Method for determining laser cladding process parameters based on defect characteristics and similarity measurement

Abstract

The present invention belongs to the field of component processing technology and discloses a method for determining laser cladding process parameters based on defect characteristics and similarity metrics. The method comprises: collecting a two-dimensional image of the defect on the component to be processed, preprocessing the collected two-dimensional image to obtain characteristic values of the defect image; inputting the obtained characteristic values of the defect image into a pre-built database; performing preliminary matching between the defect model of the component to be processed and the models in the defect characteristic library through knowledge reasoning in the database to obtain multiple preliminary matching models; and determining the laser cladding parameters in the process parameter library corresponding to the defect model with the highest similarity among the multiple preliminary matching models as the laser cladding parameters for the defect on the component to be processed. The present invention can effectively reduce trial and error time and improve the efficiency of obtaining laser cladding process parameters.

Description

Technical Field

[0001] The present invention belongs to the technical field of parts processing, and in particular relates to a method for determining laser cladding process parameters based on defect characteristics and similarity measurement. Background Art

[0002] Currently, parts and components develop various defects such as cracks, pores, wear, deformation, and pitting during use. These defects seriously affect the mechanical properties and normal service life of the parts and components. However, defective parts and components are not all useless waste. Advanced additive remanufacturing technology can be used to repair the parts and remove the defects. The repaired parts can still be used as normal products. Laser cladding forming technology is an advanced additive remanufacturing technology involving multiple disciplines. Its principle is to use a high-power density laser beam to locally melt the metal surface to form a molten pool. At the same time, powder feeding equipment is used to add coating powder to the molten pool and fuse it with the substrate surface, thereby rapidly solidifying to form a new surface metal layer with excellent performance.

[0003] When using laser cladding to repair parts, it is necessary to determine the process parameters that have a greater impact on the cladding process, such as laser power, scanning speed, and powder feeding speed. For the same part, if there are different defects, the laser cladding repair process parameters are also different. For example, when repairing wear, high-power laser cladding repair process parameters are required to clad a layer of iron-based, cobalt-based, or nickel-based alloy materials on the failed part of the part surface, so that the parts with the cladding alloy layer can be restored to their original mechanical properties. When repairing cracks, excessive laser power is not required. Instead, the powder feeding speed and scanning speed need to be appropriately increased. Therefore, the scanning speed, powder feeding speed, laser frequency, defocus amount and other process parameters of the laser cladding repair process will also be adjusted differently according to the different defects of the parts to achieve repair. Limited by knowledge rules and reasoning ability, the traditional laser cladding process takes a long time to trial and error to determine the process parameters.

[0004] Through the above analysis, the problems and defects of the existing technology are as follows:

[0005] (1) The existing methods for determining laser cladding process parameters are computationally intensive and costly, with low efficiency and high error rate.

[0006] (2) Existing matching methods are insensitive to changes in model scale and cannot distinguish individual models;

[0007] (3) The existing laser cladding process parameter design is inefficient and process reuse is difficult. Summary of the Invention

[0008] In view of the problems existing in the prior art, the present invention provides a method for determining laser cladding process parameters based on defect characteristics and similarity measurement.

[0009] The present invention is implemented as follows: a method for determining laser cladding process parameters based on defect characteristics and similarity measurement, the method for determining laser cladding process parameters based on defect characteristics and similarity measurement comprising:

[0010] Step 1: collecting a two-dimensional image of the defect of the part to be processed, and preprocessing the collected two-dimensional image of the defect of the part to be processed to obtain a feature value of the image of the defect of the part to be processed;

[0011] Step 2: Input the obtained feature values of the defect image of the part to be processed into a pre-built database; the database performs preliminary matching between the defect model of the part to be processed and the models in the defect feature library through knowledge reasoning to obtain multiple preliminary matching models;

[0012] Step three: determining the laser cladding parameters in the process parameter library corresponding to the defect model with the highest similarity among the multiple preliminary matching models as the laser cladding parameters of the defect of the component to be processed.

