Bearing surface quality detection and evaluation system based on visual cognition model

Abstract

The invention belongs to the technical field of bearing detection, and particularly relates to a bearing surface quality detection and evaluation system based on a visual cognition model, which comprises a bearing scanning image acquisition unit, a surface processing extraction unit, a visual cognition model unit, a quantitative evaluation output unit, a bearing sorting unit and a display alarm end, the bearing surface is subjected to image acquisition processing through the bearing scanning image acquisition unit and the surface processing extraction unit, and defect identification and quantitative evaluation are performed in combination with a visual cognition model and a machine learning algorithm, so that the bearing surface quality is quickly detected and evaluated, the detection efficiency is remarkably improved, the labor cost is reduced, and the detection efficiency is improved. The workload of bearing production management personnel is obviously reduced, and the difficulty of bearing production detection management is effectively reduced by carrying out gradual progressive reasonable analysis and accurate early warning on the stability and accuracy of the bearing surface quality detection evaluation process and the bearing production performance and management condition.

Description

Technical Field

[0001] The present invention relates to the technical field of bearing detection, and in particular to a bearing surface quality detection and evaluation system based on a visual cognitive model. Background Art

[0002] Bearings are core components in mechanical equipment, and their surface quality has an important impact on the operating performance and life of the equipment. Traditional bearing surface quality inspection mainly relies on manual inspection, which is not only inefficient, but also easily interfered by human factors, resulting in low accuracy of the inspection results. With the development of computer vision and artificial intelligence technology, automated inspection systems based on machine vision have gradually become mainstream.

[0003] A Chinese invention patent with publication number CN112308832A discloses a bearing quality inspection method based on machine vision. It first uses a two-dimensional inspection method to detect whether there are defects under specific pixels, and then uses three-dimensional inspection to detect local detail features. Three-dimensional inspection and reconstruction are performed on bearings with two-dimensional defect features to obtain the bearing surface defect category. It is not necessary to use three-dimensional inspection for all bearings, which improves inspection efficiency and saves inspection costs.

[0004] However, in actual application, the above-mentioned technical solution cannot realize the automated and comprehensive control of the bearing surface quality detection and evaluation process, and it is difficult to conduct a step-by-step and reasonable analysis and accurate early warning of the stability and accuracy of the bearing surface quality detection and evaluation process and the bearing production performance and management status. The intelligent level is low and the difficulty of bearing production management cannot be effectively reduced.

[0005] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0006] The purpose of the present invention is to provide a bearing surface quality detection and evaluation system based on a visual cognitive model, which solves the problems that the prior art cannot achieve automated and comprehensive control over the bearing surface quality detection and evaluation process, and it is difficult to conduct a step-by-step and reasonable analysis and accurate early warning of the stability and accuracy of the bearing surface quality detection and evaluation process as well as the bearing production performance and management status, and the intelligence level is low, and it is unable to effectively reduce the difficulty of bearing production management.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The bearing surface quality detection and evaluation system based on the visual cognitive model includes a bearing scanning and image acquisition unit, a surface treatment extraction unit, a visual cognitive model unit, a quantitative evaluation output unit, a bearing sorting unit and a display alarm terminal; the bearing scanning and acquisition unit is equipped with a high-resolution industrial camera, a multi-angle annular light source and a rotating stage to collect visible light images, high dynamic range images and three-dimensional morphology data of the bearing surface, eliminates reflection interference through an adaptive illumination compensation algorithm, generates a clear surface image with multi-scale fusion, and sends the bearing surface image to the surface treatment extraction unit;

[0009] The surface processing extraction unit performs preprocessing operations of denoising and contrast enhancement on the received bearing surface image, extracts feature information of the bearing surface using an image processing algorithm, and sends the bearing surface feature extraction information to the visual cognition model unit; the visual cognition model unit uses a convolutional neural network to construct a visual cognition model, the visual cognition model analyzes the bearing surface feature extraction information, identifies defects on the bearing surface, and sends the identification results to the quantitative evaluation output unit;

[0010] The quantitative evaluation output unit performs a quantitative evaluation of the defects according to the recognition results of the visual cognitive model, outputs the type, location and size of the defects, generates a bearing surface quality inspection report, and sends the bearing surface quality inspection report to the display alarm terminal and the bearing sorting unit; the bearing sorting unit sorts the bearings that do not meet the requirements into the scrap area, and sorts the bearings that meet the requirements into the qualified area.

