Gear crack recognition device and method

Abstract

The invention relates to a gear crack recognition device and method, and belongs to the technical field of gear recognized.The gear crack recognition method comprises the steps that clustering processing is conducted on preprocessed image data to obtain clustered image data, and therefore a Mahalanobis distance measurement method and a genetic algorithm are fused to construct a crack image training set based on the clustered image data; constructing a gear crack recognition model based on the crack image training set, obtaining real-time gear crack image data, obtaining a recognition result according to the real-time gear crack image data and the gear crack recognition model, and finally obtaining a first analysis result or a second analysis result based on the recognition result. And related early warning information is generated according to the second analysis result. By fusing the Mahalanobis distance measurement method and the genetic algorithm, the recognition precision of the gear crack can be improved, so that the crack of the machined gear can be quickly recognized, and the quality of the machined gear can be quickly evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of crack identification devices, and in particular to a gear crack identification device and method. Background Art

[0002] CNC machining technology is a highly automated manufacturing technology. Its core lies in the use of computer numerical control systems to accurately control the movement and processing of machine tools. This technology integrates the achievements of multiple fields such as computer technology, automatic control technology, sensor technology and mechanical processing technology. It is an important means to achieve high-precision, high-efficiency and high-quality processing in modern manufacturing. The basic principle of CNC machining technology is to pre-set various movements and operations in the processing process in the computer through programming. In CNC machining technology, the gear processing process is a complicated process. Since the gear is affected by the cutting force during the processing, certain cracks may be generated. Therefore, the gear needs to be cracked after the processing is completed. Small cracks are difficult to see with the naked eye, and naked eye observation consumes a lot of material, time, and manpower, which is not conducive to the rapid identification of abnormal cracked gears. Summary of the invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a gear crack identification device and method.

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

[0005] A first aspect of the present invention provides a gear crack identification method, comprising the following steps:

[0006] Acquire historical image data of gear cracks, and perform preprocessing on the historical image data of gear cracks to obtain preprocessed image data, and perform clustering on the preprocessed image data to obtain clustered image data;

[0007] Combining Mahalanobis distance metric and genetic algorithm, constructing a crack image training set based on the clustered image data;

[0008] Building a gear crack recognition model based on the crack image training set, acquiring real-time gear crack image data, and acquiring recognition results based on the real-time gear crack image data and the gear crack recognition model;

[0009] A first analysis result or a second analysis result is obtained based on the identification result, and relevant warning information is generated according to the second analysis result.

[0010] Furthermore, in the present method, historical image data of the gear crack is obtained, and preprocessing the historical image data of the gear crack is performed to obtain preprocessed image data, which specifically includes:

[0011] Acquire historical image data of gear cracks, convert the historical image data of gear cracks into grayscale images, acquire the grayscale images, introduce a Sobel operator, calculate the gradients of pixels in the grayscale images in various directions by the Sobel operator, and obtain horizontal and vertical gradients of the grayscale images;

[0012] Calculating the gradient amplitude of each pixel of the grayscale image based on the horizontal gradient and the vertical gradient, and calculating the average gradient amplitude and the gradient amplitude standard deviation of the grayscale image according to the gradient amplitude of each pixel;

[0013] Determine the spatial smoothness of the image according to the average gradient amplitude and the gradient amplitude standard deviation, and mark the area where the spatial smoothness is less than the preset spatial smoothness;

[0014] By marking, image enhancement processing is performed on the area where the spatial domain smoothness is less than the preset spatial domain smoothness, an image after image enhancement is obtained, and the image after image enhancement is output as pre-processed image data.

[0015] Furthermore, in the present method, clustering is performed on the preprocessed image data to obtain clustered image data, which specifically includes:

[0016] A fuzzy clustering algorithm is introduced, and the number of cluster centers is initialized, all pixels in the preprocessed image data are used as sample data, and the Euclidean distance value from each pixel to the cluster center is calculated based on the number of cluster centers;

[0017] Obtaining the cluster center corresponding to the minimum Euclidean distance value as the cluster of the corresponding pixel point, classifying each pixel point, obtaining the cluster corresponding to each pixel point, and segmenting the preprocessed image data according to the cluster corresponding to each pixel point;

[0018] By segmentation, clustered image data is obtained, and the clustered image data is output as an output result.

