Structural defect positioning method and system of pneumatic grinding machine

Abstract

The invention relates to the related technical field of grinding machines, in particular to a structural defect positioning method and system for a pneumatic grinding machine, and the method comprises the steps: judging the structural applicability of the grinding machine by matching detection features, and obtaining and storing an initial detection result; monitoring the operation data, calculating a loss value, and identifying abnormal parts and the probability thereof by using an abnormal positioning model; and when the probability reaches the standard, comparing the current and initial images to identify abnormity, fusing abnormal information to position structural defects and feeding back, thereby solving the technical problem that the structural health state of the pneumatic grinding machine is difficult to accurately evaluate in a complex working environment, realizing real-time monitoring of key parameters such as temperature, vibration and pressure in the operation process of the grinding machine, and improving the accuracy of the operation of the grinding machine. And in combination with the use requirement of an application scene, full-period structure defect positioning analysis is carried out, so that the scene adaptability of structure defect positioning of the pneumatic grinding machine is improved, the defects existing in the structure are found in time and accurately positioned, and the technical effects of safe and efficient operation of the pneumatic grinding machine are ensured.

Description

Technical Field

[0001] The invention relates to the technical field related to grinding machines, and in particular to a method and system for locating structural defects of a pneumatic grinding machine. Background Art

[0002] Pneumatic grinder is an industrial tool driven by compressed air. Due to its advantages of high speed, high efficiency and easy operation, it has been widely used in metal processing, automobile maintenance, shipbuilding and other industries.

[0003] When using a pneumatic grinder, compressed air enters the machine body through the air pipe interface to drive the grinding wheel to rotate at high speed. However, during long-term and high-intensity use, the pneumatic grinder is prone to various mechanical failures caused by wear, looseness, and breakage of structural parts, which seriously affect the working efficiency and safety performance of the pneumatic grinder.

[0004] For the structural defects of pneumatic grinders, it is common to regularly check the structural defects of pneumatic grinders through the maintenance records of the structural defects of pneumatic grinders, combined with the actual service life and strength of the grinders. However, regular defect inspections can only eliminate common and frequent structural defects of pneumatic grinders to a certain extent, and it is difficult to discover some hidden or sudden structural problems in real time, and subtle structural defects may be missed. In addition, working in harsh environments such as high load, high dust, and high temperature will accelerate the aging and wear of the structural parts of the pneumatic grinder, thereby aggravating the failure risks of various structural defects. It is impossible to thoroughly check out various subtle structural defects of pneumatic grinders by simply shortening the regular defect inspection cycle.

[0005] In summary, the prior art has a technical problem that it is difficult to accurately evaluate the structural health status of a pneumatic grinder in a complex working environment. Summary of the invention

[0006] The present application provides a method and system for locating structural defects of a pneumatic grinder, aiming to solve the technical problem in the prior art that it is difficult to accurately evaluate the structural health status of a pneumatic grinder in a complex working environment.

[0007] In view of the above problems, the technical solution to implement this application is:

[0008] On the one hand, the present application provides a method for locating structural defects of a pneumatic grinder, wherein the method comprises: adaptively matching the product structure of the pneumatic grinder with the scene application features through a preset detection feature set to obtain a detection result of the initial structure;

[0009] For the pneumatic grinders with qualified test results, the test results and initial images are archived, and the operation data is monitored through monitoring equipment. The operation monitoring data includes temperature information, vibration information and monitoring position, and the loss value of the operation monitoring data is evaluated, wherein the monitoring position has a positional correspondence with the component;

[0010] Based on the operation monitoring data and its loss value, abnormal analysis is performed through an abnormal location model to obtain abnormal parts and abnormal probability;

[0011] When the abnormal probability of the abnormal component reaches a preset requirement, the current structure image of the pneumatic grinder is collected, and feature recognition and comparison is performed using the current structure image and the initial image to obtain feature comparison results and abnormal components with abnormal feature comparison results;

[0012] According to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, abnormal fusion is performed to obtain structural defect positioning information for feedback.

[0013] On the other hand, the present application provides a structural defect positioning system for a pneumatic grinder, wherein the system comprises: an adaptive matching module, for adaptively matching the product structure of the pneumatic grinder with the scene application feature through a preset detection feature set to obtain a detection result of the initial structure;

[0014] An operation data monitoring module is used to archive the test results and initial images of the pneumatic grinders with qualified test results, and to monitor the operation data through monitoring equipment. The operation monitoring data includes temperature information, vibration information and monitoring position, and to evaluate the loss value of the operation monitoring data, wherein the monitoring position has a positional correspondence with the component;

[0015] An abnormality analysis module, used to perform abnormality analysis based on the operation monitoring data and its loss value through an abnormality location model to obtain abnormal parts and abnormality probabilities;

[0016] A feature recognition and comparison module is used to collect the current structure image of the pneumatic grinder when the abnormal probability of the abnormal component reaches a preset requirement, and use the current structure image and the initial image to perform feature recognition and comparison to obtain a feature comparison result and an abnormal component with an abnormal feature comparison result;

[0017] The abnormal fusion module is used to perform abnormal fusion according to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, and obtain structural defect positioning information for feedback.

