A method for optimizing the production process of off-road motorcycle parts based on intelligent regulation

By using edge detection and Hough transformation to reconstruct the three-dimensional model diagram in laser engraving equipment, combined with Bayesian network model for fault prediction, the problems of inaccurate equipment regulation and high waste rate are solved, and efficient and accurate automated production and Quality control.

CN118736329BActive Publication Date: 2024-12-06YONGKANG ZHIZHAN TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411200275.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-12-06
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

The existing laser engraving equipment lacks an intelligent detection mechanism during processing, resulting in high waste rate and inaccurate equipment regulation, which increases production costs and maintenance difficulties.

Method used

By obtaining the characteristic images of laser engraving equipment for edge detection and Hough transformation, the three-dimensional model diagram of the part is reconstructed, the part is qualified, and the equipment is regulated based on the fault analysis results. The Bayesian network model is used to predict the fault probability to achieve automated monitoring and failure prediction.

Benefits of technology

It improves the production efficiency and product quality of laser engraving equipment, reduces maintenance costs, provides efficient and accurate automation solutions, and reduces the defective yield rate and production decision-making problems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118736329B_ABST
    Figure CN118736329B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of motorcycle parts production, and in particular to a method for optimizing the production process of off-road motorcycle parts based on intelligent adjustment. Edge detection processing is performed on feature images to obtain shape features of processed parts; Hough transform processing is performed on the shape features of the processed parts to obtain a number of boundary points, and the boundary points are reduced in redundancy, and the actual shape three-dimensional model diagram of the processed parts is reconstructed according to the reduced boundary points; the processed parts are evaluated and analyzed according to the actual shape three-dimensional model diagram and the standard shape three-dimensional model diagram; if the processed parts are unqualified parts, a fault analysis is performed on the laser engraving equipment, and the laser engraving equipment is adjusted and processed according to the fault analysis results, which helps to improve production efficiency, ensure product quality, reduce maintenance costs, and provide data support for production decisions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of motorcycle accessories production, in particular to an intelligent adjustment-based production process optimization method for off-road motorcycle parts. Background Art

[0002] With the advancement of technology, the fuel tank design of off-road motorcycles is no longer limited to functionality, but is developing in the direction of personalization and aesthetics. Traditional fuel tank manufacturing processes, such as stamping or electroplating, are difficult to meet the needs of complex shapes and fine engraving. Laser engraving technology has gradually been applied to fuel tank manufacturing due to its high precision and flexibility. However, in the process of laser engraving fuel tanks by laser engraving equipment, the intelligence level of existing laser engraving equipment still needs to be improved. For example, the existing laser engraving equipment lacks the corresponding detection mechanism, and additional detection stations need to be added, which increases the investment cost of the production workshop; in addition, laser engraving equipment often produces large quantities of waste due to malfunctions during the engraving process, resulting in a high scrap rate. Therefore, how to quickly identify the engraving conditions and accurately control the equipment according to the engraving conditions is also a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a method for optimizing the production process of off-road motorcycle parts based on intelligent regulation.

[0004] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is:

[0005] The first aspect of the present invention discloses a method for optimizing the production process of off-road motorcycle parts based on intelligent adjustment, comprising the following steps:

[0006] Acquire a feature image of a part being processed in a laser engraving device at a preset time point, perform edge detection processing on the feature image, and obtain shape features of the part being processed;

[0007] Performing Hough transform processing on the shape features of the processed part to obtain a number of boundary points, performing redundancy reduction processing on the boundary points, and reconstructing a three-dimensional model diagram of the actual shape of the processed part according to the reduced-redundancy boundary points;

[0008] Obtaining a finished laser engraving engineering drawing of a processed part, and constructing a standard shape three-dimensional model drawing of the processed part according to the finished laser engraving engineering drawing; and evaluating and analyzing the processed part according to the actual shape three-dimensional model drawing and the standard shape three-dimensional model drawing;

[0009] If the processed part is qualified, the laser engraving device continues to process and produce the next part; if the processed part is unqualified, a fault analysis is performed on the laser engraving device, and the laser engraving device is adjusted and processed according to the fault analysis results.

[0010] More specifically, edge detection processing is performed on the feature image to obtain the shape features of the processed parts, specifically:

[0011] The cv2.filter2D function in the OpenCV library is used to calculate the gradients of the feature image in the horizontal and vertical directions; the horizontal gradient matrix and the vertical gradient matrix are constructed according to the gradients of the feature image in the horizontal and vertical directions;

[0012] The gradient matrix in the horizontal direction is added to the gradient matrix in the vertical direction to obtain an edge strength map; wherein the edge strength map reflects the significance of the edge of each pixel point, and the edge strength of the pixel point is related to the gradient value of the pixel;

[0013] Traversing each pixel point in the edge intensity map, and comparing the pixel value of each pixel point with the pixel values ​​of other pixels points in the preset area;

[0014] If the pixel value of a certain pixel is greater than the pixel values ​​of other pixels in the preset area, the pixel is marked as a local maximum pixel, that is, a strong edge point; if the pixel value of a certain pixel is not greater than the pixel value of one or more pixels in the preset area, the pixel is marked as a non-local maximum pixel, that is, a weak edge point;

[0015] If a certain pixel point in the edge intensity map is a strong edge point, the position node corresponding to the pixel point in the edge intensity map is set to black; if a certain pixel point in the edge intensity map is a weak edge point, the position node corresponding to the pixel point in the edge intensity map is set to white; and so on, until all the pixel points are set, a binary image is obtained;

[0016] The findContours function in the OpenCV library is used to extract the edge contours in the image from the binary image to obtain the shape features of the processed parts.