[0011] Furthermore, the display alarm terminal is communicatively connected to the stability analysis output module. The stability analysis output module is used to set a detection period, analyze the stability performance of the bearing surface quality detection and evaluation during the detection period, generate a stability qualified signal or a stability abnormality signal through the analysis, and send the stability qualified signal or the stability abnormality signal to the display alarm terminal. When the display alarm terminal receives the stability abnormality signal, it issues a warning.

[0012] Furthermore, the specific analysis process of the stability analysis output module includes:

[0013] When performing surface quality inspection and evaluation of the corresponding bearing, the moment when the bearing scanning acquisition unit performs image acquisition is marked as moment one, the moment when the bearing sorting unit completes bearing sorting is marked as moment two, and the interval between moment one and moment two is marked as the selection duration;

[0014] When the selection time exceeds the preset selection time threshold, the effectiveness evaluation symbol ZP-1 is assigned, the number of times the effectiveness evaluation symbol ZP-1 is assigned during the detection period is obtained and the ratio is calculated by calculating the ratio with the total number of bearings subjected to surface quality inspection and evaluation during the detection period to obtain the effectiveness evaluation warning value, and the average of all selection times in the detection period is calculated to obtain the timing decision value, and the effectiveness evaluation warning value and the timing decision value are numerically compared with the preset effectiveness evaluation warning threshold and the preset timing decision threshold respectively. If the effectiveness evaluation warning value or the timing decision value exceeds the corresponding preset threshold, a stability abnormality signal is generated.

[0015] Furthermore, if the effectiveness evaluation warning value and the timing decision value do not exceed the corresponding preset thresholds, the total duration of the bearing surface quality detection and evaluation during the detection period is collected and marked as the total detection time value, and the number of failures that occurred during the bearing surface quality detection and evaluation process during the detection period is marked as the detection obstacle value, and the detection obstacle value is calculated by ratio with the total detection time value to obtain the obstacle judgment value;

[0016] The stability input value is obtained by weighted summing up the effectiveness evaluation warning value, timing decision value and obstacle judgment value, and the stability input value is numerically compared with the preset stability input threshold. If the stability input value exceeds the preset stability input threshold, a stability abnormality signal is generated; if the stability input value does not exceed the preset stability input threshold, a stability qualified signal is generated.

[0017] Furthermore, the stability analysis output module is communicatively connected to the accuracy test module, and the stability analysis output module sends a stability qualified signal to the accuracy test module. When the accuracy test module receives the stability qualified signal, it analyzes the accuracy status of the bearing surface quality detection and evaluation, and generates an accuracy qualified signal or an accuracy abnormality signal through analysis, and sends the accuracy qualified signal or the accuracy abnormality signal to the display alarm terminal, which issues a warning when receiving the accuracy abnormality signal.

[0018] Furthermore, the specific analysis process of the accuracy test module includes:

[0019] Conduct several surface quality inspections and assessments on bearings known to have multiple defects, collect the recognition accuracy rate during each surface quality inspection and assessment, perform variance calculation on all recognition accuracy rates to obtain the result consistency analysis value, perform numerical comparison on the result consistency analysis value and the preset result consistency analysis threshold, and if the result consistency analysis value exceeds the preset result consistency analysis threshold, assign the recognition judgment symbol XP-1 to the test result of the corresponding bearing;

[0020] If the result consistency analysis value does not exceed the preset result consistency analysis threshold, the proportion of the number of detections whose recognition accuracy does not exceed the preset recognition accuracy threshold is marked as the recognition quasi-outlier value, and all recognition accuracies are averaged to obtain the recognition accuracy judgment value, and the recognition quasi-outlier value and the recognition accuracy judgment value are numerically compared with the preset recognition quasi-outlier threshold and the preset recognition accuracy judgment threshold, respectively. If the recognition quasi-outlier value or the recognition accuracy judgment value exceeds the corresponding preset threshold, the recognition judgment symbol XP-1 is assigned to the test result of the corresponding bearing;

[0021] After completing the test of several bearings with multiple defects, if the number of times the identification and judgment symbol XP-1 is assigned is zero, a qualified accuracy signal is generated; if the number of times the identification and judgment symbol XP-1 is assigned is not zero, an abnormal accuracy signal is generated.