[0018] In summary, one or more technical solutions provided in this application realize real-time monitoring of key parameters such as temperature, vibration, pressure, etc. of the grinding machine during operation, and conduct full-cycle structural defect location analysis in combination with the usage requirements of the application scenario to improve the scenario adaptability of the structural defect location of the pneumatic grinding machine, timely discover the defects in the structure and accurately locate them, and ensure the technical effect of safe and efficient operation of the pneumatic grinding machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A possible flow chart of a method for locating structural defects of a pneumatic grinder is provided for the present application;

[0020] Figure 2 A possible flow chart of constructing an abnormality location model in a method for locating structural defects of a pneumatic grinder is provided for this application;

[0021] Figure 3 A possible structural schematic diagram of a structural defect locating system of a pneumatic grinder is provided for the present application.

[0022] Explanation of the reference numerals: adaptive matching module M100, operation data monitoring module M200, anomaly analysis module M300, feature recognition and comparison module M400, anomaly fusion module M500. DETAILED DESCRIPTION

[0023] Embodiment 1

[0024] The present application is described in detail below with reference to the accompanying drawings. Figure 1 As shown, the present application provides a method for locating structural defects of a pneumatic grinder, wherein the method comprises:

[0025] S1: Adaptively match the product structure of the pneumatic grinder with the scene application features through the preset detection feature set to obtain the detection result of the initial structure;

[0026] Furthermore, the product structure of the pneumatic grinder is adaptively matched with the scene application characteristics through a preset detection feature set to obtain the detection result of the initial structure. The method of this application includes:

[0027] S11: According to the structure of the pneumatic grinder, the connection relationship and transmission relationship of the pneumatic grinder parts are decomposed, and the structural relationship of the pneumatic grinder is constructed;

[0028] S12: Acquire an initial image of the pneumatic grinder, where the initial image is an image of the internal structure of the pneumatic grinder before use;

[0029] S13: Analyze the scene application characteristics according to the application scene of the pneumatic grinder to obtain the scene structure fault characteristics, where the scene structure fault characteristics are distinguishing fault characteristics obtained by performing feature unidirectional comparison on multiple scene application cases, and obtain the faulty components corresponding to the distinguishing fault characteristics;

[0030] S14: Based on the faulty components corresponding to the distinguishing fault features, detection enhancement features are set, and the initial image is added to the preset detection feature set to perform structural positioning detection, and the detection results are obtained. If an abnormality is found, the positioning is marked and a reminder message is sent.

[0031] Known, common structural defects include: body design problems, machine application problems, among which the body design is unreasonable: the body design of some pneumatic grinders may not be compact enough, resulting in a large space occupied during use, which is not convenient for operation in a small or complex working environment. In addition, if the heat dissipation performance of the body is not good, it may affect the performance of the machine or even cause damage due to overheating after long-term use. In addition, if the maintenance of the grinder is not in place during use, or structural abnormalities are not discovered in time, there are safety hazards.

[0032] Based on this, the present application evaluates the current structure according to the needs of the scene or the structure prone to abnormalities: adjustments are made to the structures that do not meet the requirements to improve the adaptability of the scene; data monitoring is performed during use for the structures that meet the requirements, mainly through the changes in temperature and vibration, to synchronously monitor the damage and abnormalities of the equipment; then, case data is used for training and model building, abnormality positioning is performed based on the monitoring data, and abnormal components and abnormal probabilities are discovered in time. Furthermore, for abnormal components with a high probability of abnormality, feature comparison is performed on static equipment in combination with images to determine the displacement, looseness, wear, etc. of the components, and the abnormal probability of data analysis is integrated to locate the defect position, so as to achieve the use requirements that are closely integrated with the application scenario, and full-cycle structural defect positioning analysis is performed to improve the use adaptability of the pneumatic grinder, timely discover the defects in the structure and accurately locate them, and ensure the safe and efficient operation of the pneumatic grinder;

[0033] In general, pneumatic grinders are used in various grinding, cutting and polishing operations. The main structure usually includes key components such as the body, grinding wheel, bearings, gears, air pipe interfaces and protective devices. According to the product structure characteristics and application scenario requirements of the pneumatic grinder, a preset detection feature set is established. The preset detection feature set includes but is not limited to the specifications and dimensions, material properties, assembly clearances, and motion characteristics of key components. The target pneumatic grinder is subjected to product structure analysis, and the actual product structure parameters are compared with the preset detection feature set one by one to evaluate whether it meets the application scenario requirements and form an initial structure detection result.

[0034] The product structure of the pneumatic grinder is adaptively matched with the scene application features through a preset detection feature set to obtain the detection result of the initial structure. Specifically, the structure of the pneumatic grinder is analyzed in detail, the connection relationship and transmission relationship between the various components are clarified, a complete pneumatic grinder structural relationship model is established, and the role and mutual influence of each component in the overall structure are clarified; before the pneumatic grinder is put into formal use, a high-definition image of the internal structure of the pneumatic grinder is collected as the initial image to ensure that the image clarity is sufficient to identify key components and structural details;

[0035] In view of the application scenarios that the pneumatic grinder will face, various working conditions and environmental conditions that may be encountered are analyzed, and representative scenario application characteristics are extracted, such as load changes, dust pollution, temperature and humidity changes, etc.; through the historical failure cases, industry reports or experimental data in the data storage unit of the structural defect positioning system of the pneumatic grinder, the failure modes that have appeared in multiple scenario application cases are compared and analyzed, and the distinguishing fault characteristics are extracted, that is, those fault types that are more likely to occur in specific scenarios and their corresponding faulty components;