[0017] More specifically, the shape features of the processed part are processed by Hough transform to obtain a number of boundary points, and the boundary points are reduced in redundancy, and the actual shape three-dimensional model diagram of the processed part is reconstructed according to the reduced boundary points, specifically:

[0018] Performing polar coordinate Hough transformation on the shape features of the processed part to convert the shape features of the processed part into a plurality of straight line segments;

[0019] Each straight line segment is converted into Hough space, and the peak value of each straight line segment is found in Hough space to obtain several peak values; each peak value is reversely mapped back to the feature image to obtain several boundary points of the feature image;

[0020] Calculate the Mahalanobis distance between each boundary point, and compare the Mahalanobis distance between each boundary point with a preset threshold;

[0021] If the Mahalanobis distance between two boundary points is less than a preset threshold, the two boundary points are defined as redundant boundary points, and the two redundant boundary points are merged; and so on, until all redundant boundary points are processed and corrected, and the reduced-redundancy boundary points are obtained;

[0022] Randomly select a reduced-redundancy boundary point as a reference point, construct a three-dimensional coordinate system based on the reference point, obtain the three-dimensional coordinate values ​​of each reduced-redundancy boundary point in the three-dimensional coordinate system, and generate a boundary point coordinate collection according to the three-dimensional coordinate values ​​of each reduced-redundancy boundary point in the three-dimensional coordinate system;

[0023] A three-dimensional model diagram of the actual shape of the processed part is constructed according to the boundary point coordinate collection.

[0024] More specifically, the processed parts are evaluated and analyzed according to the actual shape three-dimensional model diagram and the standard shape three-dimensional model diagram, specifically:

[0025] Constructing a registration space, and inputting the actual shape three-dimensional model image and the standard shape three-dimensional model image into the registration space;

[0026] Retrieving reference datums of the actual shape three-dimensional model image and the standard shape three-dimensional model image in the registration space, and performing alignment processing on the reference datums of the two three-dimensional model images, so as to complete the registration process of the actual shape three-dimensional model image and the standard shape three-dimensional model image in the registration space;

[0027] Retrieving in the registration space a model area where the actual shape three-dimensional model image and the standard shape three-dimensional model image overlap with each other, and setting the overlapping model area as a white area; and retrieving in the registration space a model area where the actual shape three-dimensional model image and the standard shape three-dimensional model image do not overlap with each other, and setting the non-overlapping model area as a black area;

[0028] Calculating the area volume of the white area and the area volume of the black area; dividing the area volume of the white area by the area volume of the black area to obtain the overlap between the actual shape three-dimensional model image and the standard shape three-dimensional model image;

[0029] If the degree of overlap between the actual shape 3D model image and the standard shape 3D model image is not greater than the preset overlap threshold, the processed part is marked as an unqualified part; if the degree of overlap between the actual shape 3D model image and the standard shape 3D model image is greater than the preset overlap threshold, the processed part is marked as a qualified part.

[0030] More specifically, if the processed parts are unqualified, a fault analysis is performed on the laser engraving equipment, and the laser engraving equipment is regulated and processed according to the fault analysis results, specifically:

[0031] If the processed part is marked as a defective part, the position area corresponding to the black area model between the actual shape 3D model image and the standard shape 3D model image is obtained in the registration space, which is defined as the carving abnormal area of ​​the processed part;

[0032] Obtaining a device operation log of the laser engraving device, and obtaining, according to the device operation log, a historical processing time period corresponding to the laser engraving device performing engraving processing on the engraving abnormal area, and defining the historical processing time period as the abnormal processing time period;

[0033] and obtaining, according to the equipment operation log, historical operation parameters of each sub-equipment in the laser engraving equipment when processing is performed during the abnormal processing time period;

[0034] Constructing a Bayesian network model, and pre-training the Bayesian network model to obtain a trained Bayesian network model;

[0035] Importing the historical operating parameters of each sub-device into the trained Bayesian network model for fault prediction to obtain the fault probability of each sub-device;

[0036] If the failure probability of each sub-device in the laser engraving device is not greater than the preset failure probability, the laser engraving device continues to process and produce the next part, and the Bayesian network model is used to continuously monitor the failure status of each sub-device in the laser engraving device;

[0037] If the failure probability of one or more sub-devices in the laser engraving device is greater than a preset failure probability, the laser engraving device is controlled to stop production, and the sub-devices with a failure probability greater than the preset failure probability are marked, and fault information is generated according to the marked sub-devices, and the fault information is sent to a preset platform for display.

[0038] More specifically, a Bayesian network model is constructed, and the Bayesian network model is trained in advance to obtain a trained Bayesian network model, specifically:

[0039] Obtain the historical fault events of each sub-device through the device operation log, and calculate the similarity between the historical fault events through the cosine similarity algorithm;

[0040] Merge historical fault events with similarity greater than a preset similarity to cluster historical fault events with the same fault type to obtain a set of historical fault events of various fault types that have occurred on the sub-device;

[0041] Performing feature engineering processing on a historical fault event set of various fault types that have occurred in the sub-device to obtain fault feature data of various fault types that have occurred in the sub-device; and dividing the fault feature data into a fault feature data training set and a fault feature data test set;

[0042] Constructing a Bayesian network, initializing nodes of the Bayesian network according to various fault types that have occurred in the sub-device, and initializing edges of the Bayesian network according to a training set of fault feature data of various fault types;

[0043] The maximum likelihood estimation algorithm is introduced, and the conditional probability between each node and the edge is estimated based on the maximum likelihood estimation algorithm. A conditional probability table is constructed according to the conditional probability between each node and the edge to obtain a Bayesian network model.

[0044] Use the fault feature data test set to evaluate the accuracy and recall of the Bayesian network model. If the accuracy and recall of the Bayesian network model meet the preset requirements, it means that the model training is complete, and the trained Bayesian network model is output;

[0045] If the accuracy and recall of the Bayesian network model do not meet the preset requirements, the conditional probability table is adjusted until the accuracy and recall of the Bayesian network model meet the preset requirements, and the trained Bayesian network model is output.

