in the target casting.

[0032] Step S002: Cluster all areas to be polished, dividing all areas to be polished into multiple polishing clusters.

[0033] Step S002 further includes steps S0021-S0023: Step S0021: Use machine learning algorithms to determine the polishing type for each area to be polished.

[0034] Specifically, the machine learning algorithm in this embodiment is the random forest classification method. Using the random forest classification method, the polishing type of each area to be polished is obtained, such as: riser residue, flash / edge, surface depression, crack, etc. First, a series of features are extracted from the image and geometric data (such as size, depth, texture) of each area to form a digital feature vector of the area. Then, these feature vectors are input into a pre-trained random forest classification model. The model establishes a mapping relationship between features and types by learning from a large number of labeled polishing samples (such as known "riser residue" and "flash" image data). The model automatically analyzes and votes on the input vectors and quickly outputs the polishing type that the area is most likely to belong to, such as judging it as "surface depression" or "crack".

[0035] Step S0022: Determine the grinding parameters for each area to be ground based on the grinding type of each area.

[0036] Specifically, a process knowledge base is established: expert experience is summarized to obtain grinding parameters for different grinding types, such as: tool parameters: grinding wheel type, size, hardness, etc.; process parameters: spindle speed, feed rate, etc.; and quality parameters: surface roughness, contour accuracy, etc.

[0037] For each grinding type (including defect areas and excess material areas), corresponding grinding parameters are generated. For example, for "riser remnants", the corresponding parameters are: [large diameter ceramic grinding wheel, high power mode, low speed strong cutting, large excess material removal mode]. For "flash / edges", the corresponding parameters are: [medium diameter resin grinding wheel, medium speed dressing mode, contour following, lightweight deburring].

[0038] Step S0023: Based on the tool parameters in the polishing parameters, divide all areas to be polished into multiple clusters to obtain multiple polishing clusters.

[0039] Specifically, the grinding parameters for each area to be ground were obtained, and the areas to be ground were divided into clusters based on the grinding parameters.

[0040] Areas with the same tool parameters in the grinding parameters are recorded as the same grinding cluster. The grinding tools in the same grinding cluster are the same, so there is no need to change the grinding tools. Only the grinding speed or precision needs to be changed.

[0041] Step S003: Obtain the polishing difficulty of each area to be polished, and determine the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished.

[0042] It should be noted that: for all areas to be polished on each surface of the target casting, the areas are sorted according to the estimated polishing difficulty, and the processing paths are planned in order from easy to difficult.

[0043] This sequencing strategy is primarily based on considerations of process stability: the grinding process of castings is often accompanied by significant vibration. If the more challenging areas are machined first, the severe vibration may lead to abnormal tool wear or inaccurate process parameters. By prioritizing simpler areas with less vibration, the system can gradually enter a stable operating state, avoiding severe initial shocks. This reduces frequent tool changes due to unexpected tool wear and ensures the stable operation of feedback systems such as force control. A gradual progression from easy to difficult helps improve the consistency and reliability of the process, ultimately reducing unnecessary downtime and optimizing overall operational efficiency.

[0044] Step S003 further includes steps S0031-S0036: Step S0031: For each area to be polished, calculate the gray-level co-occurrence matrix and obtain the gray level and entropy value of the gray-level co-occurrence matrix.

[0045] It's important to note that the Gray-Level Co-occurrence Matrix (GLCM) is a mathematical tool used to quantify the texture features of an image. It describes the roughness, contrast, and regularity of the texture by statistically analyzing the frequency of simultaneous occurrences of pairs of pixels with a specific spatial relationship (e.g., horizontally adjacent pixels separated by one pixel). GLCM is a well-known technique and will not be elaborated upon here. A higher entropy value indicates greater texture complexity and surface roughness.

[0046] Step S0032: Determine the maximum entropy value using the gray level of the gray-level co-occurrence matrix.

[0047] Specifically, if the number of gray levels N is obtained when calculating the gray-level co-occurrence matrix, then the maximum entropy value is... .

[0048] Step S0033: Determine the polishing difficulty of each area to be polished by the ratio of the entropy value of the gray-level co-occurrence matrix to the maximum entropy value.