[0019] Furthermore, in this method, the Mahalanobis distance metric and the genetic algorithm are integrated to construct a crack image training set based on the clustered image data, specifically:

[0020] A Mahalanobis distance measurement method and a genetic algorithm are introduced, and crack feature extraction is performed on the clustered image data through a feature pyramid network to obtain crack feature data of the clustered image data;

[0021] Constructing and initializing a crack image training set, randomly selecting training items in the crack image training set from the crack feature data of the clustered image data, and calculating Mahalanobis distance values ​​between the training items in the crack image training set by using the Mahalanobis distance metric method;

[0022] Determine whether the Mahalanobis distance values ​​between the training items in the crack image training set are all less than a preset Mahalanobis distance threshold;

[0023] When the Mahalanobis distance values ​​between the training items in the crack image training set are all less than a preset Mahalanobis distance threshold, outputting the crack image training set;

[0024] When the Mahalanobis distance values ​​between the training items in the crack image training set are all less than the preset Mahalanobis distance threshold, the genetic algebra is set based on the genetic algorithm, and genetic iteration is performed, the Mahalanobis distance metric method is adjusted to calculate the training items in the crack image training set until they are all less than the preset Mahalanobis distance threshold, and the crack image training set is output.

[0025] Furthermore, in this method, a gear crack recognition model is constructed based on the crack image training set, real-time gear crack image data is obtained, and recognition results are obtained according to the real-time gear crack image data and the gear crack recognition model, specifically including:

[0026] Building a gear crack recognition model based on a deep neural network, and inputting the crack image training set into the gear crack recognition model for training, to obtain a trained gear crack recognition model;

[0027] Real-time gear crack image data is acquired, and the real-time gear crack image data is input into the gear crack recognition model for recognition, to acquire a recognition result.

[0028] Furthermore, in this method, obtaining a first analysis result or a second analysis result based on the recognition result, and generating relevant warning information according to the second analysis result specifically includes:

[0029] When there is no crack in the recognition result, generating prompt information that the gear is normal, taking the prompt information that the gear is normal as the first analysis result, and outputting the first analysis result;

[0030] When cracks exist in the recognition result, a prompt message indicating gear abnormality is generated, and the prompt message indicating gear abnormality is used as a second analysis result, a search tag is constructed based on the second analysis result, and a search is performed through big data based on the search tag;

[0031] The process flow related to the generation of cracks is obtained through big data retrieval, and relevant early warning information is generated based on the process flow related to the generation of cracks.

[0032] A second aspect of the present invention provides a gear crack identification device, comprising a support frame, on which a controller, a visual detector and a conveyor line are arranged.

[0033] conveying the gear to be inspected through the conveyor line;

[0034] The visual inspection instrument is used to collect a real-time gear image, and the controller is used to perform crack recognition on the real-time gear image;

[0035] The controller includes a gear crack identification method program, and the gear crack identification method program includes the steps of the gear crack identification method as described in any one of claims 1-6.

[0036] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0037] The present invention obtains historical image data of gear cracks, preprocesses the historical image data of gear cracks, obtains preprocessed image data, clusters the preprocessed image data, and obtains clustered image data, thereby integrating the Mahalanobis distance metric and the genetic algorithm, and constructs a crack image training set based on the clustered image data; then constructs a gear crack recognition model based on the crack image training set, obtains real-time gear crack image data, and obtains recognition results based on the real-time gear crack image data and the gear crack recognition model, and finally obtains the first analysis result or the second analysis result based on the recognition result, and generates relevant early warning information based on the second analysis result. The present invention can improve the recognition accuracy of gear cracks by integrating the Mahalanobis distance metric and the genetic algorithm, so that cracks in processed gears can be quickly identified and the quality of processed gears can be quickly evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] 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.

[0039] Figure 1 The overall structural diagram of the gear crack identification device is shown;

[0040] Figure 2 The overall method flow chart of the gear crack identification method is shown.