[0036] Based on the extracted scene structural fault features, additional detection enhancement features are set for parts prone to failure. The additional detection enhancement features will be added to the preset detection feature set to enhance the attention and detection of the corresponding key component status. Furthermore, the preset detection feature set containing the detection enhancement features is used to perform structural positioning detection on the initial image, and image processing and recognition technology is used to analyze whether each component meets the design standards and good status requirements;

[0037] During the inspection process, once it is found that a component does not meet the preset standards or detects signs of abnormality, the abnormal position will be immediately located and marked, and a reminder message will be automatically sent to the technician so that repair or replacement measures can be taken in time. This enables a detailed inspection of the initial structure of the pneumatic grinder, and the possible structural defects are discovered and prevented in advance. At the same time, it also enhances the ability to identify potential fault hazards that may occur in specific scenarios, thereby improving the overall reliability and safety of the equipment.

[0038] S2: For the pneumatic grinder with qualified test results, the test results and initial images are archived, and the operation data is monitored through the monitoring equipment, the operation monitoring data includes temperature information, vibration information and monitoring position, and the loss value of the operation monitoring data is evaluated, wherein the monitoring position has a positional correspondence with the component;

[0039] Furthermore, the operation monitoring data includes temperature information, vibration information and monitoring location, and the loss value of the operation monitoring data is evaluated. The method of the present application includes:

[0040] S21: Based on the operation monitoring data, integrating the data of each monitoring position according to the monitoring time to obtain a temperature monitoring data set and a vibration monitoring data set;

[0041] S22: Calculate the time series loss value for the temperature monitoring data set and the vibration monitoring data set respectively according to the connection relationship of the monitoring operation time. The loss value expression is: Obtain temperature loss value and vibration loss value; where y i (t) is the monitoring data of the t-th monitoring location, y i (t+1) is the monitoring data of the continuous time nodes at the i-th monitoring position after time t, and i and n are positive integers.

[0042] In general, for pneumatic grinders that have passed the initial structural inspection and are judged to be qualified, the inspection results shall be recorded and filed in detail, and high-definition image data of the initial structure corresponding to the qualified pneumatic grinder shall be retained for comparative analysis in the middle and late stages of the service life of the pneumatic grinder; corresponding monitoring equipment shall be installed and started to monitor the key parameters of the pneumatic grinder in real time during actual operation, mainly including temperature information (such as the temperature of key parts such as bearings and motors), vibration information (such as the vibration frequency and amplitude generated when the grinding wheel rotates), and relevant data corresponding to each monitoring position and components;

[0043] During the operation of the pneumatic grinder, monitoring data is continuously collected, and temperature change curves, vibration spectrum charts, etc. are established to ensure the integrity and accuracy of data records. Mathematical algorithms are used to evaluate the loss value of the collected operation monitoring data. For example, by calculating indicators such as the degree to which the temperature change deviates from the normal range and the number of times the vibration amplitude exceeds the threshold, the loss status of each monitoring position is quantitatively evaluated, while ensuring that the positional relationship between the monitoring position and the corresponding component is clear.

[0044] The operation monitoring data includes temperature information, vibration information and monitoring position, and the loss value of the operation monitoring data is evaluated. Specifically, various monitoring data of the pneumatic grinder during operation are collected in real time, including temperature information, vibration information and monitoring position data corresponding to the position of components; the data of each monitoring position are integrated according to the monitoring time, and a temperature monitoring data set and a vibration monitoring data set are created respectively, ensuring that each data set contains different types of monitoring data of the corresponding position at the same timestamp;

[0045] The integrated temperature monitoring data set and vibration monitoring data set are sorted according to the order of monitoring operation time, and the time connection relationship of the data is established to facilitate the calculation of the time series loss value. Further, the loss value expression is substituted In the process of obtaining the temperature loss value and the vibration loss value, the following steps are performed: a monitoring position i is selected, and the monitoring data of the i-th monitoring position at the time node t is y i (t), the monitoring data of the first continuous time node at the i-th monitoring position after time t is y i (t+1), then, bring it into the time series loop for update, repeat the above steps based on it, substitute it into the loss value expression and continuously replace it, calculate the temperature loss value of the i-th monitoring position from time node t to time node t+1;

[0046] In addition, the loss value expression It also supports the implementation of real-time monitoring and early warning mechanisms. Once the loss value of a certain location reaches a critical value, an alarm signal can be triggered immediately, and maintenance measures can be taken in time to avoid the occurrence and deterioration of potential faults.

[0047] In short, the temperature loss values ​​and vibration loss values ​​calculated at all monitoring locations within their respective time periods are collated, and the quantified temperature loss values ​​and vibration loss values ​​are combined with the initial structural inspection results and high-definition image data to conduct mid-term and late-stage status comparative analysis, providing a scientific basis for structural defect assessment.

[0048] S3: Based on the operation monitoring data and its loss value, abnormality analysis is performed through an abnormality location model to obtain abnormal parts and abnormality probabilities;

[0049] Furthermore, if Figure 2 As shown, the abnormality analysis is performed through the abnormality positioning model to obtain abnormal parts and abnormality probabilities. The method of this application includes:

[0050] S31: obtaining a structural defect case set of a pneumatic grinder, integrating data of abnormal parts according to the structural defect case set, and constructing a temperature data set and a vibration data set of each abnormal part;

[0051] S32: fitting the abnormal temperature and abnormal vibration relationship using the temperature data set and the vibration data set of each abnormal component to obtain a functional relationship;

[0052] S33: Taking the abnormal defect level of each abnormal component as the top event, taking the temperature data and the vibration data as abnormal factors, and based on the functional relationship, constructing the abnormal location model.