[0022] Furthermore, the accuracy test module is communicatively connected to the bearing production line optimization judgment module. The accuracy test module sends a qualified accuracy signal to the bearing production line optimization judgment module. When the bearing production line optimization judgment module receives the qualified accuracy signal, it analyzes the production status of the bearing production line during the detection period, and generates a bearing production qualified signal or a bearing production abnormality signal through analysis. The bearing production qualified signal or the bearing production abnormality signal is sent to the display alarm terminal. When the display alarm terminal receives the bearing production abnormality signal, it issues a warning.

[0023] Furthermore, the specific analysis process of the bearing production line optimization judgment module includes:

[0024] Collect the bearing surface quality inspection reports of all bearings entering the scrap zone during the inspection period, obtain all defect information on the corresponding bearings based on the bearing surface quality inspection reports, calculate the ratio of the number of bearings involved in the corresponding type of defects to the total number of bearings inspected during the inspection period to obtain the bearing defect matching value, compare the bearing defect matching value with the corresponding preset bearing defect matching threshold, and if the bearing defect matching value exceeds the preset bearing defect matching threshold, assign the prone judgment symbol WP-1 to the corresponding type of defect;

[0025] If the defect type corresponding to the prone judgment symbol WP-1 exists during the detection period, a bearing production abnormality signal is generated; if the defect type corresponding to the prone judgment symbol WP-1 does not exist during the detection period, the number of bearings entering the scrap area during the detection period and the total number of bearings detected during the detection period are ratio-calculated to obtain the bearing scrap analysis value, and the bearing scrap analysis value is numerically compared with the preset bearing scrap analysis threshold. If the bearing scrap analysis value exceeds the preset bearing scrap analysis threshold, a bearing production abnormality signal is generated; if the bearing scrap analysis value does not exceed the preset bearing scrap analysis threshold, a bearing production qualified signal is generated.

[0026] Furthermore, the bearing production line optimization judgment module is communicatively connected to the bearing production management judgment module. The bearing production management module sends a bearing production qualified signal to the bearing production management judgment module. When the bearing production management judgment module receives the bearing production qualified signal, it analyzes the management performance of bearing production during the detection period, and generates a management qualified signal or a management abnormality signal through analysis, and sends the management qualified signal or the management abnormality signal to the display alarm terminal. When the display alarm terminal receives the management abnormality signal, it issues a corresponding warning.

[0027] Furthermore, the specific analysis process of the bearing production management judgment module is as follows:

[0028] Monitor the bearing production line, collect the average interval time for inspection and maintenance of the corresponding production equipment in the bearing production line during the detection period, and compare it with the corresponding preset average interval time threshold, mark the production equipment whose average interval time exceeds the preset average interval time threshold as poorly managed equipment, obtain the proportion of equipment marked as poorly managed equipment during the detection period and mark it as poorly managed value;

[0029] The ratio of the average interval time for inspection and maintenance of the corresponding production equipment in the bearing production line to the corresponding preset average interval time threshold is marked as the inspection and maintenance interval value, and the inspection and maintenance interval values ​​of all production equipment are averaged to obtain the production line maintenance value, and the frequency of operation errors of the production line operators during the detection period is collected. The management judgment coefficient is obtained by weighted summation of the management non-good value, the production line maintenance value and the operation error frequency, and the management judgment coefficient is numerically compared with the preset management judgment coefficient threshold. If the management judgment coefficient exceeds the preset management judgment coefficient threshold, a management abnormality signal is generated; if the management judgment coefficient does not exceed the preset management judgment coefficient threshold, a management qualified signal is generated.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. In the present invention, by performing image acquisition and processing on the bearing surface and combining visual cognitive models and machine learning algorithms to perform defect identification and quantitative evaluation, rapid detection and evaluation of the bearing surface quality can be achieved, which significantly improves the detection efficiency and reduces labor costs, significantly reduces the workload of bearing production management personnel, and analyzes the stability performance of the bearing surface quality detection and evaluation during the detection period through the stability analysis output module. When generating a stability qualified signal, the accuracy of the bearing surface quality detection and evaluation system is tested and analyzed, which is conducive to achieving accurate evaluation of the bearing surface quality;