[0041] In the figure:

[0042] 1. Support frame, 2. Controller, 3. Visual inspection instrument, 4. Conveyor line. DETAILED DESCRIPTION

[0043] In order to more clearly understand the above-mentioned purposes, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. These drawings are all simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the structures related to the present invention. It should be noted that the embodiments in this application and the features in the embodiments can be combined with each other without conflict.

[0044] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the scope of protection of the present application. In addition, the terms "first", "second", etc. are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the invention, unless otherwise specified, "multiple" means two or more.

[0045] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0046] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. The preferred embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thoroughly understood.

[0047] like Figure 1As shown, the first aspect of the present invention provides a gear crack identification device, including a support frame 1, on which a controller 2, a visual detector 3 and a conveyor line 4 are arranged.

[0048] The gear to be inspected is conveyed via conveyor line 4;

[0049] The real-time gear image is collected by the visual inspection device 3, and crack recognition is performed on the real-time gear image by the controller 2;

[0050] The controller 2 includes a gear crack identification method program, and the gear crack identification method program includes the following steps:

[0051] S102: acquiring historical image data of gear cracks, and preprocessing the historical image data of gear cracks to acquire preprocessed image data, and performing clustering processing on the preprocessed image data to acquire clustered image data;

[0052] S104: integrating Mahalanobis distance metric and genetic algorithm to construct a crack image training set based on the clustered image data;

[0053] S106: constructing a gear crack recognition model based on the crack image training set, acquiring real-time gear crack image data, and acquiring recognition results according to the real-time gear crack image data and the gear crack recognition model;

[0054] S108: Obtain the first analysis result or the second analysis result based on the identification result, and generate relevant warning information according to the second analysis result.

[0055] It should be noted that the present invention can improve the recognition accuracy of gear cracks by integrating the Mahalanobis distance metric method and the genetic algorithm, thereby making it possible to quickly identify cracks in processed gears and quickly perform quality assessment on the processed gears.

[0056] Furthermore, in the present method, historical image data of the gear crack is obtained, and the historical image data of the gear crack is preprocessed to obtain preprocessed image data, which specifically includes:

[0057] Obtain historical image data of gear cracks, convert the historical image data of gear cracks into grayscale images, obtain the grayscale images, introduce the Sobel operator, calculate the gradients of pixels in the grayscale images in various directions through the Sobel operator, and obtain the horizontal gradient and vertical gradient of the grayscale images;

[0058] The gradient amplitude of each pixel of the grayscale image is calculated based on the horizontal gradient and the vertical gradient, and the average gradient amplitude and the gradient amplitude standard deviation of the grayscale image are calculated according to the gradient amplitude of each pixel;

[0059] Determine the spatial smoothness of the image according to the average gradient amplitude and the gradient amplitude standard deviation, and mark the area where the spatial smoothness is less than the preset spatial smoothness;

[0060] By marking, image enhancement processing is performed on the area whose spatial domain smoothness is less than the preset spatial domain smoothness, an image after image enhancement is obtained, and the image after image enhancement is output as preprocessed image data.

[0061] It should be noted that this method can obtain an enhanced image, making the features of the image more obvious.

[0062] Furthermore, in the present method, clustering is performed on the preprocessed image data to obtain clustered image data, which specifically includes:

[0063] The fuzzy clustering algorithm is introduced, and the number of cluster centers is initialized. All the pixels in the preprocessed image data are used as sample data, and the Euclidean distance value from each pixel to the cluster center is calculated based on the number of cluster centers.

[0064] The cluster center corresponding to the minimum Euclidean distance value is obtained as the cluster of the corresponding pixel point, and each pixel point is classified to obtain the cluster corresponding to each pixel point, and the preprocessed image data is segmented according to the cluster corresponding to each pixel point;

[0065] Through segmentation, clustered image data is obtained, and the clustered image data is output as an output result.

[0066] It should be noted that this method can segment the clustered image data to make the crack features more obvious.