[0053] In general, the operation monitoring data of the pneumatic grinder during operation is collected in real time, including but not limited to temperature information, vibration information and monitoring position data corresponding to the position of components; based on the monitoring data, the loss value of each component is calculated, and the loss value reflects the wear, fatigue and other conditions of the component in the working state, as well as the degree of deviation compared with the normal working state;

[0054] The calculated loss value is input into a pre-trained anomaly localization model for analysis. The anomaly localization model identifies components that may have abnormalities and calculates the probability of an abnormality in each component. When the abnormal probability of a component reaches a preset threshold, it indicates that the corresponding component may have a more serious structural defect.

[0055] Anomaly analysis is performed through the anomaly location model to obtain abnormal parts and abnormal probabilities. Specifically, a collection of structural defect cases of pneumatic grinders is collected, including various known structural defect cases and their related temperature and vibration data. The data of abnormal parts in each case are integrated, and for each abnormal part, a corresponding temperature data set and vibration data set are constructed, and the temperature and vibration conditions of the same type of parts when abnormalities occur are collected; the temperature data set and vibration data set of each abnormal part are jointly analyzed using machine learning methods, and attempts are made to fit the relationship between abnormal temperature and abnormal vibration through regression analysis, time series analysis or any other feasible algorithm. The functional relationship can reveal how abnormal temperature and abnormal vibration work together to cause the occurrence of structural defects;

[0056] Taking the abnormal defect level of abnormal parts as the top-level event, that is, taking the severity of the defect as the benchmark, temperature data and vibration data are included in the model as abnormal factors affecting the occurrence of defects; based on the fitted relationship function between abnormal temperature and abnormal vibration, an abnormal location model is constructed. The abnormal location model can calculate the abnormal probability based on the temperature and vibration data monitored in real time, and then identify possible structural abnormal parts;

[0057] The constructed abnormality location model is applied to the real-time monitoring data, and the abnormality probability of each component is calculated based on the real-time monitored temperature and vibration information. Furthermore, with the accumulation of more actual monitoring data, the abnormality location model is continuously optimized and iterated, so that the abnormality location model can locate the structural defects of the pneumatic grinder more accurately and quickly, and realize the effective prediction and location of structural defects in the operation of the pneumatic grinder.

[0058] S4: When the abnormal probability of the abnormal component reaches a preset requirement, a current structural image of the pneumatic grinder is collected, and feature recognition and comparison is performed using the current structural image and the initial image to obtain a feature comparison result and an abnormal component with an abnormal feature comparison result;

[0059] Further, the current structural image of the pneumatic grinder is collected, and feature recognition and comparison is performed using the current structural image and the initial image to obtain feature comparison results and abnormal parts with abnormal feature comparison results. The method of the present application includes:

[0060] S41: performing component labeling based on the initial image;

[0061] S42: aligning the current structure image with the initial image according to the component annotation to determine a logo comparison relationship;

[0062] S43: using a pre-trained convolution kernel model to compare the identification position features of the current structure image and the initial image to obtain a deviation recognition result;

[0063] S44: performing deviation positioning and deviation data extraction based on the deviation identification result to determine abnormal parts with abnormalities.

[0064] On the whole, when it is confirmed that the abnormality probability reaches the abnormality probability threshold corresponding to the preset requirements, the current structural image of the pneumatic grinder is immediately collected and compared with the initial structural image before the pneumatic grinder is put into formal use; using image recognition and feature extraction technology, feature recognition and comparison are performed on the current structural image and the initial structural image to find out whether there are differences or abnormal features between the current structural image and the initial structural image, so as to further confirm and identify abnormal parts that may have structural defects.

[0065] The current structural image of the pneumatic grinder is collected, and feature recognition and comparison is performed using the current structural image and the initial image to obtain feature comparison results and abnormal parts with abnormal feature comparison results. Specifically, detailed parts annotation is performed based on the initial image, including information such as key parts names, numbers, positions and shapes, and a parts database is established; the current structural image of the pneumatic grinder is obtained, and according to the parts annotation information, the current structural image is geometrically aligned with the initial image to ensure that the same parts in the two images are in corresponding positions, so as to facilitate subsequent feature comparison;

[0066] Using a pre-trained convolutional neural network model (CNN), iterative training is performed on the structural image of the pneumatic grinder, so that the model can effectively identify and extract features in the structural image of the pneumatic grinder; the aligned current structural image and the initial image are input into the trained convolutional neural network model, and the image features of the same parts of the current structural image and the initial image are compared using the pre-trained convolution kernel model, and the differences in texture, shape, boundary, etc. between the current structural image and the initial image of the same parts are compared, and further, the feature differences (including texture feature differences, shape feature differences, and boundary feature differences) between the current structural image and the initial image at the identification positions of each part are analyzed to obtain the deviation recognition result, and the identified feature differences are accurately located to clarify the specific position and range of the parts that have undergone abnormal changes in the current structural image;

[0067] Key deviation data are extracted from the deviation identification results, such as quantitative information such as the degree of deformation (shape feature difference), wear area (texture feature difference), and fracture position (boundary feature difference) of components. Based on the deviation data and the preset threshold, it is determined which components are indeed abnormal and marked as abnormal components, thus achieving an accurate comparison between the current structure and the initial structure of the pneumatic grinder, and effectively identifying and locating potential structural defects or abnormal components.