[0013] Furthermore, in step 1, collecting a two-dimensional image of the defect of the component to be processed, and preprocessing the collected two-dimensional image of the defect of the component to be processed includes:

[0014] (1) The part to be processed is placed on the inspection platform, and the ultrasonic probe placed nearby is used to contact the surface defects of the part. A pulse signal is fed back each time a different position of the surface defect of the part is touched; a two-dimensional image of the defect of the part to be processed is collected through the feedback pulse signal;

[0015] (2) storing the collected two-dimensional images of defects of the parts to be processed, and converting all the two-dimensional images of defects of the parts to be processed into three-dimensional images by using coordinate transformation, and projecting the three-dimensional images of defects of the parts to be processed into two-dimensional space by using the three-dimensional space spherical coordinate transformation method through Prewitt edge extraction, MEM binarization or other image feature extraction methods to generate multi-view two-dimensional images;

[0016] (3) Using a deep convolutional neural network to extract and classify image features of the multi-view two-dimensional images.

[0017] Furthermore, in step 2, the database construction method includes:

[0018] First, the component defect models, including cracks, pores, wear, and scratches, are determined, and a defect model library is constructed. The image feature values of defect length, circularity, and depth are obtained for the crack defect model, and the image feature values of circularity, depth, and pore diameter are obtained for the pore defect model.

[0019] Secondly, obtain the defect equivalent area circle radius, circularity, texture entropy, and second-order moment image eigenvalues of the wear defect model; obtain the defect area and depth image eigenvalues of the scratch defect model;

[0020] Then, a defect feature library is constructed based on the acquired image feature values; and laser cladding repair parameters corresponding to different image feature values in the defect feature library are determined, and a process parameter library is constructed based on the laser cladding repair parameters;

[0021] Finally, a database is constructed based on the defect model library, defect feature library and process parameter library.

[0022] Furthermore, in step 2, the database performs preliminary matching between the defect model of the component to be processed and the model in the defect feature library through knowledge reasoning, including:

[0023] 1) Using a 3D point cloud recognition and matching algorithm, the point cloud feature representations of the defect model of the part to be processed and the model in the defect feature library are obtained respectively;

[0024] 2) calculating the Euclidean distance between each model in the defect feature library and the defect model of the component to be processed using the point cloud feature representation, and calculating the similarity using the Euclidean distance;

[0025] 3) Sort the models in the defect feature library by similarity from large to small, and select the M models with the largest similarity as the preliminary matching models, as follows:

[0026]

[0027]

[0028] Among them, Hl A =[Hl A1 ,Hl A2 ,Hl A3 …Hl An ] represents the feature identifier of any model A in the defect feature library; H1 B =[Hl B1 ,Hl B2 ,Hl B3 …Hl Bn ] represents the feature identifier of the defect model B to be matched; Sim represents the similarity between the defect model of the part to be processed and the feature identifier of the model point cloud in the defect feature library; the value range of M is 3 to 10;

[0029] 4) Repeat step 3) to obtain the similarity between each model in the defect feature library and the defect model of the part to be processed.

[0030] Furthermore, in step 1), using a three-dimensional point cloud recognition and matching algorithm to respectively obtain the point cloud feature representations of the defect model of the component to be processed and the model in the defect feature library includes:

[0031] 1.1) The point cloud library reads any OBJ file of a 3D model in the defect feature library, extracts the point cloud dataset of the 3D model, and uses the ISS extraction method to extract the key points in the point cloud dataset:

[0032] 1.1.1) Establish an LRF for any point in the point cloud dataset and determine a spherical search range with radius r centered at the point as the neighborhood of the point. All points within the neighborhood are considered to be neighboring points.

[0033] 1.1.2) Use the following formula to calculate the weight of each neighboring point in the neighborhood:

[0034]

[0035] 1.1.3) Calculate the covariance matrix of each neighboring point in the neighborhood using the following formula:

[0036]

[0037] 1.1.4) Calculate the eigenvalues {λ1,λ2,λ3} of the covariance matrix of each neighboring point, where λ1≥λ2≥λ3;

[0038] 1.1.5) Set thresholds ε1 and ε2, and if the following equation is satisfied, the point is considered a key point p j :

[0039]

[0040] 1.2) Search the neighborhood of any key point in the point cloud dataset, construct a LRF based on the key point, and translate the coordinates of the neighboring points of the key point to the LRF;

[0041] 1.3) Project the adjacent points along the three LRF coordinate axes to obtain three frames of projection point cloud grid statistics; divide the projection point cloud grid statistics into N P ×N P grids, obtaining points and the coordinates of the points in each grid, and obtaining a discrete projection statistical graph;

[0042] 1.4) Calculate n in the projection statistics graph i and n j Angle between <n i , n j >, where n j Represents the normal of the neighboring point, n i Represents the normal of the key point; <n i , nj The value range of > is divided into N using [0, π] q (The present invention sets N q =100) subintervals, count the points distributed in each subinterval of the grid; regard each subinterval of the grid as a surface element, and there are N p Each bin is considered to have N p boxes, each box contains a measurement value; expand the grid map into 1×N in a certain order q ×N P ×N P dimensional array, normalized to generate the normal histogram H n ;