[0067] Furthermore, in this method, the Mahalanobis distance metric and the genetic algorithm are integrated to construct a crack image training set based on the clustered image data, specifically:

[0068] The Mahalanobis distance metric and genetic algorithm are introduced, and crack features are extracted from clustered image data through a feature pyramid network to obtain crack feature data of the clustered image data;

[0069] Constructing and initializing a crack image training set, randomly selecting training items in the crack image training set from crack feature data of clustered image data, and calculating Mahalanobis distance values ​​between training items in the crack image training set by using Mahalanobis distance metric method;

[0070] Determine whether the Mahalanobis distance values ​​between the training items in the crack image training set are all less than a preset Mahalanobis distance threshold;

[0071] When the Mahalanobis distance values ​​between the training items in the crack image training set are all less than the preset Mahalanobis distance threshold, the crack image training set is output;

[0072] When the Mahalanobis distance values ​​between the training items in the crack image training set are all less than the preset Mahalanobis distance threshold, the genetic algebra is set based on the genetic algorithm, and genetic iteration is performed to adjust the Mahalanobis distance metric method to calculate the training items in the crack image training set until they are all less than the preset Mahalanobis distance threshold, and the crack image training set is output.

[0073] It should be noted that, since the crack feature data of the clustered image data contains other feature data, when the Mahalanobis distance values ​​between the training items in the crack image training set are all less than the preset Mahalanobis distance threshold, the genetic algebra is set based on the genetic algorithm, and genetic iteration is performed to adjust the Mahalanobis distance metric method to calculate the training items in the crack image training set until they are all less than the preset Mahalanobis distance threshold, so that the data of the training set remains highly identical, which can avoid the influence of multi-scale feature data on the model.

[0074] Furthermore, in this method, a gear crack recognition model is constructed based on a crack image training set, real-time gear crack image data is obtained, and recognition results are obtained according to the real-time gear crack image data and the gear crack recognition model, specifically including:

[0075] A gear crack recognition model is constructed based on a deep neural network, and a crack image training set is input into the gear crack recognition model for training to obtain a trained gear crack recognition model;

[0076] Real-time gear crack image data is obtained, and the real-time gear crack image data is input into a gear crack recognition model for recognition to obtain a recognition result.

[0077] Furthermore, in this method, obtaining the first analysis result or the second analysis result based on the recognition result, and generating relevant warning information according to the second analysis result, specifically includes:

[0078] When there is no crack in the recognition result, a prompt message indicating that the gear is normal is generated, and the prompt message indicating that the gear is normal is used as the first analysis result, and the first analysis result is output;

[0079] When cracks are found in the recognition result, a prompt message indicating gear abnormality is generated, and the prompt message indicating gear abnormality is used as the second analysis result. A search tag is constructed based on the second analysis result, and a search is performed through big data based on the search tag.

[0080] The process flow related to the generation of cracks is obtained through big data retrieval, and relevant early warning information is generated based on the process flow related to the generation of cracks.

[0081] It should be noted that the process flow related to the generation of cracks includes quenching, cutting, grinding and other processes. This method can quickly find abnormal data.

[0082] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0083] The above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the aforementioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A gear crack identification method, characterized in that: The following steps are involved: Acquire historical image data of gear cracks, and perform preprocessing on the historical image data of gear cracks to obtain preprocessed image data, and perform clustering on the preprocessed image data to obtain clustered image data; Combining Mahalanobis distance metric and genetic algorithm, constructing a crack image training set based on the clustered image data; Building a gear crack recognition model based on the crack image training set, acquiring real-time gear crack image data, and acquiring recognition results based on the real-time gear crack image data and the gear crack recognition model; A first analysis result or a second analysis result is obtained based on the identification result, and relevant warning information is generated according to the second analysis result.