[0068] S5: Based on the abnormal parts and abnormal probabilities, feature comparison results, and abnormal parts with abnormal feature comparison results, abnormal fusion is performed to obtain structural defect positioning information for feedback.

[0069] Furthermore, according to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, abnormal fusion is performed to obtain structural defect location information. The method of the present application includes:

[0070] S51: According to the abnormal parts and abnormal probabilities, feature comparison results, and abnormal parts with abnormal feature comparison results, data integration is performed according to the abnormal parts to obtain abnormal probabilities and feature comparison results of each abnormal part;

[0071] S52: inputting the abnormal probability of the abnormal component and the feature comparison result into a correlation analysis channel to obtain abnormal support;

[0072] S53: Arrange the abnormal support degrees in positive order, filter the abnormal support degrees according to preset filtering conditions, locate the deviation in the filtered feature comparison results as the structural defect location information, the preset filtering conditions correspond to the size of the abnormal probability, wherein the total abnormal support degrees obtained by filtering are not less than half of the proportional relationship of the abnormal probability.

[0073] On the whole, the abnormal parts, abnormal probability, feature comparison results and abnormal features confirmed by comparison are comprehensively analyzed and integrated to obtain more accurate structural defect location information; the final structural defect information is fed back to maintenance personnel for timely repair and replacement to ensure the normal operation and safety of the pneumatic grinder, realizing real-time monitoring, positioning and feedback of structural defects of the pneumatic grinder, and improving equipment maintenance efficiency and safety.

[0074] According to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, abnormal fusion is performed to obtain structural defect location information. Specifically, according to the abnormal parts and abnormal probabilities, feature comparison results and information on parts with abnormal features in the feature comparison results, data of all abnormal parts are integrated, the abnormal probability and feature comparison results of each abnormal part are calculated, and a detailed abnormal parts database is established;

[0075] The abnormal probability and feature comparison results of the integrated abnormal parts are input into the correlation analysis channel. The correlation analysis channel can be an expert evaluation channel, or an AI evaluation channel, machine learning, etc., to determine the difference changes in the current image recognition, whether the difference changes in image recognition will cause the current abnormal probability. Furthermore, a large support indicates that a certain feature difference change has a close relationship with the abnormal probability. For example, the support corresponding to the boundary feature is 0.7. Correspondingly, if the boundary feature difference becomes larger, the abnormal probability will increase with the boundary feature difference. By comparing the support of comparison features such as texture features, shape features, and boundary features, the comparison feature with the largest support is determined, and the comparison feature with the largest support is used as the defect location.

[0076] In addition, it may also be the abnormal probability caused by the feature change of multiple comparison features. The main features in the feature change of multiple comparison features are screened out, and the screening is carried out in the order of support. For example, the abnormal probability caused by the feature change of multiple comparison features is 60%. In the order of support from large to small, the support corresponding to the first ranking is 20, the support corresponding to the second ranking is 18, the support corresponding to the third ranking is 11, the support corresponding to the fourth ranking is 6, and the support corresponding to the remaining multiple items is 5. If the abnormal probability corresponding to the cumulative support corresponding to the screening number is not less than 50% of the abnormal probability caused by the feature change of multiple comparison features, then the support corresponding to the first ranking is 20, and the support corresponding to the second ranking is 18, that is, the comparison feature with the first support ranking and the comparison feature with the second support ranking need to be used as the structural defect positioning information;

[0077] The total support of the abnormalities obtained by screening is not less than half of the proportional relationship of the abnormal probability. Half of the abnormal probability is the minimum requirement. It can be adaptively adjusted for high control accuracy. For example, the preset screening requirement is not less than 60%. Optionally, if the abnormal probability is 80%, the sum of the support of the parameters obtained by screening is not less than 48%. According to the 48% proportion requirement, the corresponding number of parameters are screened out. Specifically, the support corresponding to the first ranking is 20, the support corresponding to the second ranking is 18, and the support corresponding to the third ranking is 11, that is, the first-ranked comparison feature, the second-ranked comparison feature, and the third-ranked comparison feature need to be used as the structural defect positioning information; the accumulated support corresponding to the first-ranked comparison feature, the second-ranked comparison feature, and the third-ranked comparison feature is 49, among which 49 / 100×100%>60%×80%. The closer the abnormal probability corresponding to the accumulated support corresponding to the screening number is to the abnormal probability caused by the feature changes of multiple comparison features, the higher the defect positioning accuracy is.