[0046] The second aspect of the present invention discloses an off-road motorcycle parts production process optimization system based on intelligent adjustment, the off-road motorcycle parts production process optimization system comprises a memory and a processor, the memory stores an off-road motorcycle parts production process optimization method program, when the off-road motorcycle parts production process optimization method program is executed by the processor, the following steps are implemented:

[0047] Acquire a feature image of a part being processed in a laser engraving device at a preset time point, perform edge detection processing on the feature image, and obtain shape features of the part being processed;

[0048] Performing Hough transform processing on the shape features of the processed part to obtain a number of boundary points, performing redundancy reduction processing on the boundary points, and reconstructing a three-dimensional model diagram of the actual shape of the processed part according to the reduced-redundancy boundary points;

[0049] Obtaining a finished laser engraving engineering drawing of a processed part, and constructing a standard shape three-dimensional model drawing of the processed part according to the finished laser engraving engineering drawing; and evaluating and analyzing the processed part according to the actual shape three-dimensional model drawing and the standard shape three-dimensional model drawing;

[0050] If the processed part is qualified, the laser engraving device continues to process and produce the next part; if the processed part is unqualified, a fault analysis is performed on the laser engraving device, and the laser engraving device is adjusted and processed according to the fault analysis results.

[0051] The present invention solves the technical defects existing in the background technology, and has the following beneficial effects: obtaining the characteristic image of the processed part in the laser engraving equipment at a preset time node, performing edge detection processing on the characteristic image, and obtaining the shape feature of the processed part; performing Hough transform processing on the shape feature of the processed part to obtain a number of boundary points, and performing redundancy reduction processing on the boundary points, and reconstructing the actual shape three-dimensional model diagram of the processed part according to the reduced boundary points; obtaining the laser engraving finished engineering drawing of the processed part, and constructing the standard shape three-dimensional model diagram of the processed part according to the laser engraving finished engineering drawing; evaluating and analyzing the processed part according to the actual shape three-dimensional model diagram and the standard shape three-dimensional model diagram; if the processed part is a qualified part, the laser engraving equipment continues to process and produce the next part; if the processed part is an unqualified part, the laser engraving equipment is analyzed for faults, and the laser engraving equipment is regulated and processed according to the fault analysis results. The present invention can provide an efficient, accurate and automated solution for the production process of laser engraving equipment, which helps to improve production efficiency, ensure product quality, reduce maintenance costs, and provide data support for production decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0053] Figure 1 A flow chart of the overall method for optimizing the production process of off-road motorcycle parts;

[0054] Figure 2 A partial method flow chart of a method for optimizing the production process of off-road motorcycle parts;

[0055] Figure 3 System block diagram of a system for optimizing the production process of off-road motorcycle parts. DETAILED DESCRIPTION

[0056] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0057] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0058] like Figure 1 As shown, the first aspect of the present invention discloses a method for optimizing the production process of off-road motorcycle parts based on intelligent adjustment, comprising the following steps:

[0059] S102: acquiring a feature image of a part being processed in a laser engraving device at a preset time point, performing edge detection processing on the feature image, and obtaining shape features of the part being processed;

[0060] S104: performing Hough transform processing on the shape features of the processed part to obtain a number of boundary points, performing redundancy reduction processing on the boundary points, and reconstructing a three-dimensional model diagram of the actual shape of the processed part according to the reduced-redundancy boundary points;

[0061] S106: Obtaining a finished laser engraving engineering drawing of the processed part, constructing a standard shape three-dimensional model drawing of the processed part according to the finished laser engraving engineering drawing; and evaluating and analyzing the processed part according to the actual shape three-dimensional model drawing and the standard shape three-dimensional model drawing;

[0062] S108: If the processed part is a qualified part, the laser engraving device continues to process and produce the next part; if the processed part is an unqualified part, a fault analysis is performed on the laser engraving device, and the laser engraving device is adjusted and processed according to the fault analysis result.

[0063] More specifically, edge detection processing is performed on the feature image to obtain the shape features of the processed parts, specifically:

[0064] The cv2.filter2D function in the OpenCV library is used to calculate the gradients of the feature image in the horizontal and vertical directions; the horizontal gradient matrix and the vertical gradient matrix are constructed according to the gradients of the feature image in the horizontal and vertical directions;

[0065] Among them, the cv2.filter2D function can apply various custom convolution kernels to filter images. It achieves filtering effects by performing convolution operations on images. In the OpenCV library, the cv2.filter2D function is a general image filtering function, and its operation process and principle will not be described in detail here.

[0066] The gradient matrix in the horizontal direction is added to the gradient matrix in the vertical direction to obtain an edge strength map; wherein the edge strength map reflects the significance of the edge of each pixel point, and the edge strength of the pixel point is related to the gradient value of the pixel;

[0067] Traversing each pixel point in the edge intensity map, and comparing the pixel value of each pixel point with the pixel values ​​of other pixels points in the preset area;

[0068] If the pixel value of a certain pixel is greater than the pixel values ​​of other pixels in the preset area, the pixel is marked as a local maximum pixel, that is, a strong edge point; if the pixel value of a certain pixel is not greater than the pixel value of one or more pixels in the preset area, the pixel is marked as a non-local maximum pixel, that is, a weak edge point;

[0069] If a certain pixel point in the edge intensity map is a strong edge point, the position node corresponding to the pixel point in the edge intensity map is set to black; if a certain pixel point in the edge intensity map is a weak edge point, the position node corresponding to the pixel point in the edge intensity map is set to white; and so on, until all the pixel points are set, a binary image is obtained;

[0070] The findContours function in the OpenCV library is used to extract the edge contours in the image from the binary image to obtain the shape features of the processed parts.