[0049] Specifically, firstly, a grayscale image of the surface of the area j to be polished is acquired, and the number of grayscale levels N used for calculation is determined (e.g., 16 or 32 levels), that is, the continuous 256 grayscale levels are compressed into N discrete levels to simplify the calculation; then, based on this number of grayscale levels, a grayscale co-occurrence matrix is ​​constructed by statistically analyzing the grayscale combination probabilities of pairs of pixels with fixed spatial relationships in the image; finally, the entropy value is calculated based on the grayscale co-occurrence matrix, and the maximum value of this entropy value is... The entropy value is normalized by calculating the ratio of the entropy value to the maximum entropy value. The normalized entropy value of each area to be polished is used as the polishing difficulty of that area.

[0050] Step S0034: Obtain the polishing clusters of each surface of the target casting.

[0051] Specifically, after obtaining the polishing difficulty of all areas to be polished on surface k, multiple polishing clusters of surface k are obtained.

[0052] Step S0035: The average value of the polishing difficulty of all areas to be polished in each polishing cluster of each surface is determined as the polishing difficulty of each polishing cluster of each surface.

[0053] Specifically, for a polishing cluster h on surface k, the average polishing difficulty of all polishing areas of polishing cluster h distributed on surface k is obtained as the polishing difficulty of polishing cluster h on surface k.

[0054] Step S0036: Sum the polishing difficulty of the surface where each polishing cluster is located, and determine the sum as the overall polishing difficulty of each polishing cluster.

[0055] Specifically, the polishing cluster h is distributed on surfaces k and f. Therefore, the polishing difficulty of the polishing cluster h on surface k is summed with the polishing difficulty of the polishing cluster h on surface f to obtain the overall polishing difficulty of the polishing cluster h.

[0056] Step S004: Obtain the set of surfaces where each polishing cluster is located, and use the set of surfaces where each polishing cluster is located to obtain the minimum conversion cost of each polishing cluster.

[0057] It should be noted that when determining the polishing order of the polishing clusters, two core optimization objectives need to be balanced: Prioritize process stability (microscopically, from easy to difficult): On the same surface, prioritize grinding areas that are easier to grind in order to reduce initial vibration and stabilize the process system.

[0058] Global efficiency priority ("reducing clamping and switching" on a macro level): At the global level, prioritize the processing of clusters that are "low in overall difficulty" and "spatially concentrated" to minimize the number and magnitude of repositioning (rotation / movement) of the robot or workpiece (such as a positioner), thereby reducing non-processing time.

[0059] In other words, for a defect cluster to be polished, its "priority" is inversely proportional to the "overall polishing difficulty" and directly proportional to the "spatial distribution concentration". That is, the easier the cluster is to polish and the more concentrated its distribution, the higher the priority should be given to processing.

[0060] Obtain the set of surfaces where each grinding cluster is located: Determine which surfaces of the target casting all areas to be ground in grinding cluster h are distributed (such as surface k, surface f, surface b), then the set of surfaces where grinding cluster h is located is F = {kfb}.

[0061] Establish an inter-surface transformation model: Obtain the optimal flipping path between any two related surfaces, quantified by the minimum composite angle (or axial motion) required for the positioner to rotate. For example, a 90° rotation around the X-axis is required to move from surface k to surface f.

[0062] Step S004 further includes steps S0041-S0043: Step S0041: For each set of surfaces containing a grinding cluster, calculate the conversion cost between any two surfaces, and the maximum value among all the conversion costs between any two surfaces.

[0063] Specifically, for the set of surfaces F={kfb} containing the grinding cluster h, calculate the conversion cost between each pair of surfaces, expressed as the sum of the conversion angles. For example, to move from surface k to surface f, a horizontal rotation of 45° and a vertical rotation of 70° are required, so the conversion cost is 45° + 70° = 115°. The conversion cost from surface k to surface b is 130°, and the conversion cost from surface f to surface b is 90°. Therefore, the maximum conversion cost among all pairs of surfaces is 130°.

[0064] Step S0042: Determine the fixed conversion cost between two surfaces as the ratio of the conversion cost between any two surfaces to the maximum value among all conversion costs between any two surfaces.

[0065] Specifically, each conversion cost is normalized by the maximum conversion cost: the ratio of the conversion cost 90° from surface f to surface b to the maximum conversion cost 130° is determined as the fixed conversion cost between surfaces f and b.