[0043] 1.5) Calculate the average curvature of each grid in the projection statistics map, assign the value of the average curvature to each grid; assign the value of 1 to the grid without points; each grid has a measurement value, and the grid map is expanded to 1×N in a certain order p ×N p dimensional array, normalized to generate the curvature histogram H a ;

[0044] 1.6) Calculate the average density of each grid point in the projection statistical map, assign the average density value of the points to each grid, and assign a value of 1 to grids without points; each grid has a measurement value, and the grid map is expanded to 1×N in a certain order. p ×N p dimensional array, normalized to generate the average density histogram H b ;

[0045] 1.7) H from three frames n 、H a and H b The arrays are stitched together as follows:

[0046] H=[c1H n ,c2H a ,c3H b ];

[0047] Among them, H represents the final histogram of the representation, c1, c2 and c 3分别 Represents the weight used to adjust the proportion of normal, curvature and average density in feature description; the curvature weight c2 is equal to the average density weight c3;

[0048] 1.8) Repeat steps 1.1) to 1.7) n times to generate a histogram of the representation for each key point of the 3D model, and collect the obtained n histograms of the representation to form a new feature set Hl = [H1, H2, H3 ... H n ], the feature set is a point cloud feature representation of the three-dimensional model.

[0049] Another object of the present invention is to provide a laser cladding process parameter determination system based on defect characteristics and similarity measurement that implements the laser cladding process parameter determination method based on defect characteristics and similarity measurement, the laser cladding process parameter determination system based on defect characteristics and similarity measurement comprising:

[0050] Detection system and database;

[0051] The detection system includes: an image acquisition module and an image processing module;

[0052] The image acquisition module includes: a detection platform, an ultrasonic probe, a control machine, and an acquisition module; it is used to acquire two-dimensional images of defects in the parts to be processed;

[0053] The image processing module includes: a pixel processing unit, an image filtering and noise reduction unit, and a visual feature extraction unit; it is used to process the collected image of the part to be processed to obtain the defect contour;

[0054] The database is used to construct a defect model library, a defect feature library and a process parameter library to store defect model information, defect feature information and process parameter information respectively.

[0055] Furthermore, the image acquisition module includes:

[0056] Testing platform, used to place parts to be processed and ultrasonic probes;

[0057] The ultrasonic probe is used to contact the surface defects of the parts to be processed and feedback a pulse signal to the control machine each time it contacts a different position of the surface defect of the parts;

[0058] A control unit is used to control the motion trajectory of the ultrasonic probe on the detection platform; and to control the speed and position parameters of the ultrasonic probe on the detection platform;

[0059] The acquisition module is used to receive feedback pulses, collect two-dimensional images of component defects through the received feedback pulses, and transmit the collected images to the image processing module.

[0060] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for determining laser cladding process parameters based on defect characteristics and similarity metrics.

[0061] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the method for determining laser cladding process parameters based on defect characteristics and similarity metrics.

[0062] Another object of the present invention is to provide an information data processing terminal, which is used to implement the laser cladding process parameter determination system based on defect characteristics and similarity measurement.

[0063] In combination with the above technical solutions and the technical problems solved, please analyze the advantages and positive effects of the technical solutions to be protected by the present invention from the following aspects:

[0064] First, in view of the technical problems existing in the above-mentioned prior art and the difficulty of solving these problems, this paper closely combines the technical solutions to be protected by the present invention and the results and data during the research and development process, and analyzes in detail and in depth how the technical solutions of the present invention solve the technical problems and some creative technical effects brought about by solving the problems. The specific description is as follows:

[0065] After obtaining the preliminary matching model, the present invention combines the laser cladding process parameter database to determine the defect repair process parameters and selects them as the final matching model. This solves the defects of the existing matching method that are insensitive to changes in model scale and cannot distinguish individual models. At the same time, it makes full use of the laser cladding process parameter database to solve the bottlenecks of low design efficiency and difficulty in process reuse for designers, provides scientific guidance for laser cladding process design and recommends appropriate (optimal) laser cladding process parameters according to defects.