[0071] Among them, the findContours function can detect the contours of white or black objects in an image. White objects refer to areas with higher brightness than the surrounding areas in a binary image, while black objects refer to areas with lower brightness than the surrounding areas. In the OpenCV library, the findContours function is a general function for detecting contours in an image. Its operation process and principle will not be described in detail here.

[0072] It should be noted that the image processing technology based on the OpenCV library is used to identify and extract the edge features of the surface of the off-road motorcycle tank. By calculating the gradient of the feature image, the gradient matrix in the horizontal and vertical directions is constructed, and then merged to form an edge intensity map, which reflects the prominence of the edge of the pixel point. Then, by comparing the pixel values, the strong edge points and the weak edge points are distinguished and represented in black and white on the binary image. Finally, the findContours function is used to extract the edge contour from the binary image to obtain the geometric shape information of the part, that is, the curve contour feature of the part. This method can realize the feature extraction of the complex surface curve of the tank, obtain the shape characteristics of the tank, and provide accurate data support for the subsequent construction of the tank model, thereby improving production efficiency.

[0073] More specifically, the shape features of the processed part are processed by Hough transform to obtain a number of boundary points, and the boundary points are reduced in redundancy, and the actual shape three-dimensional model diagram of the processed part is reconstructed according to the reduced boundary points, specifically:

[0074] Performing polar coordinate Hough transformation on the shape features of the processed part to convert the shape features of the processed part into a plurality of straight line segments;

[0075] Each straight line segment is converted into Hough space, and the peak value of each straight line segment is found in Hough space to obtain several peak values; each peak value is reversely mapped back to the feature image to obtain several boundary points of the feature image;

[0076] Calculate the Mahalanobis distance between each boundary point, and compare the Mahalanobis distance between each boundary point with a preset threshold;

[0077] If the Mahalanobis distance between two boundary points is less than a preset threshold, the two boundary points are defined as redundant boundary points, and the two redundant boundary points are merged; and so on, until all redundant boundary points are processed and corrected, and the reduced-redundancy boundary points are obtained;

[0078] Randomly select a reduced-redundancy boundary point as a reference point, construct a three-dimensional coordinate system based on the reference point, obtain the three-dimensional coordinate values ​​of each reduced-redundancy boundary point in the three-dimensional coordinate system, and generate a boundary point coordinate collection according to the three-dimensional coordinate values ​​of each reduced-redundancy boundary point in the three-dimensional coordinate system;

[0079] A three-dimensional model diagram of the actual shape of the processed part is constructed according to the boundary point coordinate collection.

[0080] The Chinese name of Hough transform is Hough transform. The boundary point coordinate collection is imported into 3D software such as CAD and SolidWorks to build a model, thereby building a 3D model diagram of the actual shape of the processed part.

[0081] It should be noted that, first, the shape features of the processed parts (fuel tanks) are converted into several straight line segments by performing polar coordinate Hough transformation. Then, these straight line segments are transformed into Hough space, and the boundary points of the feature image are obtained by finding the peaks of the straight line segments. Next, the Mahalanobis distance between these boundary points is calculated and compared with the preset threshold to identify redundant boundary points and merge them, and finally obtain the reduced boundary points. By randomly selecting a reduced boundary point as a reference point, a three-dimensional coordinate system is constructed, and the three-dimensional coordinate values ​​of each reduced boundary point in the three-dimensional coordinate system are obtained. Finally, a collection of boundary point coordinates is generated according to these three-dimensional coordinate values, and a three-dimensional model diagram of the actual shape of the processed parts is constructed using these data. This method can accurately extract and reduce the shape features of the processed parts and construct a high-precision three-dimensional model diagram. Specifically, through polar coordinate Hough transformation and peak detection, straight line segment features can be effectively extracted from complex shapes, and redundant boundary points can be identified and reduced by calculating the Mahalanobis distance and comparing the threshold, thereby improving the accuracy of feature extraction. Finally, by constructing a three-dimensional model based on a collection of boundary point coordinates, high-precision modeling of the actual shape of the machined parts can be achieved.

[0082] It should also be noted that Hough transform is an effective feature extraction method that identifies the features of a specific shape by mapping points in image space to parameter space. However, in practical applications, especially when the shape features of machined parts are processed by Hough transform, the obtained boundary points may be redundant. This is mainly because: (1) In the actual image acquisition process, due to factors such as ambient light and equipment accuracy, noise may be introduced into the image, causing the Hough transform to detect some points in non-target areas, thereby generating redundant boundary points. (2) The shape of the machined parts may be very complex and contain multiple similar or close feature points. When extracting these feature points, the Hough transform may regard them as independent boundary points, but in fact they may be different parts of the same feature. The existence of redundant boundary points will bring a series of adverse effects on the subsequent 3D model reconstruction: (1) Redundant boundary points will cause the shape of the 3D model to deviate from the shape of the actual part, thereby reducing the accuracy and reliability of the model. (2) Redundant boundary points will increase the amount of calculation in the model reconstruction process, prolong the processing time, and reduce efficiency. (3) Redundant boundary points will interfere with subsequent feature matching, model optimization and other steps, increasing the difficulty and complexity of processing. Therefore, after Hough transform processing, the obtained boundary points need to be reduced in redundancy to improve the accuracy of feature extraction and the quality of subsequent three-dimensional model reconstruction.