[0032] 2. In the present invention, when a qualified accuracy signal is received, the production status of the bearing production line during the detection period is analyzed by the bearing production line optimization judgment module, and when a bearing production abnormality signal is generated, the subsequent supervision of the bearing production line is strengthened to ensure the quality of the produced bearings and reduce the production cost of the bearings. When a qualified bearing production signal is generated, the management performance of the bearing production during the detection period is analyzed by the bearing production management judgment module, and when a management abnormality signal is generated, the management and control of the bearing production process is strengthened to further ensure the smooth progress of the subsequent production process of the bearing and improve the bearing quality, with a high level of intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0034] Figure 1 This is a system block diagram of Embodiment 1 of the present invention;

[0035] Figure 2 This is a system block diagram of Embodiment 2 and Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Embodiment 1: Figure 1 As shown, the bearing surface quality detection and evaluation system based on the visual cognitive model proposed in the present invention includes a bearing scanning and imaging unit, a surface treatment extraction unit, a visual cognitive model unit, a quantitative evaluation output unit, a bearing sorting unit, a stability analysis output module, an accuracy test module and a display alarm terminal;

[0038] The bearing scanning acquisition unit is equipped with a high-resolution industrial camera, a multi-angle annular light source and a rotating stage to collect visible light images, high dynamic range images and three-dimensional morphology data of the bearing surface. It eliminates reflection interference through an adaptive illumination compensation algorithm, generates a clear surface image with multi-scale fusion, and sends the bearing surface image to the surface processing extraction unit.

[0039] The surface processing extraction unit performs preprocessing operations of denoising and contrast enhancement on the received bearing surface image, extracts feature information of the bearing surface, such as edge, texture, etc., using an image processing algorithm, and sends the bearing surface feature extraction information to the visual cognition model unit;

[0040] The visual cognition model unit uses a convolutional neural network (CNN) or other deep learning architecture to build a visual cognition model (the model is trained by a large amount of bearing surface image data so that it can accurately identify various types of surface defects). The visual cognition model analyzes the bearing surface feature extraction information, identifies the defects on the bearing surface, and sends the identification results to the quantitative evaluation output unit;

[0041] The quantitative evaluation output unit performs a quantitative evaluation of the defects based on the recognition results of the visual cognitive model, outputs detailed information such as the type, location and size of the defects, generates a bearing surface quality inspection report, and sends the bearing surface quality inspection report to the display alarm terminal and the bearing sorting unit; the bearing sorting unit sorts the bearings that do not meet the requirements into the scrap area, and sorts the bearings that meet the requirements into the qualified area, and realizes the automatic sorting of the bearings at the end of the bearing surface quality inspection and evaluation.

[0042] The present invention adopts high-resolution industrial cameras and advanced image processing algorithms to achieve high-precision detection of tiny defects on the bearing surface. It also combines visual cognitive models and machine learning algorithms to perform intelligent evaluation of detected defects and output detailed defect information, providing strong support for subsequent bearing quality control and improvement. It also realizes rapid detection of bearings through an automated system, greatly improving detection efficiency, reducing labor costs, and significantly reducing the workload of bearing production management personnel.