[0066] Second, considering the technical solution as a whole or from the perspective of the product, the technical effects and advantages of the technical solution to be protected by the present invention are described in detail as follows:

[0067] This invention can effectively reduce trial-and-error time and improve the efficiency of obtaining laser cladding process parameters. By constructing a similarity calculation function, extracting appropriate feature indicators, and calculating the similarity between defect feature models, the invention then finds a set of models similar to a given model based on the similarity values. Finally, matching models are displayed based on the similarity evaluation results. The laser cladding process corresponding to the defect with the highest similarity is recommended as the laser cladding process for the defect in the component to be repaired, significantly reducing computational effort and costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Schematic diagram of a method for determining laser cladding process parameters based on defect characteristics and similarity metrics provided by an embodiment of the present invention;

[0069] Figure 2This is a flow chart of a method for determining laser cladding process parameters based on defect characteristics and similarity metrics provided by an embodiment of the present invention;

[0070] Figure 3 3D data CNN model provided by an embodiment of the present invention;

[0071] Figure 4 This is a flowchart of detecting defects in parts to be repaired provided by an embodiment of the present invention

[0072] Figure 5 This is a flow chart of a three-dimensional point cloud recognition and matching algorithm provided by an embodiment of the present invention;

[0073] Figure 6 is the final representation histogram provided by the embodiment of the present invention;

[0074] Figure 7 This is a comparison diagram of the preferred matching defect model (crack) provided by an embodiment of the present invention;

[0075] Figure 8 This is a comparison diagram of the preferred matching defect model (pore) provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0077] 1. Explanatory Examples In order to enable those skilled in the art to fully understand how to implement the present invention, this section provides an illustrative example that expands upon the technical solutions of the claims.

[0078] like Figure 1-Figure 2 As shown, the method for determining laser cladding process parameters based on defect characteristics and similarity metrics provided by an embodiment of the present invention includes:

[0079] S101, collecting a two-dimensional image of a defect of a component to be processed, and preprocessing the collected two-dimensional image of the defect of the component to be processed to obtain a characteristic value of the image of the defect of the component to be processed;

[0080] S102, inputting the obtained feature values of the defect image of the part to be processed into a pre-built database; the database performs preliminary matching between the defect model of the part to be processed and the models in the defect feature database through knowledge reasoning to obtain multiple preliminary matching models;

[0081] S103 , determining the laser cladding parameters in the process parameter library corresponding to the defect model with the highest similarity among the multiple preliminary matching models as the laser cladding parameters of the defect of the component to be processed.

[0082] The laser cladding process parameter determination system based on defect characteristics and similarity measurement provided by an embodiment of the present invention includes:

[0083] Detection system and database;

[0084] The detection system includes: an image acquisition module and an image processing module;

[0085] The image acquisition module includes: a detection platform, an ultrasonic probe, a control machine, and an acquisition module; and is used to acquire two-dimensional images of defects in parts to be processed.

[0086] The image processing module includes: pixel processing, image filtering and noise reduction unit and visual feature extraction unit; it is used to process the collected images of the parts to be processed to obtain the defect contour;

[0087] The database is used to construct a defect model library, a defect feature library and a process parameter library to store defect model information, defect feature information and process parameter information respectively.

[0088] The image acquisition module provided by the embodiment of the present invention includes:

[0089] Testing platform, used to place parts to be processed and ultrasonic probes;

[0090] The ultrasonic probe is used to contact the surface defects of the parts to be processed and feedback a pulse signal to the control machine each time it contacts a different position of the surface defect of the parts;

[0091] A control unit is used to control the motion trajectory of the ultrasonic probe on the detection platform; and to control the speed and position parameters of the ultrasonic probe on the detection platform;

[0092] The acquisition module is used to receive feedback pulses, collect two-dimensional images of component defects through the received feedback pulses, and transmit the collected images to the image processing module.

[0093] The method for determining laser cladding process parameters based on defect characteristics and similarity metrics provided by an embodiment of the present invention specifically includes:

[0094] S1. Place the part to be matched on the inspection platform, start the control machine and set the ultrasonic probe's movement speed, position and other parameters to control the ultrasonic probe's contact with the part's surface defects, ensure that the acquisition module can capture the part defect image, and that the area where the part to be matched is placed is at an appropriate position in the center of the image range shown, and obtain the two-dimensional image pixel information.

[0095] S2. Based on the collected and stored pixel values of the two-dimensional image, the size and resolution of the reconstructed three-dimensional image are measured. The coordinates of the two-dimensional pixel points are transformed into three-dimensional coordinates through the mapping matrix. Each frame of the two-dimensional image is mapped to the corresponding position of the three-dimensional image. The reconstructed three-dimensional coordinate points are combined together to form a three-dimensional image of the part to be repaired. The defects of the part to be repaired are marked, and the three-dimensional model of the defective part is saved and output in OBJ format.