[0083] More specifically, if Figure 2As shown, the processed parts are evaluated and analyzed according to the actual shape three-dimensional model diagram and the standard shape three-dimensional model diagram, specifically:

[0084] S202: constructing a registration space, and inputting the actual shape three-dimensional model image and the standard shape three-dimensional model image into the registration space;

[0085] S204: retrieving reference datums of the actual shape three-dimensional model image and the standard shape three-dimensional model image in the registration space, and performing alignment processing on the reference datums of the two three-dimensional model images, so as to complete the registration process of the actual shape three-dimensional model image and the standard shape three-dimensional model image in the registration space;

[0086] S206: searching in the registration space for a model area where the actual shape 3D model image and the standard shape 3D model image overlap with each other, and setting the overlapping model area to a white area; and searching in the registration space for a model area where the actual shape 3D model image and the standard shape 3D model image do not overlap with each other, and setting the non-overlapping model area to a black area;

[0087] S208: calculating the area volume of the white area and the area volume of the black area; dividing the area volume of the white area by the area volume of the black area to obtain the overlap between the actual shape three-dimensional model image and the standard shape three-dimensional model image;

[0088] S210: If the degree of overlap between the actual shape 3D model image and the standard shape 3D model image is not greater than a preset overlap threshold, the processed part is marked as an unqualified part; if the degree of overlap between the actual shape 3D model image and the standard shape 3D model image is greater than a preset overlap threshold, the processed part is marked as a qualified part.

[0089] It should be noted that, first, a registration space is constructed by three-dimensional software, and the three-dimensional model image of the actual shape and the three-dimensional model image of the standard shape are simultaneously input into the registration space. In the registration space, the model registration process is completed by retrieving the reference datum of the two models (the reference datum is obtained by the relevant technical personnel) and performing alignment processing. Then, the area where the actual shape model image and the standard shape model image overlap with each other (white area) and the area where they do not overlap (black area) are identified in the registration space. By calculating the volume of these two areas and calculating their ratio, the overlap between the actual shape three-dimensional model image and the standard shape three-dimensional model image is obtained. Finally, according to the preset overlap threshold, the processed parts are marked as qualified or unqualified. The three-dimensional model images of the processed parts are registered and compared in an automated manner, which greatly improves the detection efficiency and accuracy. By calculating the overlap, the difference between the actual shape model image and the standard shape model image can be quantified, which provides an objective basis for quality assessment, is simple to operate, and is easy to implement in actual production. Overall, this method provides an efficient, accurate, and automated solution for the quality inspection of processed parts, which helps to improve production efficiency and product quality.

[0090] More specifically, if the processed parts are unqualified, a fault analysis is performed on the laser engraving equipment, and the laser engraving equipment is regulated and processed according to the fault analysis results, specifically:

[0091] If the processed part is marked as a defective part, the position area corresponding to the black area model between the actual shape 3D model image and the standard shape 3D model image is obtained in the registration space, which is defined as the carving abnormal area of ​​the processed part;

[0092] Obtaining a device operation log of the laser engraving device, and obtaining, according to the device operation log, a historical processing time period corresponding to the laser engraving device performing engraving processing on the engraving abnormal area, and defining the historical processing time period as the abnormal processing time period;

[0093] and obtaining, according to the equipment operation log, historical operation parameters of each sub-equipment in the laser engraving equipment when processing is performed during the abnormal processing time period;

[0094] Constructing a Bayesian network model, and pre-training the Bayesian network model to obtain a trained Bayesian network model;

[0095] Importing the historical operating parameters of each sub-device into the trained Bayesian network model for fault prediction to obtain the fault probability of each sub-device;

[0096] If the failure probability of each sub-device in the laser engraving device is not greater than the preset failure probability, the laser engraving device continues to process and produce the next part, and the Bayesian network model is used to continuously monitor the failure status of each sub-device in the laser engraving device;

[0097] If the failure probability of one or more sub-devices in the laser engraving device is greater than a preset failure probability, the laser engraving device is controlled to stop production, and the sub-devices with a failure probability greater than the preset failure probability are marked, and fault information is generated according to the marked sub-devices, and the fault information is sent to a preset platform for display.

[0098] The equipment operation log of laser engraving equipment is a detailed document that records the equipment's operating status, operation process and performance parameters. These logs usually include basic equipment information, operation history, processing tasks, fault records, maintenance records and other aspects.

[0099] It should be noted that when a processed part is judged to be unqualified, the system will identify the black area between the actual shape 3D model image and the standard shape 3D model image in the registration space, and define these areas as the abnormal engraving area of ​​the processed part. Then, the system will collect the operation log of the laser engraving equipment, and obtain the historical processing time period (abnormal processing time period) corresponding to the processing of the abnormal engraving area, as well as the historical operating parameters of each sub-equipment during this time period. Next, the system uses the pre-trained Bayesian network model to analyze the historical operating parameters of each sub-equipment and predict their probability of failure. If the failure probability of all sub-equipment is lower than the preset threshold, the system will continue to allow the laser engraving equipment to process and produce, and continue to use the Bayesian network model to monitor the equipment status. If it is found that the failure probability of any sub-equipment exceeds the preset threshold, the system will immediately stop production, mark the failed sub-equipment, generate fault information, and send it to the preset platform for display. Through the Bayesian network model, the system can predict the probability of failure of each sub-device in the laser engraving equipment, so as to discover potential problems in advance; during the production process, the system can monitor the equipment status in real time, promptly discover and handle abnormal situations, avoid the phenomenon of mass production of defective products due to equipment failure, reduce economic losses, and effectively reduce the defect rate. At the same time, when the equipment fails, the system can quickly locate the fault and provide fault information to facilitate maintenance personnel to diagnose and handle the fault. Overall, this system can provide an efficient, accurate and automated solution for fault prediction and monitoring of laser engraving equipment, which helps to improve production efficiency, ensure product quality, reduce maintenance costs, and provide data support for production decisions.