[0043] The stability analysis output module sets the detection period, analyzes the stability performance of the bearing surface quality detection and evaluation during the detection period, generates a stability qualified signal or a stability abnormality signal through analysis, and sends the stability qualified signal or the stability abnormality signal to the display alarm terminal. When the display alarm terminal receives the stability abnormality signal, it issues an early warning to remind the management personnel to strengthen the subsequent detection supervision, and optimize the detection and evaluation system in time to ensure the stability of the subsequent detection and evaluation process; the specific analysis process of the stability analysis output module is as follows:

[0044] When performing surface quality inspection and evaluation of the corresponding bearing, the moment when the bearing scanning and acquisition unit performs image acquisition is marked as moment one, the moment when the bearing sorting unit completes bearing sorting is marked as moment two, and the interval between moment one and moment two is marked as selection time; wherein, the larger the value of the selection time, the slower the operation efficiency of the surface quality inspection and evaluation of the corresponding bearing;

[0045] Compare the selection time with the preset selection time, assign the performance evaluation symbol ZP-1 when the selection time exceeds the preset selection time threshold, obtain the number of times the performance evaluation symbol ZP-1 is assigned during the detection period, and calculate the ratio of it to the total number of bearings subjected to surface quality inspection and evaluation during the detection period to obtain the performance evaluation warning value, and calculate the average of all selection times during the detection period to obtain the selection time decision value;

[0046] The effectiveness evaluation warning value and the timing decision value are numerically compared with the preset effectiveness evaluation warning threshold and the preset timing decision threshold. If the effectiveness evaluation warning value or the timing decision value exceeds the corresponding preset threshold, it indicates that the operational stability of the bearing surface quality inspection and evaluation operation is poor, and a stability abnormality signal is generated.

[0047] Furthermore, if the effectiveness evaluation warning value and the timing decision value do not exceed the corresponding preset thresholds, the total duration of the bearing surface quality detection and evaluation during the detection period is collected and marked as the total detection time value, and the number of failures that occurred during the bearing surface quality detection and evaluation process during the detection period is marked as the detection obstacle value, and the detection obstacle value is calculated by ratio with the total detection time value to obtain the obstacle judgment value;

[0048] The stability input error value is obtained by weighted summing the effectiveness evaluation warning value, the timing decision value and the obstacle judgment value; that is, by assigning preset weight coefficients with values ​​greater than zero to the effectiveness evaluation warning value, the timing decision value and the obstacle judgment value, respectively, the effectiveness evaluation warning value, the timing decision value and the obstacle judgment value are multiplied by the corresponding preset weight coefficients, and the three groups of product results are summed to obtain the stability input error value; and the larger the value of the stability input error value is, the worse the overall operational stability of the bearing surface quality detection and evaluation operation is;

[0049] The stability input deviation value is numerically compared with the preset stability input deviation threshold. If the stability input deviation value exceeds the preset stability input deviation threshold, it indicates that the operational stability of the bearing surface quality detection and evaluation operation is generally poor, and a stability abnormality signal is generated; if the stability input deviation value does not exceed the preset stability input deviation threshold, it indicates that the operational stability of the bearing surface quality detection and evaluation operation is generally good, and a stability qualified signal is generated.

[0050] Furthermore, the stability analysis output module is connected in communication with the accuracy test module. The stability analysis output module sends a stability qualified signal to the accuracy test module. When the accuracy test module receives the stability qualified signal, it analyzes the accuracy status of the bearing surface quality detection and evaluation, and generates an accuracy qualified signal or an accuracy abnormality signal through analysis.

[0051] And send the accuracy qualified signal or accuracy abnormal signal to the display alarm terminal. When the display alarm terminal receives the accuracy abnormal signal, it issues an early warning to remind the management personnel to optimize and improve the defect recognition detection model, improve the subsequent recognition accuracy, and facilitate the accurate evaluation of the bearing surface quality. The specific analysis process of the accuracy test module is as follows:

[0052] Conduct several surface quality inspections and assessments on bearings known to have multiple defects, collect the recognition accuracy rate of each surface quality inspection and assessment, perform variance calculation on all recognition accuracy rates to obtain the result consistency analysis value, and numerically compare the result consistency analysis value with the preset result consistency analysis threshold. If the result consistency analysis value exceeds the preset result consistency analysis threshold, indicating that the differences between the recognition results of the corresponding bearing are large, the recognition judgment symbol XP-1 is assigned to the test result of the corresponding bearing;