[0096] S3. Use the three-dimensional space spherical coordinate transformation method (rotation matrix) to construct a multi-view two-dimensional image mapping model of the three-dimensional model of the defective part to be repaired.

[0097] S4. Use image feature extraction technologies such as Prewitt edge extraction and MEM binarization to generate multi-perspective two-dimensional images of component defects. Then, use a deep convolutional neural network (CNN) to extract and classify the three-dimensional morphological images of the defects in the components to be repaired.

[0098] S5. For example, using two models for detecting component defects, cracks and pores, the three image feature values for the detected cracks are 23 mm, 5.5 mm, and 0.5 mm, respectively. The three image feature values for the detected pores are 1.0 mm, 1.1 mm, and 2.0 mm, respectively, for their circularity (B1), depth (B2), and diameter (B3). These image features are entered into the database.

[0099] S6. The database performs preliminary matching on the defect model of the component to be matched and several models selected from the defect feature library as preliminary matching models.

[0100] S61 adopts a three-dimensional point cloud recognition and matching algorithm to obtain the point cloud feature representations of the defect model of the component to be matched and the model in the defect feature library. The basic invention is to use a programming library called point cloud library to read the OBJ file of the three-dimensional model of the component defect, extract the key points of the model through the internal morphological description (Intrinsic Shape Signatures, abbreviated as ISS) extraction method, and then detect its key points to construct the local reference frame (Local Reference Frame, abbreviated as LRF) to obtain three LRF coordinate axis planes. Each adjacent point is then projected along the three LRF coordinate axis planes to generate a projected point cloud grid statistical map. The density and average curvature are calculated based on the points that fall into each projection statistical map, and the normal of the midpoint of each face element is calculated. At the same time, these values are sorted into a 1×N dimensional array in a certain order. The array can be regarded as a histogram, and a histogram descriptor is generated. Finally, the normal histogram, curvature histogram, and average density histogram are constructed through projection statistics, and then spliced into a feature histogram. The feature histograms of all key points of the component 3D model are combined together to serve as the point cloud feature representation of the component 3D model. Specifically, it includes:

[0101] The S611 point cloud library reads any OBJ format file of a 3D model in the defect feature library. Assume that any 3D model in the defect feature library consists of i points (p i ) is collected from the point cloud dataset (also called point cloud P), and then the key points p in the point cloud P of the model are extracted by the ISS extraction method. j , the basic operation of the ISS extraction method is as follows:

[0102] S6111 takes any point in the point cloud P to establish an LRF, sets a spherical search range with a radius r (r is preferably 5 to 10) centered on the point as the neighborhood of the point, and all points in the neighborhood are neighboring points.

[0103] S6112 calculates the weight of each neighboring point in the neighborhood:

[0104]

[0105] S6113 calculates the covariance matrix of each neighboring point in the neighborhood:

[0106]

[0107] S6114 calculates the eigenvalues {λ1, λ2, λ3} of the covariance matrix of each neighboring point, where λ1≥λ2≥λ3.

[0108] S6115 finally sets the thresholds ε1 and ε2 (the present invention sets ε1 = ε2 = 0.975), and if the following equations are satisfied at the same time, the point is considered to be a key point p j :

[0109]

[0110] S612 searches for the neighborhood of any key point in the point cloud P, constructs an LRF based on the key point, and translates the coordinates of the neighboring points of the key point to the LRF.

[0111] S613 projects the neighboring points along the three LRF coordinate axes respectively, and obtains a three-frame projection point cloud grid statistical map.

[0112] S614 divides the projected point cloud grid statistics map into N P ×N P A grid is formed, and points and their coordinates are obtained in each grid, thereby obtaining a discrete projection statistical graph.

[0113] S615 Calculate n in projection statistics i and n j Angle between <n i , n j >, where n j Defined as the normal of the neighboring point, n i is the normal of the key point. Then <n i , n j The value range of > can be divided into N using [0, π] q (The present invention sets N q = 100) subintervals, the points distributed in each subinterval of the grid can be counted. Each subinterval of the grid can be regarded as a surface element, so there are N p (The present invention sets N p =100) bins. Each bin can be considered as having N p There are boxes, each box has a measurement value. The grid is expanded into 1×N in a certain order. q ×N P ×N P dimensional array, normalized to generate the normal histogram H n .