[0066] Step S0043: Treat each surface as a node, the connection between nodes as an edge, and the fixed conversion cost between any two surfaces as the edge value. Use the shortest path algorithm to obtain the minimum conversion cost for each polishing cluster.

[0067] Specifically, the shortest path algorithm is used to quickly find the route with the minimum cumulative cost from one starting point to another ending point in a weighted graph (such as a road network or communication network). The shortest path algorithm is a well-known technique and will not be elaborated upon here. Through the shortest path algorithm, the shortest path from one surface to all surfaces is obtained. The shortest path can represent the minimum transformation cost to complete all areas to be polished within a polishing cluster.

[0068] The shortest path algorithm in this embodiment is the ant colony algorithm: the surface containing the polishing cluster h is abstracted as a path node, and the fixed transformation cost between any two surfaces is defined as the edge value connecting the two nodes, thus constructing a weighted network graph representing the transformation relationship between surfaces. Subsequently, the system calls the ant colony algorithm to solve this network: the algorithm simulates multiple "ants" moving between nodes according to the optimal transformation cost and releasing pheromones. Through multiple iterations, pheromones gradually accumulate on the low-cost path, eventually guiding the algorithm to converge to the shortest path that visits all nodes and has the lowest total transformation cost. The total cost of this path is the minimum transformation cost to complete all the areas to be polished in a polishing cluster, and its corresponding node sequence is the optimal surface visiting order.

[0069] Step S005: Determine the comprehensive cost of each polishing cluster based on the overall polishing difficulty and minimum conversion cost of each polishing cluster.

[0070] It should be noted that the overall cost of a polishing cluster = w1 × overall polishing difficulty of the polishing cluster + w2 × minimum conversion cost of the polishing cluster, where w1 represents the cluster concentration and w2 represents the cluster dispersion. For each polishing cluster, the more dispersed the elements in the cluster are, and the more different surfaces they are distributed on, the more the conversion cost needs to be considered.

[0071] Step S005 further includes steps S0051-S0055: Step S0051: Determine the ratio of the number of elements in the surface set of each grinding cluster to the total number of surfaces of the target casting as the ratio of the number of elements in the surface set of each grinding cluster.

[0072] Specifically, for the set of surfaces F={kfb} containing the grinding cluster h, where the number of elements is 3 and the total number of surfaces of the target casting is 4, then 3 / 4 is determined as the ratio of the surfaces containing the grinding cluster h. The larger the ratio, the more surfaces the different areas to be ground of this grinding cluster are distributed on, and the more the conversion cost needs to be considered.

[0073] Step S0052: The mean of the ratios of all surfaces containing polished clusters is determined as the cluster dispersion degree, and the difference between 1 and the cluster dispersion degree is determined as the cluster concentration degree.

[0074] Specifically, the ratio of the surface containing each polishing cluster is calculated, and then the mean of the ratios of all polishing clusters is calculated. The mean ratio is taken as the cluster dispersion degree w2. The larger w2 is, the greater the dispersion of the overall distribution of all clusters. (1-w2) is denoted as the cluster concentration degree w1.

[0075] Step S0053: The product of the cluster concentration degree and the overall polishing difficulty of each polishing cluster is determined as the first product.

[0076] Specifically, the overall polishing difficulty of the w1×polishing cluster is denoted as the first product.

[0077] Step S0054: Determine the second product by multiplying the cluster dispersion degree by the minimum conversion cost of each polishing cluster.

[0078] Specifically, the minimum conversion cost of w2×polishing cluster is denoted as the second product.

[0079] Step S0055: The sum of the first product and the second product is used to determine the overall cost for each polishing cluster.

[0080] Specifically, the overall cost of a polishing cluster = w1 × the overall polishing difficulty of the polishing cluster + w2 × the minimum conversion cost of the polishing cluster.

[0081] Step S006: Based on the comprehensive cost of each polishing cluster, obtain the processing order of each area to be polished.

[0082] Step S006 further includes steps S0061-S0063: Step S0061: Sort the polishing clusters according to the order of comprehensive cost from low to high, and obtain the processing order of each polishing cluster.