[0078] Through the correlation analysis channel, the abnormal support of each abnormal component is calculated, and the abnormal support reflects the degree of consistency between the possibility of the component abnormality and its feature comparison result; the calculated abnormal support is arranged in order from high to low, and then the abnormal support is screened according to the preset screening conditions, and the screening conditions include the threshold of abnormal probability, the abnormal support is greater than a certain value, or the total abnormal support must reach a certain proportion of the abnormal probability. Correspondingly, the smaller the screening number, the faster the corresponding recognition speed. Limitedly, the abnormal probability corresponding to the total abnormal support is not less than half of the abnormal probability caused by the feature changes of multiple comparison features;

[0079] Among the screened parts with higher abnormal support, the specific structural defect position is determined according to the deviation positioning information of the feature comparison results; the deviation positioning in the screened feature comparison results is used as the structural defect positioning information, so as to accurately identify the structural defect parts in the pneumatic grinder that need to be focused on and repaired, and combine the abnormal probability, feature comparison results and abnormal support, so as to accurately locate the structural defects of the pneumatic grinder.

[0080] Furthermore, the present application method also includes:

[0081] According to the occurrence time sequence relationship between the detection result of the initial structure and the operation monitoring data, a time sequence chain is constructed, the state of each component is used as the node state information, the operation monitoring data is used as the action information of the node state, the historical case data is used to fit the state influence relationship, the node reward value is calculated, and a Markov chain prediction model is constructed;

[0082] The Markov chain prediction model is used to predict the abnormal probability based on the current operation monitoring data, and the abnormal probability of the parts is obtained for feedback.

[0083] According to the detection results of the initial structure, the basic state of each component is determined, and combined with the operation monitoring data, a component state timing chain is constructed; in the component state timing chain, each component is regarded as a node, its state is regarded as the node state information, and the operation monitoring data such as temperature and vibration are regarded as the action information of the node state changing over time; using historical case data, the transition rules between the states of each component are analyzed, and according to the state changes of the components at different times, a state transfer matrix is ​​constructed, which reflects the probability of the component state changing from one moment to the next.

[0084] Conduct in-depth mining of historical data to fit the causal relationship between component states, that is, how a change in the state of a component affects the states of other components, thereby providing a basis for the Markov chain prediction model; in the Markov decision process, define the node reward value, which can be a score representing the normal operation, slight abnormality or serious failure of the component, and calculate the reward value of each node (each component) in different states according to the state transfer matrix and the state influence relationship; construct a Markov chain prediction model based on the state transfer matrix, node reward value and the state influence relationship between components, and the Markov chain prediction model can predict the possibility of component state transfer and the potential structural defects caused by it in the future;

[0085] Input the current operation monitoring data into the constructed Markov chain prediction model, and predict the probability of abnormality of each component in the future period according to the current state and state transition law; evaluate the health status of the component according to the abnormal probability predicted by the model. Using the Markov chain prediction model, the current monitoring and analysis data are used to make a later prediction based on the time series relationship, and feedback is given to nodes with a higher probability, so that maintenance and structural adjustments can be carried out in advance based on probability.

[0086] In summary, the beneficial effects of the embodiments of the present application are:

[0087] 1. Real-time monitoring of key parameters such as temperature, vibration, and pressure of the grinding machine during operation can improve the ability to discover problems and improve the accuracy and timeliness of structural defect identification.

[0088] 2. Discover and locate potential structural defects through the abnormal positioning model, discover subtle structural changes, and improve positioning accuracy.

[0089] 3. Due to the use of abnormal parts and abnormal probability, feature comparison results and abnormal parts with abnormal feature comparison results, data integration is performed according to abnormal parts to obtain the abnormal probability and feature comparison results of each abnormal part; the abnormal probability and feature comparison results of the abnormal parts are input into the correlation analysis channel to obtain the abnormal support; the abnormal support is arranged in positive order, and the abnormal support is screened according to the preset screening conditions, and the deviation in the screened feature comparison results is located as the structural defect positioning information. The preset screening conditions correspond to the size of the abnormal probability, and the total abnormal support obtained by screening is not less than half of the proportional relationship of the abnormal probability. Accurately identify the structural defect parts that need to be focused on and repaired in the pneumatic grinder, combine the abnormal probability, feature comparison results and abnormal support, so as to accurately locate the structural defects of the pneumatic grinder.

[0090] Embodiment 2

[0091] Based on the same inventive concept as the method for locating a structural defect of a pneumatic grinder in the aforementioned embodiment, Figure 3 As shown, an embodiment of the present application provides a structural defect positioning system for a pneumatic grinding machine, wherein the system comprises:

[0092] The adaptive matching module M100 is used to adaptively match the product structure of the pneumatic grinder with the scene application characteristics through a preset detection feature set to obtain the detection result of the initial structure;

[0093] The operation data monitoring module M200 is used to archive the test results and initial images of the pneumatic grinder with qualified test results, and monitor the operation data through monitoring equipment. The operation monitoring data includes temperature information, vibration information and monitoring position, and the loss value of the operation monitoring data is evaluated, wherein the monitoring position has a positional correspondence with the component;

[0094] The abnormality analysis module M300 is used to perform abnormality analysis based on the operation monitoring data and its loss value through an abnormality location model to obtain abnormal parts and abnormality probabilities;

[0095] The feature recognition and comparison module M400 is used to collect the current structure image of the pneumatic grinder when the abnormal probability of the abnormal component reaches the preset requirement, and use the current structure image and the initial image to perform feature recognition and comparison to obtain the feature comparison result and the abnormal component with abnormal feature comparison result;

[0096] The abnormal fusion module M500 is used to perform abnormal fusion according to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, and obtain structural defect positioning information for feedback.