[0114] S616 calculates the average curvature of each grid in the projection statistics map and assigns the value of the average curvature to each grid. The grid without a point is assigned a value of 1. Each grid has a measurement value, and the grid map is expanded to 1×N in a certain order. p ×N p dimensional array, normalized to generate the curvature histogram H a .

[0115] S617 calculates the average density of points in each grid in the projection statistical map, assigns the average density value of the points to each grid, and assigns a value of 1 to grids without points. Each grid has a measurement value, and the grid map is expanded to 1×N in a certain order. p ×N p dimensional array, normalized to generate the average density histogram H b .

[0116] S618 comes from three frames of H n 、H a and H b The arrays can be stitched together and represented as follows:

[0117] H=[c1H n ,c2H a ,c3H b ]

[0118] Where H is the final representation histogram, c1, c2 and c3 are weights used to adjust the ratio of normal, curvature and average density in the feature description; the normal weight c1 is 0.5-0.7, more preferably 0.6, and the curvature weight c2 is equal to the average density weight c3, that is, c1=c3=0.2.

[0119] S619 repeats the method from S611 to S618 n times (n is preferably 5×10 5 ~1×10 6 ), generate a histogram of the representation for each key point of the model, and collect the obtained n histograms of the representation together to form a new feature set Hl = [H1, H2, H3…H n ], the feature set is the point cloud feature representation of the model.

[0120] S62 calculates the Euclidean distance between each model in the defect feature library and the defect model of the component to be matched using the point cloud feature identifier, and then calculates the similarity using the Euclidean distance. The models in the defect feature library are sorted from large to small according to similarity, and several models with the largest similarity are selected as preliminary matching models. This includes:

[0121] S621 records the feature identifier of any model A in the defect feature library as H1 A =[Hl A1 ,Hl A2 ,Hl A3 …Hl An ], the feature representation of the defect model B to be matched is Hl B =[Hl B1 ,Hl B2 ,Hl B3 …Hl Bn], the Euclidean distance is used to measure the similarity Sim between the defect model of the component to be matched and the model point cloud feature representation in the defect feature library:

[0122]

[0123] S622 repeats S621 to obtain the similarity between each model in the defect feature library and the defect model of the component to be matched;

[0124] S623 sorts the models in the defect feature library by similarity from largest to smallest, and selects the top M models with the largest similarity as preliminary matching models, where M is set to 3 to 10, and more preferably 5.

[0125] S7. The laser cladding process in the process parameter library corresponding to the defect with the highest similarity is recommended as the laser cladding process for the defect of the component to be repaired.

[0126] 2. Application Examples: In order to demonstrate the creativity and technical value of the technical solution of the present invention, this section provides application examples of the claimed technical solution on specific products or related technologies.

[0127] The method for determining laser cladding process parameters based on defect characteristics and similarity metrics provided in an embodiment of the present invention is applied to a computer device, wherein the computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for determining laser cladding process parameters based on defect characteristics and similarity metrics.

[0128] The method for determining laser cladding process parameters based on defect characteristics and similarity metrics provided by an embodiment of the present invention is applied to a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the method for determining laser cladding process parameters based on defect characteristics and similarity metrics.

[0129] The method for determining laser cladding process parameters based on defect characteristics and similarity measurement provided by an embodiment of the present invention is applied to an information data processing terminal, and the information data processing terminal is used to implement the laser cladding process parameter determination system based on defect characteristics and similarity measurement.

[0130] 3. Evidence of the effects of the embodiments: The embodiments of the present invention have achieved some positive effects during the development or use process, and indeed have great advantages over the existing technology. The following content describes them with reference to the data, charts, etc. of the experimental process.

[0131] The highest similarity of the crack defect model detected by the present invention is 0.988247, and the corresponding laser cladding process for the defects of the parts to be repaired is a laser power of 2.0-2.5kW, a spot diameter of 2.5-3.2mm, a scanning speed of 6-12mm / s, etc., and further preferably a laser power of 2.3kW, a spot diameter of 3.0mm, a scanning speed of 10mm / s, etc.; the highest similarity of the porosity defect model detected is 0.999899, and the corresponding laser cladding process for the defects of the parts to be repaired is a laser power of 700-1000W, a spot diameter of 2.0-2.2mm, a scanning speed of 8-14mm / s, etc., and further preferably a laser power of 800W, a spot diameter of 2.0mm, a scanning speed of 10mm / s, etc.