[0083] Specifically, the grinding clusters are sorted in ascending order of comprehensive cost. The lower the comprehensive cost, the lower the overall processing cost of the grinding cluster, and the higher the priority should be given to processing it. The processing order of each grinding cluster is obtained according to the ascending order of comprehensive cost, and each grinding cluster is processed in order of comprehensive cost from low to high.

[0084] Step S0062: For each grinding cluster, sort the surfaces containing the grinding clusters in descending order of the number of areas to be ground to obtain the processing order of the surfaces containing the grinding clusters.

[0085] Specifically, for the grinding cluster h: the surfaces containing the grinding cluster are processed sequentially in descending order of the number of areas to be ground.

[0086] Step S0063: For each surface where a grinding cluster is located, sort the areas to be ground in order of increasing grinding difficulty to obtain the processing order of each area to be ground.

[0087] Specifically, for a grinding cluster h, there are n areas to be ground on the surface f. These areas are sorted in order of increasing grinding difficulty as the grinding order.

[0088] If the grinding clusters are distributed on other surfaces, the positioner is planned to rotate or the robot is moved to the next surface. The grinding sequence of different areas to be ground on the surface is determined by the grinding difficulty, until all surfaces of the grinding cluster are ground.

[0089] After completing the current sanding cluster, switch to the next highest priority sanding cluster according to the processing order of sanding clusters, and complete the sanding.

[0090] Step S007: Obtain the polishing path for each area to be polished.

[0091] It should be noted that for any area to be polished, the polishing trajectory can be quickly planned by using the key feature points of the surface of the target casting to be polished.

[0092] This embodiment achieves the grinding of the area to be ground by grinding feature points. The grinding trajectory should reflect the basic shape of the grinding surface as much as possible, which requires selecting appropriate feature points for grinding.

[0093] This embodiment combines the feature points with their positions in the CAD design model to obtain the correction degree. The greater the correction degree, the more the corresponding feature point needs to be polished, and the more the corresponding feature point needs to be retained. At the same time, combined with the corner point positions in the CAD design model, feature points near the corner points are retained, which helps to polish the casting to meet the CAD design model. Then, the feature points that need to be retained are selected for polishing based on the degree of polishing required and their proximity to the corner points.

[0094] Step S007 further includes steps S0071-S0079: Step S0071: For each area to be polished, obtain multiple feature points and the corresponding point of each feature point in the CAD design model.

[0095] Specifically, for each polished area, feature points are obtained through SIFT detection: First, the surface image of the polished area m is acquired, and then the SIFT (Scale Invariant Feature Transform) algorithm is used for automatic analysis. This algorithm simulates the process of the human eye observing an image at different scales, and can stably locate local extreme points in the image that have significant texture, distinct corners, or edge intersections. These points are the "feature points" that characterize the contour morphology of the area. The SIFT algorithm is a well-known technique and will not be elaborated upon here.

[0096] For each feature point, the corresponding point is found on the CAD design model through normal projection. Here, the corresponding point refers to the intersection point obtained by projecting a feature point extracted from the actual casting surface perpendicularly onto the design model (CAD) surface along the surface normal direction (or the opposite direction) at that point. Simply put, it's about finding the theoretically correct location of each "marker point" on the actual object on the design blueprint.

[0097] Step S0072: Determine the Euclidean distance between each feature point and its corresponding point in the CAD design model.

[0098] Specifically, the Euclidean distance between each feature point and its corresponding point is denoted as d. The larger d is, the more polishing is required for that feature point.

[0099] Step S0073: Obtain the maximum value of the Euclidean distance between the feature point and the corresponding point.

[0100] Step S0074: Determine the first distance for each feature point by the ratio of the Euclidean distance between each feature point and its corresponding point to the maximum value of the Euclidean distance between the feature point and its corresponding point.

[0101] Specifically, the first distance of each feature point is denoted as D. The distance is then normalized using the maximum value of d, and subsequent calculations are performed using this normalized value.

[0102] Step S0075: Obtain multiple corner points in the CAD design model.