[0100] More specifically, a Bayesian network model is constructed, and the Bayesian network model is trained in advance to obtain a trained Bayesian network model, specifically:

[0101] Obtain the historical fault events of each sub-device through the device operation log, and calculate the similarity between the historical fault events through the cosine similarity algorithm;

[0102] Merge historical fault events with similarity greater than a preset similarity to cluster historical fault events with the same fault type to obtain a set of historical fault events of various fault types that have occurred on the sub-device;

[0103] Performing feature engineering processing on a historical fault event set of various fault types that have occurred in the sub-device to obtain fault feature data of various fault types that have occurred in the sub-device; and dividing the fault feature data into a fault feature data training set and a fault feature data test set;

[0104] Constructing a Bayesian network, initializing nodes of the Bayesian network according to various fault types that have occurred in the sub-device, and initializing edges of the Bayesian network according to a training set of fault feature data of various fault types;

[0105] The maximum likelihood estimation algorithm is introduced, and the conditional probability between each node and the edge is estimated based on the maximum likelihood estimation algorithm. A conditional probability table is constructed according to the conditional probability between each node and the edge to obtain a Bayesian network model.

[0106] Use the fault feature data test set to evaluate the accuracy and recall of the Bayesian network model. If the accuracy and recall of the Bayesian network model meet the preset requirements, it means that the model training is complete, and the trained Bayesian network model is output;

[0107] If the accuracy and recall of the Bayesian network model do not meet the preset requirements, the conditional probability table is adjusted until the accuracy and recall of the Bayesian network model meet the preset requirements, and the trained Bayesian network model is output.

[0108] It should be noted that the Bayesian network model is a graphical model based on probability theory, which uses a directed acyclic graph (i.e., nodes and edges) to represent the dependencies and conditional probabilities between variables. In the fault prediction of laser engraving equipment, the Bayesian network model can use the statistical laws and probability relationships implicit in the historical operating data to predict the possible failures of each sub-equipment in the future and provide data support for the maintenance and repair of the equipment. The Bayesian network model is a commonly used prediction model in equipment failure prediction, and its fault prediction principle will not be further elaborated here.

[0109] It should be noted that the historical fault events of each sub-device are first obtained through the device operation log, and the cosine similarity algorithm is used to calculate the similarity between these events. Events with high similarity are merged to achieve fault type clustering. Then, feature engineering is performed on the clustered historical fault events to extract fault feature data, and the data is divided into training sets and test sets. Next, a Bayesian network is constructed, the network nodes and edges are initialized, and the conditional probabilities between nodes and edges are estimated using the maximum likelihood estimation algorithm, a conditional probability table is constructed, and a Bayesian network model is obtained. Finally, the test set is used to evaluate the accuracy and recall of the model. If the preset requirements are met, the trained model is output; otherwise, the conditional probability table is adjusted until the requirements are met, and the Bayesian network training process is completed.

[0110] In addition, the method further comprises the following steps:

[0111] At a preset second time point, the processed part in the laser engraving device is scanned by an ultrasonic detector, and an ultrasonic signal fed back by the processed part is obtained;

[0112] Performing feature extraction processing on the ultrasonic signal to obtain crack feature data of the processed part; the crack feature data includes crack length, depth, width and position;

[0113] Calculating the crack volume value of the processed part according to the crack feature data, and performing ratio processing on the crack volume value and a preset value to obtain the crack concentration value of the processed part;

[0114] Determining whether the crack concentration value is greater than a preset crack concentration value, if the crack concentration value is not greater than the preset crack concentration value, marking the processed part as a qualified part, and continuing to put the processed part into subsequent process steps for processing and production;

[0115] If the crack concentration value is greater than a preset crack concentration value, a crack characteristic model diagram of the processed part is constructed based on the crack characteristic data;

[0116] Acquire finished product engineering drawing information of the processed part after subsequent processing steps, and construct a finished product model diagram of the processed part according to the finished product engineering drawing information;

[0117] Integrate the crack feature model map with the finished product model map, and if there is an overlapping area between the model area of ​​the crack feature model map and the model area of ​​the finished product model map, retain the crack feature model map with the overlapping area in the finished product model map;

[0118] If the model area of ​​the crack feature model diagram does not overlap with the model area of ​​the finished product model diagram, the crack feature model diagram without the overlapping area is removed from the finished product model diagram;

[0119] After the integration is completed, a simulated processing model diagram is obtained, and the crack volume value of the remaining cracks in the simulated processing model diagram is calculated, and the crack volume value of the remaining cracks is processed by ratio with a preset value to obtain a simulated crack concentration value of the processed part after subsequent processing steps;

[0120] If the simulated crack concentration value is still greater than the preset crack concentration value, the processed part is marked as an unqualified part and the processed part is immediately scrapped; if the simulated crack concentration value is not greater than the preset crack concentration value, the processed part is marked as a qualified part and the processed part continues to be put into subsequent process steps for processing and production.

[0121] Among them, the subsequent processing steps include drilling, trimming, grinding, etc.

[0122] First, use an ultrasonic detector to scan the processed parts to obtain ultrasonic signals. Then, extract the features of these signals to obtain crack feature data, such as the length, depth, width and position of the crack. Next, calculate the crack volume value and compare it with the preset value to obtain the crack concentration value. If the crack concentration value is not greater than the preset value, the processed part is marked as qualified; otherwise, a crack feature model diagram is constructed. Subsequently, a finished product model diagram is constructed based on the finished product engineering drawing information of the processed parts. The crack feature model diagram is integrated with the finished product model diagram, and the crack features are retained or eliminated to obtain a simulated processing model diagram. The crack volume value of the remaining cracks in the simulated processing model diagram is calculated to obtain the simulated crack concentration value. If the simulated crack concentration value is still greater than the preset value, the processed part is marked as unqualified; otherwise, it is marked as qualified.