[0097] Furthermore, the adaptive matching module M100 is used to perform the following method:

[0098] According to the structure of the pneumatic grinder, the connection relationship and transmission relationship of the pneumatic grinder parts are decomposed, and the structural relationship of the pneumatic grinder is constructed;

[0099] Acquire an initial image of the pneumatic grinder, wherein the initial image is an image of the internal structure of the pneumatic grinder before use;

[0100] Perform scenario application feature analysis according to the application scenario of the pneumatic grinder to obtain scenario structure fault features, where the scenario structure fault features are distinguishing fault features obtained by performing feature unidirectional comparison of multiple scenario application cases, and obtain faulty components corresponding to the distinguishing fault features;

[0101] Based on the faulty parts corresponding to the distinguishing fault features, detection enhancement features are set, added to the preset detection feature set to perform structural positioning detection on the initial image, and the detection results are obtained. If an abnormality exists, the positioning is marked and a reminder message is sent.

[0102] Furthermore, the operation data monitoring module M200 is used to perform the following method:

[0103] Based on the operation monitoring data, data of each monitoring position is integrated according to the monitoring time to obtain a temperature monitoring data set and a vibration monitoring data set;

[0104] The temperature monitoring data set and the vibration monitoring data set are respectively calculated for the time series loss value according to the connection relationship of the monitoring operation time. The loss value expression is: Obtain temperature loss value and vibration loss value; where y i (t ) is the monitoring data of the t-th monitoring location, y i (t+1) is the monitoring data of the continuous time nodes at the i-th monitoring position after time t, and i and n are positive integers.

[0105] Furthermore, the abnormality analysis module M300 is used to perform the following method:

[0106] Obtain a structural defect case set of a pneumatic grinder, integrate data of abnormal parts according to the structural defect case set, and construct a temperature data set and a vibration data set of each abnormal part;

[0107] Using the temperature data set and the vibration data set of each abnormal component to fit the abnormal temperature and abnormal vibration relationship, a functional relationship is obtained;

[0108] The abnormal defect level of each abnormal component is taken as the top event, and the temperature data and vibration data are taken as abnormal factors. Based on the functional relationship, the abnormal location model is constructed.

[0109] Furthermore, the feature recognition and comparison module M400 is used to perform the following method:

[0110] Performing component labeling based on the initial image;

[0111] According to the component annotations, the current structure image is aligned with the initial image to determine a logo comparison relationship;

[0112] Using a pre-trained convolution kernel model, the current structure image and the initial image are compared for identification position features to obtain a deviation recognition result;

[0113] Deviation positioning and deviation data extraction are performed based on the deviation identification result to determine abnormal parts with abnormalities.

[0114] Furthermore, the abnormal fusion module M500 is used to perform the following method:

[0115] According to the abnormal parts and abnormal probabilities, feature comparison results, and abnormal parts with abnormal feature comparison results, data integration is performed according to the abnormal parts to obtain the abnormal probability and feature comparison results of each abnormal part;

[0116] Inputting the abnormal probability of the abnormal component and the feature comparison result into a correlation analysis channel to obtain abnormal support;

[0117] The abnormal support is arranged in positive order, and the abnormal support is screened according to preset screening conditions. The deviation in the screened feature comparison result is located as the structural defect location information. The preset screening condition corresponds to the size of the abnormal probability, wherein the total abnormal support obtained by screening is not less than half of the proportional relationship of the abnormal probability.

[0118] Furthermore, the structural defect locating system of the pneumatic grinder is used to perform the following method:

[0119] According to the occurrence time sequence relationship between the detection result of the initial structure and the operation monitoring data, a time sequence chain is constructed, the state of each component is used as the node state information, the operation monitoring data is used as the action information of the node state, the historical case data is used to fit the state influence relationship, the node reward value is calculated, and a Markov chain prediction model is constructed;

[0120] The Markov chain prediction model is used to predict the abnormal probability based on the current operation monitoring data, and the abnormal probability of the parts is obtained for feedback.

[0121] In summary, any step can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor, and no unnecessary restrictions are made here.

[0122] Furthermore, the above-mentioned technical scheme only reflects the preferred technical scheme of the technical scheme of the embodiment of the present application. Some changes that may be made to some parts thereof by technicians in this technical field all reflect the novel principles of the embodiment of the present application. Obviously, technicians in this field can make various changes and modifications to the present application without departing from the scope of the present application.

Claims

1. A method for locating structural defects of a pneumatic grinder, characterized in that: The method comprises: The product structure of the pneumatic grinder is adaptively matched with the scene application features through the preset detection feature set to obtain the detection result of the initial structure; For the pneumatic grinders with qualified test results, the test results and initial images are archived, and the operation data is monitored through monitoring equipment. The operation monitoring data includes temperature information, vibration information and monitoring position, and the loss value of the operation monitoring data is evaluated, wherein the monitoring position has a positional correspondence with the component; Based on the operation monitoring data and its loss value, abnormal analysis is performed through an abnormal location model to obtain abnormal parts and abnormal probability; When the abnormal probability of the abnormal component reaches a preset requirement, the current structure image of the pneumatic grinder is collected, and feature recognition and comparison is performed using the current structure image and the initial image to obtain feature comparison results and abnormal components with abnormal feature comparison results; According to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, abnormal fusion is performed to obtain structural defect positioning information for feedback.