2. A gear crack identification method according to claim 1, characterized in that: Acquiring historical image data of the gear crack, and preprocessing the historical image data of the gear crack to obtain preprocessed image data, specifically including: Acquire historical image data of gear cracks, convert the historical image data of gear cracks into a grayscale image, acquire the grayscale image, introduce a Sobel operator, calculate the gradients of pixels in the grayscale image in various directions by the Sobel operator, and obtain the horizontal gradient and vertical gradient of the grayscale image; Calculating the gradient amplitude of each pixel of the grayscale image based on the horizontal gradient and the vertical gradient, and calculating the average gradient amplitude and the gradient amplitude standard deviation of the grayscale image according to the gradient amplitude of each pixel; Determine the spatial domain smoothness of the image according to the average gradient amplitude and the gradient amplitude standard deviation, and mark the area where the spatial domain smoothness is less than the preset spatial domain smoothness; By marking, image enhancement processing is performed on the area where the spatial domain smoothness is less than the preset spatial domain smoothness, an image after image enhancement is obtained, and the image after image enhancement is output as pre-processed image data.

3. A gear crack identification method according to claim 1, characterized in that: Performing clustering processing on the preprocessed image data to obtain clustered image data specifically includes: A fuzzy clustering algorithm is introduced, and the number of cluster centers is initialized, all pixels in the preprocessed image data are used as sample data, and the Euclidean distance value from each pixel to the cluster center is calculated based on the number of cluster centers; Obtaining the cluster center corresponding to the minimum Euclidean distance value as the cluster of the corresponding pixel point, classifying each pixel point, obtaining the cluster corresponding to each pixel point, and segmenting the preprocessed image data according to the cluster corresponding to each pixel point; By segmentation, clustered image data is obtained, and the clustered image data is output as an output result.

4. A gear crack identification method according to claim 1, characterized in that: The Mahalanobis distance metric and the genetic algorithm are integrated to construct a crack image training set based on the clustered image data, specifically: A Mahalanobis distance metric method and a genetic algorithm are introduced, and crack feature extraction is performed on the clustered image data through a feature pyramid network to obtain crack feature data of the clustered image data; Constructing and initializing a crack image training set, randomly selecting training items in the crack image training set from the crack feature data of the clustered image data, and calculating Mahalanobis distance values ​​between the training items in the crack image training set by using the Mahalanobis distance metric method; Determine whether the Mahalanobis distance values ​​between the training items in the crack image training set are all less than a preset Mahalanobis distance threshold; When the Mahalanobis distance values ​​between the training items in the crack image training set are all less than a preset Mahalanobis distance threshold, outputting the crack image training set; When the Mahalanobis distance values ​​between the training items in the crack image training set are all less than the preset Mahalanobis distance threshold, the genetic algebra is set based on the genetic algorithm, and genetic iteration is performed, the Mahalanobis distance metric method is adjusted to calculate the training items in the crack image training set until they are all less than the preset Mahalanobis distance threshold, and the crack image training set is output.

5. A gear crack identification method according to claim 1, characterized in that: Building a gear crack recognition model based on the crack image training set, acquiring real-time gear crack image data, and acquiring recognition results according to the real-time gear crack image data and the gear crack recognition model, specifically includes: Building a gear crack recognition model based on a deep neural network, and inputting the crack image training set into the gear crack recognition model for training, to obtain a trained gear crack recognition model; Real-time gear crack image data is acquired, and the real-time gear crack image data is input into the gear crack recognition model for recognition, to acquire a recognition result.

6. A gear crack identification method according to claim 1, characterized in that: Obtaining a first analysis result or a second analysis result based on the identification result, and generating relevant warning information according to the second analysis result, specifically includes: When there is no crack in the recognition result, generating prompt information that the gear is normal, taking the prompt information that the gear is normal as the first analysis result, and outputting the first analysis result; When cracks exist in the recognition result, a prompt message indicating gear abnormality is generated, and the prompt message indicating gear abnormality is used as a second analysis result, a search tag is constructed based on the second analysis result, and a search is performed through big data based on the search tag; The process flow related to the generation of cracks is obtained through big data retrieval, and relevant early warning information is generated based on the process flow related to the generation of cracks.

7. A gear crack identification device, characterized in that: It includes a support frame, on which a controller, a visual detector and a conveyor line are arranged. conveying the gear to be inspected through the conveyor line; The visual inspection instrument is used to collect a real-time gear image, and the controller is used to perform crack recognition on the real-time gear image; The controller includes a gear crack identification method program, and the gear crack identification method program includes the steps of the gear crack identification method as described in any one of claims 1-6.

Images