[0103] Specifically, corner points are extracted from the CAD design model, that is, some sharp vertices or points with large curvature on the model are obtained: by analyzing the geometric relationship between each vertex and its adjacent face in the CAD design model, vertices with abrupt curvature changes or sharp angles between adjacent faces are automatically identified; specifically, when the angle between the normal vectors of the adjacent triangular faces at a vertex is less than a preset threshold (e.g., 150 degrees), or when the curvature of the vertex in a local area is significantly higher than that of the surrounding area, it is determined to be a "corner point" that represents a key turning point of the model contour.

[0104] Step S0076: Determine the minimum Euclidean distance between each feature point and all corner points as the second distance for each feature point.

[0105] Specifically, the second distance of each feature point is denoted as L. The smaller L is, the more necessary it is to grind the feature point, which helps to shape the key contours of the casting.

[0106] Step S0077: Determine the retention rate of each feature point based on the first distance and the second distance of each feature point.

[0107] Step S0077 further includes steps S0771-S0774: Step S0771: The mean of the first distances of all feature points is determined as the first weight, and the difference between 1 and the first weight is determined as the second weight.

[0108] Specifically, Q1 is denoted as the first weight. The larger this value is, the more the Euclidean distance between the feature point and its corresponding point needs to be considered when retaining the feature points. 1-Q1 is denoted as the second weight.

[0109] Step S0772: The product of the first weight and the first distance of each feature point is determined as the third product.

[0110] Specifically, Q1×D is denoted as the third product.

[0111] Step S0773: The product of the second weight and the second distance of each feature point is determined as the fourth product.

[0112] Specifically, (1-Q1)×L is denoted as the fourth product.

[0113] Step S0774: The sum of the third and fourth products is used to determine the retention rate for each feature point.

[0114] Specifically, for each feature point, the result of the calculation of Q1×D+(1-Q1)×L is used as the retention degree of that feature point.

[0115] Step S0078: Feature points with a retention rate greater than the preset retention rate threshold are identified as polishing feature points.

[0116] It should be noted that the preset retention threshold is set according to specific circumstances, and a value of 0.6 is preferred here. Feature points with a retention rate greater than 0.6 are retained as polishing feature points.

[0117] Step S0079: Use the TSP algorithm to find the shortest path that traverses all grinding feature points, and determine the shortest path as the grinding path for each area to be ground.

[0118] Specifically, the Traveling Salesman Problem (TSP) algorithm aims to find the shortest possible route for a traveling salesman to visit all cities once and return to the starting point. It is a classic mathematical method for optimizing the order of paths and is a well-known technique, so it will not be elaborated on here.

[0119] First, multiple polishing feature points are considered as "stations" that must be visited, and the spatial distance between stations is defined as "journey cost". Then, the Traveling Salesman Problem (TSP) algorithm is used to calculate the point set and automatically find a continuous visiting sequence that can connect all feature points and has the shortest total travel distance. Finally, the system directly converts this optimized path into a polishing trajectory that the robot can execute, that is, the polishing path of each area to be polished.

[0120] Step S008: Polish the areas to be polished according to the processing order and polishing path of each area to be polished.

[0121] After obtaining the polishing path, control the polishing robot to execute the polishing steps: System Readiness and Calibration: Workpiece Confirmation: Ensure the casting is fixed in the programmed position; Tool Calibration: Accurately measure the tip position of the grinding tool; Sensor Preparation: If force control is used, perform zero-point calibration on the force sensor; Safety Settings: Set the robot's movement range and safety zone.

[0122] The robot executes each segment of the grinding path sequentially: Rapid positioning: moves unloaded to a safe position near the starting point of the current grinding area; Contacting the workpiece: slowly approaches the workpiece surface; Force control mode: switches to force control after contact, maintains preset pressure and begins grinding; Position mode: moves strictly according to the programmed trajectory; Following grinding: moves along the planned path and automatically adjusts parameters such as spindle speed and feed rate according to different areas, and fine-tunes the posture in real time based on force feedback (in force control) or compensates for tool length based on the wear model; Tool lifting and transfer: after completing the current segment, lifts the tool and quickly moves to the starting point of the next area.

[0123] Process monitoring: Real-time monitoring: Monitors signals such as spindle load, grinding force, and vibration; Abnormal handling: Automatically pauses or reverts when encountering overload or abnormality.

[0124] Task completed: After all paths have been executed, the robot returns to a safe position and the system is ready to operate.