2. The method according to claim 1, characterized in that The method of adaptively matching the product structure of the pneumatic grinder with the scene application features through the preset detection feature set to obtain the detection result of the initial structure includes: According to the structure of the pneumatic grinder, the connection relationship and transmission relationship of the pneumatic grinder parts are decomposed, and the structural relationship of the pneumatic grinder is constructed; Acquire an initial image of the pneumatic grinder, wherein the initial image is an image of the internal structure of the pneumatic grinder before use; Perform scenario application feature analysis according to the application scenario of the pneumatic grinder to obtain scenario structure fault features, where the scenario structure fault features are distinguishing fault features obtained by performing feature unidirectional comparison of multiple scenario application cases, and obtain faulty components corresponding to the distinguishing fault features; Based on the faulty components corresponding to the distinguishing fault features, detection enhancement features are set, added to the preset detection feature set to perform structural positioning detection on the initial image, and the detection results are obtained. If an abnormality is found, the positioning is marked and a reminder message is sent.

3. The method according to claim 1, characterized in that The operation monitoring data includes temperature information, vibration information and monitoring location, and the loss value evaluation of the operation monitoring data includes: Based on the operation monitoring data, data of each monitoring position is integrated according to the monitoring time to obtain a temperature monitoring data set and a vibration monitoring data set; The temperature monitoring data set and the vibration monitoring data set are respectively calculated for the time series loss value according to the connection relationship of the monitoring operation time. The loss value expression is: Obtain temperature loss value and vibration loss value; where y i (t ) is the monitoring data of the t-th monitoring location, y i (t+1) is the monitoring data of the continuous time nodes at the i-th monitoring position after time t, and i and n are positive integers.

4. The method according to claim 3, characterized in that The abnormality analysis is performed by using the abnormality location model to obtain abnormal parts and abnormality probabilities, including: Obtain a structural defect case set of a pneumatic grinder, integrate data of abnormal parts according to the structural defect case set, and construct a temperature data set and a vibration data set of each abnormal part; Using the temperature data set and the vibration data set of each abnormal component to fit the abnormal temperature and abnormal vibration relationship, a functional relationship is obtained; The abnormal defect level of each abnormal component is taken as the top event, and the temperature data and vibration data are taken as abnormal factors. Based on the functional relationship, the abnormal location model is constructed.

5. The method according to claim 1, characterized in that The method of collecting the current structural image of the pneumatic grinder, using the current structural image and the initial image to perform feature recognition and comparison, and obtaining feature comparison results and abnormal parts with abnormal feature comparison results, includes: Performing component labeling based on the initial image; According to the component annotations, the current structure image is aligned with the initial image to determine a logo comparison relationship; Using a pre-trained convolution kernel model, the current structure image and the initial image are compared for identification position features to obtain a deviation recognition result; Deviation positioning and deviation data extraction are performed based on the deviation identification result to determine abnormal parts with abnormalities.

6. The method according to claim 5, characterized in that According to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, abnormal fusion is performed to obtain structural defect location information, including: According to the abnormal parts and abnormal probabilities, feature comparison results, and abnormal parts with abnormal feature comparison results, data integration is performed according to the abnormal parts to obtain the abnormal probability and feature comparison results of each abnormal part; Inputting the abnormal probability of the abnormal component and the feature comparison result into a correlation analysis channel to obtain abnormal support; The abnormal support is arranged in positive order, and the abnormal support is screened according to preset screening conditions. The deviation in the screened feature comparison result is located as the structural defect location information. The preset screening condition corresponds to the size of the abnormal probability, wherein the total abnormal support obtained by screening is not less than half of the proportional relationship of the abnormal probability.

7. The method according to claim 1, characterized in that The method further comprises: According to the occurrence time sequence relationship between the detection result of the initial structure and the operation monitoring data, a time sequence chain is constructed, the state of each component is used as the node state information, the operation monitoring data is used as the action information of the node state, the historical case data is used to fit the state influence relationship, the node reward value is calculated, and a Markov chain prediction model is constructed; The Markov chain prediction model is used to predict the abnormal probability based on the current operation monitoring data, and the abnormal probability of the parts is obtained for feedback.

8. A structural defect positioning system for a pneumatic grinder, characterized in that: A method for locating structural defects of a pneumatic grinder according to any one of claims 1 to 7, the system comprising: An adaptive matching module is used to adaptively match the product structure of the pneumatic grinder with the scene application features through a preset detection feature set to obtain a detection result of the initial structure; An operation data monitoring module is used to archive the test results and initial images of the pneumatic grinders with qualified test results, and to monitor the operation data through monitoring equipment. The operation monitoring data includes temperature information, vibration information and monitoring position, and to evaluate the loss value of the operation monitoring data, wherein the monitoring position has a positional correspondence with the component; An abnormality analysis module, used to perform abnormality analysis based on the operation monitoring data and its loss value through an abnormality location model to obtain abnormal parts and abnormality probabilities; A feature recognition and comparison module is used to collect the current structure image of the pneumatic grinder when the abnormal probability of the abnormal component reaches a preset requirement, and use the current structure image and the initial image to perform feature recognition and comparison to obtain a feature comparison result and an abnormal component with an abnormal feature comparison result; The abnormal fusion module is used to perform abnormal fusion according to the abnormal parts and abnormal probabilities, feature comparison results and abnormal parts with abnormal feature comparison results, and obtain structural defect positioning information for feedback.

Images