[0123] It should be noted that during the laser engraving process, due to the concentrated laser energy and extremely high temperature, the material may be partially melted or the thermal stress may be too large, which may cause micro cracks inside the material and form cracks. If the number of cracks is large, the overall crack concentration will exceed the standard, resulting in the crack concentration of the part not meeting the standard. However, after subsequent processing steps, such as drilling, trimming and grinding, the crack concentration of the part may be reduced: For example, during the drilling process, the rotary cutting action of the drill bit can help remove part of the material in the crack area and reduce the impact of the crack on the overall structure. For example, by cutting the edge, the exposed part of the crack can be removed, which directly reduces the area of ​​the crack and thus reduces the crack concentration. In summary, through these subsequent processing steps, the impact of the crack can be repaired or mitigated, and the cracks of the cracks and the impact of the cracks on the overall performance can be reduced, so that the crack concentration of the processed parts can meet the standards again. Therefore, in the laser engraving process step, if the crack concentration of the processed parts is too high, it is necessary to further analyze whether the crack concentration of the processed parts can return to the normal value after the subsequent processing steps. If it can return to the normal value, the processed parts are marked as qualified parts, and the processed parts can continue to be put into the subsequent process steps for processing and production. In this way, the scrap cost can be effectively reduced; however, if the crack concentration of the processed parts cannot return to the normal value after the subsequent processing steps, even if the parts are continued to be put into the subsequent processing steps, the finished products processed by them will eventually need to be scrapped (because their crack concentration is still too high). Therefore, in the present invention, if it is identified that the crack concentration of the processed parts cannot return to the normal value, the processed parts are scrapped in the laser engraving process step to prevent the processed parts from flowing into the subsequent steps, thereby effectively reducing production costs.

[0124] like Figure 3 As shown, the second aspect of the present invention discloses an off-road motorcycle parts production process optimization system based on intelligent adjustment, the off-road motorcycle parts production process optimization system includes a memory 60 and a processor 80, the memory 60 stores an off-road motorcycle parts production process optimization method program, when the off-road motorcycle parts production process optimization method program is executed by the processor 80, the following steps are implemented:

[0125] Acquire a feature image of a part being processed in a laser engraving device at a preset time point, perform edge detection processing on the feature image, and obtain shape features of the part being processed;

[0126] Performing Hough transform processing on the shape features of the processed part to obtain a number of boundary points, performing redundancy reduction processing on the boundary points, and reconstructing a three-dimensional model diagram of the actual shape of the processed part according to the reduced-redundancy boundary points;

[0127] Obtaining a finished laser engraving engineering drawing of a processed part, and constructing a standard shape three-dimensional model drawing of the processed part according to the finished laser engraving engineering drawing; and evaluating and analyzing the processed part according to the actual shape three-dimensional model drawing and the standard shape three-dimensional model drawing;

[0128] If the processed part is qualified, the laser engraving device continues to process and produce the next part; if the processed part is unqualified, a fault analysis is performed on the laser engraving device, and the laser engraving device is adjusted and processed according to the fault analysis results.