[0132] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware portion can be implemented using dedicated logic; the software portion can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will appreciate that the above-mentioned devices and methods can be implemented using computer-executable instructions and / or contained in processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0133] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with this technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for determining laser cladding process parameters based on defect characteristics and similarity metrics, characterized in that: The method for determining laser cladding process parameters based on defect characteristics and similarity measurement includes: Step 1: collecting a two-dimensional image of the defect of the part to be processed, and preprocessing the collected two-dimensional image of the defect of the part to be processed to obtain a feature value of the image of the defect of the part to be processed; In the step 1, collecting a two-dimensional image of the defect of the component to be processed, and preprocessing the collected two-dimensional image of the defect of the component to be processed includes: (1) The part to be processed is placed on the inspection platform, and the ultrasonic probe placed nearby is used to contact the surface defects of the part. A pulse signal is fed back each time a different position of the surface defect of the part is touched; a two-dimensional image of the defect of the part to be processed is collected through the feedback pulse signal; (2) storing the collected two-dimensional images of defects of the parts to be processed, and converting all the two-dimensional images of defects of the parts to be processed into three-dimensional images by using coordinate transformation, and projecting the three-dimensional images of defects of the parts to be processed into two-dimensional space by using the three-dimensional space spherical coordinate transformation method through Prewitt edge extraction, MEM binarization or other image feature extraction methods to generate multi-view two-dimensional images; (3) extracting and classifying image features of the multi-view two-dimensional images using a deep convolutional neural network; Step 2: Input the obtained feature values of the defect image of the part to be processed into a pre-built database; the database performs preliminary matching between the defect model of the part to be processed and the models in the defect feature library through knowledge reasoning to obtain multiple preliminary matching models; In the second step, the database performs preliminary matching between the defect model of the component to be processed and the model in the defect feature library through knowledge reasoning, including: 1) Using a 3D point cloud recognition and matching algorithm, the point cloud feature representations of the defect model of the part to be processed and the model in the defect feature library are obtained respectively; 2) calculating the Euclidean distance between each model in the defect feature library and the model of the component to be processed using the point cloud feature representation, and obtaining the similarity using the Euclidean distance calculation; 3) Sort the models in the defect feature library by similarity from large to small, and select the M models with the largest similarity as the preliminary matching models, as follows: Among them, Hl A =[Hl A1 ,Hl A2 ,Hl A3 …Hl An ] represents the feature identifier of any model A in the defect feature library; H1 B =[Hl B1 ,Hl B2 ,Hl B3 …Hl Bn ] represents the feature identifier of the defect model B to be matched; Sim represents the similarity between the feature identifiers of the part model to be processed and the model point cloud in the defect feature library; the value range of M is 3 to 10; 4) Repeat step 3) to obtain the similarity between each model in the defect feature library and the model of the part to be processed; Step three: determining the laser cladding parameters in the process parameter library corresponding to the defect model with the highest similarity among the multiple preliminary matching models as the laser cladding parameters of the defect of the component to be processed.

2. The method for determining laser cladding process parameters based on defect characteristics and similarity metrics according to claim 1, wherein: In the step 2, the database construction method includes: First, the component defect models, including cracks, pores, wear, and scratches, are determined, and a defect model library is constructed. The image feature values of defect length, circularity, and depth are obtained for the crack defect model, and the image feature values of circularity, depth, and pore diameter are obtained for the pore defect model. Secondly, obtain the defect equivalent area circle radius, circularity, texture entropy, and second-order moment image eigenvalues of the wear defect model; obtain the defect area and depth image eigenvalues of the scratch defect model; Then, a defect feature library is constructed based on the acquired image feature values; and laser cladding repair parameters corresponding to different image feature values in the defect feature library are determined, and a process parameter library is constructed based on the laser cladding repair parameters; Finally, a database is constructed based on the defect model library, defect feature library and process parameter library.