[0125] Please see Figure 2 , Figure 2 This is a schematic diagram of a module for an adaptive grinding machine for complex castings according to the present invention, including a processor and a memory. The processor is used to process instructions stored in the memory to implement the monitoring process of the following modules: The acquisition module 100 is used to acquire the area to be polished of the target casting.

[0126] The analysis module 200 is used to cluster all areas to be polished, dividing them into multiple polishing clusters; obtain the polishing difficulty of each area to be polished, and determine the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished; obtain the set of surfaces in which each polishing cluster is located, and obtain the minimum conversion cost of each polishing cluster using the set of surfaces in which each polishing cluster is located; determine the comprehensive cost of each polishing cluster based on the overall polishing difficulty and minimum conversion cost of each polishing cluster; obtain the processing order of each area to be polished based on the comprehensive cost of each polishing cluster; and obtain the polishing path of each area to be polished.

[0127] The polishing module 300 is used to polish the areas to be polished according to the processing order and polishing path of each area to be polished.

[0128] In summary, in this embodiment of the invention, a 3D scanning model of the casting is obtained and compared with the design model to identify all areas to be polished. Next, each area undergoes multi-dimensional feature quantification and process classification, and based on material properties and geometric morphology, it is clustered into several "process families" that can share the same set of polishing methods, tool components, and process parameters. Then, collaborative path planning is performed within each process family: on the one hand, polishing paths that satisfy the specific geometric characteristics of the area are generated; on the other hand, the processing paths for multiple areas within the family are optimized to minimize idle travel and processing time. Finally, cross-process family scheduling optimization is performed at the global level, comprehensively planning the processing sequence, tool change points, and robot posture switching strategies between different process families, thereby achieving the comprehensive goals of minimizing tool changes, smoothing process switching, and shortening the total processing path, thus improving the polishing quality and efficiency of complex castings.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive machining method for grinding complex castings, characterized in that, The method includes the following steps: Obtain the area of ​​the target casting to be polished; All areas to be polished are clustered and divided into multiple polishing clusters; Obtain the polishing difficulty of each area to be polished, and determine the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished; Obtain the set of surfaces for each polishing cluster, and use the set of surfaces for each polishing cluster to obtain the minimum conversion cost for each polishing cluster; The overall cost of each polishing cluster is determined based on the overall polishing difficulty and minimum switching cost of each polishing cluster. Based on the comprehensive cost of each polishing cluster, the processing order of each area to be polished is obtained; Obtain the polishing path for each area to be polished; Polish the areas to be polished according to the processing order and polishing path of each area.

2. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The specific steps for obtaining the grinding area of ​​the target casting are as follows: Acquire the point cloud 3D model of the target casting, the CAD design model, and images of each surface of the target casting; The point cloud 3D model of the target casting is compared with the CAD design model to obtain the allowance area; Defect detection is performed on the surface images of the target casting to obtain the defect areas; The areas with excess material and the areas with defects are identified as the areas to be polished in the target casting.

3. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The specific steps involved in clustering all areas to be polished into multiple polishing clusters are as follows: Machine learning algorithms are used to determine the polishing type for each area to be polished. Based on the grinding type of each area to be ground, determine the grinding parameters for each area to be ground; Based on the tool parameters in the polishing parameters, all areas to be polished are divided into multiple clusters, resulting in multiple polishing clusters.

4. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The specific steps for obtaining the polishing difficulty of each area to be polished, and determining the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished, are as follows: For each area to be polished, calculate the gray-level co-occurrence matrix and obtain the gray level and entropy value of the gray-level co-occurrence matrix; The maximum entropy value is determined by using the gray levels of the gray-level co-occurrence matrix; The ratio of the entropy value of the gray-level co-occurrence matrix to the maximum entropy value is used to determine the polishing difficulty of each area to be polished. Obtain the polishing clusters for each surface of the target casting; The average polishing difficulty of all areas to be polished in each polishing cluster of each surface is determined as the polishing difficulty of each polishing cluster of each surface. The polishing difficulty of the surface containing each polishing cluster is summed, and the sum is used to determine the overall polishing difficulty of each polishing cluster.

5. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The specific steps for obtaining the minimum conversion cost of each polishing cluster by utilizing the surface set where each polishing cluster belongs are as follows: For each set of surfaces containing a polishing cluster, calculate the conversion cost between any two surfaces, and the maximum value among all pairs of conversion costs between surfaces; The fixed conversion cost between two surfaces is determined by the ratio of the conversion cost between any two surfaces to the maximum value among all conversion costs between any two surfaces. Treat each surface as a node, the lines connecting nodes as edges, and the fixed conversion cost between any two surfaces as edge values. Use the shortest path algorithm to obtain the minimum conversion cost for each polishing cluster.

6. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The process of determining the comprehensive cost of each polishing cluster based on its overall polishing difficulty and minimum conversion cost includes the following specific steps: The ratio of the number of elements in the surface set of each grinding cluster to the total number of surfaces of the target casting is determined as the ratio of the number of elements in the surface set of each grinding cluster. The mean of the ratios of all surfaces containing polished clusters is determined as the degree of cluster dispersion, and the difference between 1 and the degree of cluster dispersion is determined as the degree of cluster concentration. The first product is determined by multiplying the cluster concentration degree by the overall polishing difficulty of each polishing cluster. The second product is determined by multiplying the cluster dispersion degree by the minimum conversion cost of each polishing cluster; The sum of the first and second products is used to determine the overall cost for each polishing cluster.

7. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The specific steps for obtaining the processing order of each area to be polished based on the comprehensive cost of each polishing cluster are as follows: The grinding clusters are sorted according to their overall cost from low to high to obtain the processing order for each grinding cluster. For each grinding cluster, the surfaces containing the grinding cluster are sorted in descending order of the number of areas to be ground, thus obtaining the processing order of the surfaces containing the grinding cluster. For each surface containing a grinding cluster, the areas to be ground are sorted in order of increasing grinding difficulty to obtain the processing order for each area to be ground.

8. The adaptive machining method for grinding complex castings according to claim 1, characterized in that, The specific steps for obtaining the polishing path for each area to be polished are as follows: For each area to be polished, obtain multiple feature points and the corresponding point of each feature point in the CAD design model; The Euclidean distance between each feature point and its corresponding point in the CAD design model is determined as the Euclidean distance between each feature point and its corresponding point. Obtain the maximum value of the Euclidean distance between the feature point and its corresponding point; The first distance for each feature point is determined by the ratio of the Euclidean distance between each feature point and its corresponding point to the maximum value of the Euclidean distance between the feature point and its corresponding point. Obtain multiple corner points in a CAD design model; The minimum Euclidean distance between each feature point and all corner points is determined as the second distance for each feature point; The retention rate of each feature point is determined based on the first and second distances between each feature point. Feature points with a retention rate greater than a preset retention rate threshold are identified as polishing feature points; The TSP algorithm is used to find the shortest path that traverses all the grinding feature points, and the shortest path is determined as the grinding path for each area to be ground.

9. The adaptive machining method for grinding complex castings according to claim 8, characterized in that, The specific steps for determining the retention rate of each feature point based on the first and second distances are as follows: The mean of the first distances of all feature points is determined as the first weight, and the difference between 1 and the first weight is determined as the second weight. The product of the first weight and the first distance of each feature point is determined as the third product; The product of the second weight and the second distance of each feature point is determined as the fourth product; The sum of the third and fourth products is used to determine the retention rate for each feature point.

10. An adaptive machining equipment for grinding complex castings, characterized in that, Includes a processor and a memory, the processor being used to process instructions stored in the memory to implement the processing of the following modules: The acquisition module is used to acquire the area to be polished of the target casting; The analysis module is used to cluster all areas to be polished, dividing all areas to be polished into multiple polishing clusters; Obtain the polishing difficulty of each area to be polished, and determine the overall polishing difficulty of each polishing cluster based on the polishing difficulty of each area to be polished; obtain the set of surfaces where each polishing cluster is located, and use the set of surfaces where each polishing cluster is located to obtain the minimum conversion cost of each polishing cluster; determine the comprehensive cost of each polishing cluster based on the overall polishing difficulty and minimum conversion cost of each polishing cluster; obtain the processing order of each area to be polished based on the comprehensive cost of each polishing cluster; obtain the polishing path of each area to be polished. The polishing module is used to polish the areas to be polished according to the processing order and polishing path of each area.

Images