[0053] If the result consistency analysis value does not exceed the preset result consistency analysis threshold, the proportion of the number of detections whose recognition accuracy does not exceed the preset recognition accuracy threshold is marked as the recognition quasi-outlier value, and all recognition accuracies are averaged to obtain the recognition accuracy judgment value, and the recognition quasi-outlier value and the recognition accuracy judgment value are numerically compared with the preset recognition quasi-outlier threshold and the preset recognition accuracy judgment threshold, respectively. If the recognition quasi-outlier value or the recognition accuracy judgment value exceeds the corresponding preset threshold, indicating that the defect recognition accuracy for the corresponding bearing is poor, the recognition judgment symbol XP-1 is assigned to the test result of the corresponding bearing;

[0054] After completing the test of several bearings with multiple defects, if the number of times the identification judgment symbol XP-1 is assigned is zero, it indicates that the recognition performance for bearing surface defects is good, and a qualified accuracy signal is generated; if the number of times the identification judgment symbol XP-1 is assigned is not zero, it indicates that the recognition performance for bearing surface defects is poor, which is not conducive to ensuring accurate assessment of bearing surface quality, and an abnormal accuracy signal is generated.

[0055] Embodiment 2: Figure 2 As shown, the difference between this embodiment and the first embodiment is that the accuracy test module is communicatively connected to the bearing production line optimization judgment module, the accuracy test module sends an accuracy qualified signal to the bearing production line optimization judgment module, and the bearing production line optimization judgment module analyzes the production status of the bearing production line during the detection period when receiving the accuracy qualified signal;

[0056] Through analysis, a bearing production qualified signal or a bearing production abnormal signal is generated, and the bearing production qualified signal or the bearing production abnormal signal is sent to the display alarm terminal. When the display alarm terminal receives the bearing production abnormal signal, it issues an early warning to remind management personnel to strengthen subsequent supervision of the bearing production line, ensure the quality of the produced bearings, and reduce the production cost of the bearings; the specific analysis process of the bearing production line optimization judgment module is as follows:

[0057] Collect the bearing surface quality inspection reports of all bearings entering the scrap zone during the inspection period, obtain all defect information on the corresponding bearings based on the bearing surface quality inspection reports, calculate the ratio of the number of bearings involved in the corresponding type of defects to the total number of bearings inspected during the inspection period to obtain the bearing defect matching value, compare the bearing defect matching value with the corresponding preset bearing defect matching threshold, and if the bearing defect matching value exceeds the preset bearing defect matching threshold, assign the prone judgment symbol WP-1 to the corresponding type of defect;

[0058] If there is a defect type corresponding to the prone judgment symbol WP-1 during the detection period, it indicates that there is a frequently occurring bearing defect type in the production process of the bearing production line during the detection period, and it is necessary to timely analyze the cause of the corresponding type of defect and overcome it, then a bearing production abnormality signal is generated; if there is no defect type corresponding to the prone judgment symbol WP-1 during the detection period, then the number of bearings entering the scrap area during the detection period and the total number of bearings detected during the detection period are calculated by ratio to obtain the bearing scrap analysis value;

[0059] The bearing scrap analysis value is numerically compared with the preset bearing scrap analysis threshold. If the bearing scrap analysis value exceeds the preset bearing scrap analysis threshold, it indicates that the quality of the bearings produced during the detection period is poor, and a bearing production abnormality signal is generated; if the bearing scrap analysis value does not exceed the preset bearing scrap analysis threshold, it indicates that the quality of the bearings produced during the detection period is good, and a bearing production qualified signal is generated.

[0060] Embodiment 3: Figure 2 As shown, the difference between this embodiment and the first and second embodiments is that the bearing production line optimization judgment module is communicatively connected to the bearing production management judgment module, the bearing production management module sends the bearing production qualified signal to the bearing production management judgment module, and the bearing production management judgment module analyzes the management performance of bearing production during the detection period when receiving the bearing production qualified signal;

[0061] Through analysis, a qualified management signal or a management abnormality signal is generated, and the qualified management signal or the management abnormality signal is sent to the display alarm terminal. When the display alarm terminal receives the management abnormality signal, it issues a corresponding warning to remind management personnel to strengthen the control of the bearing production process, and subsequently promptly overhaul the production equipment and train the production line operators to further ensure the smooth progress of the subsequent production process of the bearing and improve the quality of the bearing. The specific analysis process of the bearing production management judgment module is as follows:

[0062] Monitor the bearing production line, collect the average interval time for inspection and maintenance of the corresponding production equipment in the bearing production line during the detection period, and compare it with the corresponding preset average interval time threshold, mark the production equipment whose average interval time exceeds the preset average interval time threshold as poorly managed equipment, obtain the proportion of equipment marked as poorly managed equipment during the detection period and mark it as poorly managed value;

[0063] The ratio of the average interval time for inspection and maintenance of the corresponding production equipment in the bearing production line to the corresponding preset average interval time threshold is marked as the inspection and maintenance interval value, and the inspection and maintenance interval values ​​of all production equipment are averaged to obtain the production line maintenance value, as well as the frequency of operation errors of the production line operators during the detection period (i.e., the number of errors recorded by the production line operators during the detection period);

[0064] The management judgment coefficient is obtained by weighted summing the management non-good value, the production line maintenance value and the operation error frequency; that is, the management non-good value, the production line maintenance value and the operation error frequency are respectively assigned preset weight coefficients with values ​​greater than zero, and the management non-good value, the production line maintenance value and the operation error frequency are respectively multiplied by the corresponding preset weight coefficients, and the three sets of product results are summed to obtain the management judgment coefficient; and the larger the value of the management judgment coefficient, the worse the overall management status of bearing production during the detection period is;

[0065] The management judgment coefficient is numerically compared with the preset management judgment coefficient threshold. If the management judgment coefficient exceeds the preset management judgment coefficient threshold, it indicates that the management status of bearing production during the detection period is generally poor, and a management abnormality signal is generated; if the management judgment coefficient does not exceed the preset management judgment coefficient threshold, it indicates that the management status of bearing production during the detection period is generally good, and a management qualification signal is generated.

[0066] The working principle of the present invention is as follows: when in use, by performing image acquisition and processing on the bearing surface and combining visual cognitive models and machine learning algorithms to perform defect identification and quantitative evaluation, rapid detection and evaluation of the bearing surface quality can be achieved, detection efficiency can be significantly improved, labor costs can be reduced, and the workload of bearing production management personnel can be significantly reduced. The stability performance of the bearing surface quality detection and evaluation during the detection period is analyzed through the stability analysis output module, and when a stability abnormality signal is generated, subsequent detection supervision is strengthened and the detection and evaluation system is optimized in a timely manner. When a stability qualified signal is generated, the accuracy of the bearing surface quality detection and evaluation system is tested and analyzed through the accuracy test module, and when an accuracy abnormality signal is generated, the defect recognition detection model is optimized and improved to improve the subsequent recognition accuracy, which is conducive to the accurate evaluation of the bearing surface quality, and has a high level of intelligence.

[0067] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. Bearing surface quality detection and evaluation system based on visual cognitive model, characterized by: It includes a bearing scanning and image acquisition unit, a surface treatment extraction unit, a visual cognition model unit, a quantitative evaluation output unit, a bearing sorting unit and a display alarm terminal; the bearing scanning and acquisition unit acquires visible light images, high dynamic range images and three-dimensional morphological data of the bearing surface, and generates a clear surface image with multi-scale fusion; The surface processing extraction unit performs preprocessing operations on the received bearing surface image and extracts characteristic information of the bearing surface. The visual cognition model unit uses a convolutional neural network to construct a visual cognition model, and the visual cognition model identifies defects on the bearing surface. The quantitative evaluation output unit performs quantitative evaluation on the defects, generates a bearing surface quality inspection report, and sends the bearing surface quality inspection report to the display alarm terminal and the bearing sorting unit; the bearing sorting unit sorts the bearings that do not meet the requirements into the scrap area, and sorts the bearings that meet the requirements into the qualified area.

2. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 1 is characterized in that: The display alarm terminal is communicatively connected to the stability analysis output module. The stability analysis output module is used to set a detection period, analyze the stability performance of the bearing surface quality detection and evaluation during the detection period, generate a stability qualified signal or a stability abnormal signal through analysis, and send the stability qualified signal or the stability abnormal signal to the display alarm terminal.

3. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 2 is characterized in that: The specific analysis process of the stability analysis output module includes: When conducting surface quality inspection and evaluation of the corresponding bearings, an effectiveness evaluation symbol ZP-1 is assigned when the selection time exceeds the preset selection time threshold, the number of times the effectiveness evaluation symbol ZP-1 is assigned during the inspection period is obtained and the ratio is calculated by calculating the effectiveness evaluation warning value with the total number of bearings subjected to surface quality inspection and evaluation during the inspection period, and the average of all selection times during the inspection period is calculated to obtain the timing decision value. If the effectiveness evaluation warning value or the timing decision value exceeds the corresponding preset threshold, a stability abnormality signal is generated.

4. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 3 is characterized in that: If the effectiveness evaluation warning value and the timing decision value do not exceed the corresponding preset thresholds, the stability input abnormality value is obtained by weighted summation of the effectiveness evaluation warning value, the timing decision value and the obstacle judgment value. If the stability input abnormality value exceeds the preset stability input abnormality threshold, a stability abnormality signal is generated; If the stability input deviation value does not exceed the preset stability input deviation threshold, a stability qualified signal is generated.

5. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 2 is characterized in that: The stability analysis output module is communicatively connected to the accuracy test module. When the accuracy test module receives the stability qualified signal, it analyzes the accuracy status of the bearing surface quality detection and evaluation, generates an accuracy qualified signal or an accuracy abnormality signal through the analysis, and sends the accuracy qualified signal or the accuracy abnormality signal to the display alarm terminal.

6. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 5 is characterized in that: The specific analysis process of the accuracy test module includes: Conduct several surface quality inspections and assessments on bearings known to have multiple defects. If the result consistency analysis value exceeds the preset result consistency analysis threshold, the identification judgment symbol XP-1 is assigned to the test result of the corresponding bearing; if the result consistency analysis value does not exceed the preset result consistency analysis threshold, when the identification quasi-difference value or the identification judgment value exceeds the corresponding preset threshold, the identification judgment symbol XP-1 is assigned to the test result of the corresponding bearing; After completing the test on several bearings with multiple defects, if the number of times the identification judgment symbol XP-1 is assigned is zero, a qualified accuracy signal is generated; otherwise, an abnormal accuracy signal is generated.

7. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 5 is characterized in that: The accuracy test module is communicatively connected to the bearing production line optimization judgment module. When the bearing production line optimization judgment module receives the accuracy qualified signal, it analyzes the production status of the bearing production line during the detection period, generates a bearing production qualified signal or a bearing production abnormality signal through analysis, and sends the bearing production qualified signal or the bearing production abnormality signal to the display alarm terminal.

8. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 7 is characterized in that: The specific analysis process of the bearing production line optimization judgment module includes: If a defect type corresponding to the prone judgment symbol WP-1 exists during the detection period, a bearing production abnormality signal is generated; if a defect type corresponding to the prone judgment symbol WP-1 exists during the detection period, a bearing production abnormality signal is generated when the bearing scrap analysis value exceeds the preset bearing scrap analysis threshold; otherwise, a bearing production qualified signal is generated.

9. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 7 is characterized in that: The bearing production line optimization judgment module is communicatively connected to the bearing production management judgment module. When the bearing production management judgment module receives the bearing production qualified signal, it analyzes the management performance of the bearing production during the detection period, generates a management qualified signal or a management abnormality signal through the analysis, and sends the management qualified signal or the management abnormality signal to the display alarm terminal.

10. The bearing surface quality detection and evaluation system based on the visual cognitive model according to claim 9 is characterized in that: The specific analysis process of the bearing production management judgment module is as follows: The management judgment coefficient is calculated by weighted summing up the management non-good value, production line maintenance value and operation error frequency. If the management judgment coefficient exceeds the preset management judgment coefficient threshold, a management abnormality signal is generated; if the management judgment coefficient does not exceed the preset management judgment coefficient threshold, a management qualification signal is generated.

Images