3. The method for determining laser cladding process parameters based on defect characteristics and similarity metrics according to claim 1, wherein: In the step 1), the three-dimensional point cloud recognition and matching algorithm is used to respectively obtain the point cloud feature representation of the defect model of the part to be processed and the model in the defect feature library, including: 1.1) The point cloud library reads any OBJ file of a 3D model in the defect feature library, extracts the point cloud dataset of the 3D model, and uses the ISS extraction method to extract the key points in the point cloud dataset: 1.1.1) Establish an LRF for any point in the point cloud dataset and determine a spherical search range with radius r centered at the point as the neighborhood of the point. All points within the neighborhood are considered to be neighboring points. 1.1.2) Use the following formula to calculate the weight of each neighboring point in the neighborhood: 1.1.3) Calculate the covariance matrix of each neighboring point in the neighborhood using the following formula: 1.1.4) Calculate the eigenvalues {λ1,λ2,λ3} of the covariance matrix of each neighboring point, where λ1≥λ2≥λ3; 1.1.5) Set thresholds ε1 and ε2, and if the following equation is satisfied, the point is considered a key point p j : 1.2) Search the neighborhood of any key point in the point cloud dataset, construct a LRF based on the key point, and translate the coordinates of the neighboring points of the key point to the LRF; 1.3) Project the adjacent points along the three LRF coordinate axes to obtain three frames of projection point cloud grid statistics; divide the projection point cloud grid statistics into N P ×N P grids, obtaining points and the coordinates of the points in each grid, and obtaining a discrete projection statistical graph; 1.4) Calculate n in the projection statistics graph i and n j Angle between <n i , n j >, where n j Represents the normal of the neighboring point, n i Represents the normal of the key point; <n i , n j The value range of > is divided into N using [0, π] q subintervals, count the points distributed in each subinterval of the grid; regard each subinterval of the grid as a surface element, and there are N p Each bin is considered to have N p boxes, each box contains a measurement value; expand the grid map into 1×N in a certain order q ×N P ×N P dimensional array, normalized to generate the normal histogram H n ; 1.5) Calculate the average curvature of each grid in the projection statistics map, assign the value of the average curvature to each grid; assign the value of 1 to the grid without points; each grid has a measurement value, and the grid map is expanded to 1×N in a certain order p ×N p dimensional array, normalized to generate the curvature histogram H a ; 1.6) Calculate the average density of each grid point in the projection statistical map, assign the average density value of the points to each grid, and assign a value of 1 to grids without points; each grid has a measurement value, and the grid map is expanded to 1×N in a certain order. p ×N p dimensional array, normalized to generate the average density histogram H b ; 1.7) H from three frames n 、H a and H b The arrays are stitched together as follows: H=[c1H n ,c2H a ,c3H b ]; Where H represents the final histogram of the representation, c1, c2, and c3 represent the weights used to adjust the proportions of normal, curvature, and average density in the feature description, respectively; the curvature weight c2 is equal to the average density weight c3; 1.8) Repeat steps 1.1) to 1.7) n times to generate a histogram of the representation for each key point of the 3D model, and collect the obtained n histograms of the representation to form a new feature set Hl = [H1, H2, H3 ... H n ], the feature set is a point cloud feature representation of the three-dimensional model.

4. A laser cladding process parameter determination system based on defect characteristics and similarity measurement, which implements the laser cladding process parameter determination method based on defect characteristics and similarity measurement according to any one of claims 1 to 3, characterized in that: The laser cladding process parameter determination system based on defect characteristics and similarity measurement includes: Detection system and database; The detection system includes: an image acquisition module and an image processing module; The image acquisition module includes: a detection platform, an ultrasonic probe, a control machine, and an acquisition module; it is used to acquire two-dimensional images of defects in the parts to be processed; The image processing module includes: a pixel processing unit, an image filtering and noise reduction unit, and a visual feature extraction unit; it is used to process the collected image of the part to be processed to obtain the defect contour; The database is used to construct a defect model library, a defect feature library and a process parameter library to store defect model information, defect feature information and process parameter information respectively.

5. The laser cladding process parameter determination system based on defect characteristics and similarity measurement according to claim 4, characterized in that: The image acquisition module includes: Testing platform, used to place parts to be processed and ultrasonic probes; The ultrasonic probe is used to contact the surface defects of the parts to be processed and feedback a pulse signal to the control machine each time it contacts a different position of the surface defect of the parts; A control unit is used to control the motion trajectory of the ultrasonic probe on the detection platform; and to control the speed and position parameters of the ultrasonic probe on the detection platform; The acquisition module is used to receive feedback pulses, collect two-dimensional images of component defects through the received feedback pulses, and transmit the collected images to the image processing module.

6. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method for determining laser cladding process parameters based on defect characteristics and similarity metrics as described in any one of claims 1 to 3.

7. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method for determining laser cladding process parameters based on defect characteristics and similarity metrics as described in any one of claims 1 to 3.

8. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the laser cladding process parameter determination system based on defect characteristics and similarity measurement as described in any one of claims 4-5.

Images