[0129] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0130] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0131] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0132] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0133] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0134] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for optimizing the production process of off-road motorcycle parts based on intelligent adjustment, characterized in that: The following steps are involved: Acquire a feature image of a part being processed in a laser engraving device at a preset time point, perform edge detection processing on the feature image, and obtain shape features of the part being processed; Performing Hough transform processing on the shape features of the processed part to obtain a number of boundary points, performing redundancy reduction processing on the boundary points, and reconstructing a three-dimensional model diagram of the actual shape of the processed part according to the reduced-redundancy boundary points; Obtaining a finished laser engraving engineering drawing of a processed part, and constructing a standard shape three-dimensional model drawing of the processed part according to the finished laser engraving engineering drawing; and evaluating and analyzing the processed part according to the actual shape three-dimensional model drawing and the standard shape three-dimensional model drawing; If the processed part is qualified, the laser engraving device continues to process and produce the next part; if the processed part is unqualified, the laser engraving device is analyzed for failure, and the laser engraving device is adjusted and processed according to the failure analysis results; The feature image is subjected to edge detection processing to obtain the shape features of the processed parts, specifically: The cv2.filter2D function in the OpenCV library is used to calculate the gradients of the feature image in the horizontal and vertical directions; the horizontal gradient matrix and the vertical gradient matrix are constructed according to the gradients of the feature image in the horizontal and vertical directions; The gradient matrix in the horizontal direction is added to the gradient matrix in the vertical direction to obtain an edge strength map; wherein the edge strength map reflects the significance of the edge of each pixel point, and the edge strength of the pixel point is related to the gradient value of the pixel; Traversing each pixel point in the edge intensity map, and comparing the pixel value of each pixel point with the pixel values ​​of other pixels points in the preset area; If the pixel value of a certain pixel is greater than the pixel values ​​of other pixels in the preset area, the pixel is marked as a local maximum pixel, that is, a strong edge point; if the pixel value of a certain pixel is not greater than the pixel value of one or more pixels in the preset area, the pixel is marked as a non-local maximum pixel, that is, a weak edge point; If a certain pixel point in the edge intensity map is a strong edge point, the position node corresponding to the pixel point in the edge intensity map is set to black; if a certain pixel point in the edge intensity map is a weak edge point, the position node corresponding to the pixel point in the edge intensity map is set to white; and so on, until all the pixel points are set, a binary image is obtained; Use the findContours function in the OpenCV library to extract the edge contours in the image from the binary image and obtain the shape features of the processed parts; The process of evaluating and analyzing the processed parts according to the actual shape three-dimensional model diagram and the standard shape three-dimensional model diagram is as follows: Constructing a registration space, and inputting the actual shape three-dimensional model image and the standard shape three-dimensional model image into the registration space; Retrieving reference datums of the actual shape three-dimensional model image and the standard shape three-dimensional model image in the registration space, and performing alignment processing on the reference datums of the two three-dimensional model images, so as to complete the registration process of the actual shape three-dimensional model image and the standard shape three-dimensional model image in the registration space; Retrieving in the registration space a model area where the actual shape three-dimensional model image and the standard shape three-dimensional model image overlap with each other, and setting the overlapping model area as a white area; and retrieving in the registration space a model area where the actual shape three-dimensional model image and the standard shape three-dimensional model image do not overlap with each other, and setting the non-overlapping model area as a black area; Calculating the area volume of the white area and the area volume of the black area; dividing the area volume of the white area by the area volume of the black area to obtain the overlap between the actual shape three-dimensional model image and the standard shape three-dimensional model image; If the overlap between the actual shape 3D model image and the standard shape 3D model image is not greater than a preset overlap threshold, the processed part is marked as a defective part; if the overlap between the actual shape 3D model image and the standard shape 3D model image is greater than a preset overlap threshold, the processed part is marked as a qualified part; If the processed parts are unqualified, the laser engraving equipment is analyzed for faults, and the laser engraving equipment is regulated and processed according to the results of the fault analysis, specifically: If the processed part is marked as a defective part, the position area corresponding to the black area model between the actual shape 3D model image and the standard shape 3D model image is obtained in the registration space, which is defined as the carving abnormal area of ​​the processed part; Obtaining a device operation log of the laser engraving device, and obtaining, according to the device operation log, a historical processing time period corresponding to the laser engraving device performing engraving processing on the engraving abnormal area, and defining the historical processing time period as the abnormal processing time period; and obtaining, according to the equipment operation log, historical operation parameters of each sub-equipment in the laser engraving equipment when processing is performed during the abnormal processing time period; Constructing a Bayesian network model, and pre-training the Bayesian network model to obtain a trained Bayesian network model; Importing the historical operating parameters of each sub-device into the trained Bayesian network model for fault prediction to obtain the fault probability of each sub-device; If the failure probability of each sub-device in the laser engraving device is not greater than the preset failure probability, the laser engraving device continues to process and produce the next part, and the Bayesian network model is used to continuously monitor the failure status of each sub-device in the laser engraving device; If the failure probability of one or more sub-devices in the laser engraving device is greater than a preset failure probability, the laser engraving device is controlled to stop production, and the sub-devices with a failure probability greater than the preset failure probability are marked, and fault information is generated according to the marked sub-devices, and the fault information is sent to a preset platform for display; Among them, a Bayesian network model is constructed, and the Bayesian network model is trained in advance to obtain a trained Bayesian network model, specifically: Obtain the historical fault events of each sub-device through the device operation log, and calculate the similarity between the historical fault events through the cosine similarity algorithm; Merge historical fault events with similarity greater than a preset similarity to cluster historical fault events with the same fault type to obtain a set of historical fault events of various fault types that have occurred on the sub-device; Performing feature engineering processing on a historical fault event set of various fault types that have occurred in the sub-device to obtain fault feature data of various fault types that have occurred in the sub-device; and dividing the fault feature data into a fault feature data training set and a fault feature data test set; Constructing a Bayesian network, initializing nodes of the Bayesian network according to various fault types that have occurred in the sub-device, and initializing edges of the Bayesian network according to a training set of fault feature data of various fault types; The maximum likelihood estimation algorithm is introduced, and the conditional probability between each node and the edge is estimated based on the maximum likelihood estimation algorithm. A conditional probability table is constructed according to the conditional probability between each node and the edge to obtain a Bayesian network model. Use the fault feature data test set to evaluate the accuracy and recall of the Bayesian network model. If the accuracy and recall of the Bayesian network model meet the preset requirements, it means that the model training is complete, and the trained Bayesian network model is output; If the accuracy and recall of the Bayesian network model do not meet the preset requirements, the conditional probability table is adjusted until the accuracy and recall of the Bayesian network model meet the preset requirements, and the trained Bayesian network model is output.

2. The method for optimizing the production process of off-road motorcycle parts based on intelligent adjustment according to claim 1, characterized in that: The shape features of the processed part are processed by Hough transform to obtain a number of boundary points, and the boundary points are reduced in redundancy. The actual shape three-dimensional model diagram of the processed part is reconstructed according to the reduced boundary points, specifically: Performing polar coordinate Hough transformation on the shape features of the processed part to convert the shape features of the processed part into a plurality of straight line segments; Each straight line segment is converted into Hough space, and the peak value of each straight line segment is found in Hough space to obtain several peak values; each peak value is reversely mapped back to the feature image to obtain several boundary points of the feature image; Calculate the Mahalanobis distance between each boundary point, and compare the Mahalanobis distance between each boundary point with a preset threshold; If the Mahalanobis distance between two boundary points is less than a preset threshold, the two boundary points are defined as redundant boundary points and the two redundant boundary points are merged; This process is repeated in this way until all redundant boundary points are processed and corrected, and the reduced-redundancy boundary points are obtained. Randomly select a reduced-redundancy boundary point as a reference point, construct a three-dimensional coordinate system based on the reference point, obtain the three-dimensional coordinate values ​​of each reduced-redundancy boundary point in the three-dimensional coordinate system, and generate a boundary point coordinate collection according to the three-dimensional coordinate values ​​of each reduced-redundancy boundary point in the three-dimensional coordinate system; A three-dimensional model diagram of the actual shape of the processed part is constructed according to the boundary point coordinate collection.

3. An off-road motorcycle parts production process optimization system based on intelligent adjustment, characterized in that: The off-road motorcycle parts production process optimization system includes a memory and a processor, wherein the memory stores an off-road motorcycle parts production process optimization method program. When the off-road motorcycle parts production process optimization method program is executed by the processor, the off-road motorcycle parts production process optimization method steps based on intelligent adjustment as described in any one of claims 1-2 are implemented.