A method for evaluating the quality of rubber bellows processing

Abstract

The present invention relates to the technical field of rubber bellows processing quality inspection. The present invention relates to a method for evaluating the effect of rubber bellows processing quality inspection. The method comprises the following steps: S1, obtaining process parameters of rubber bellows production equipment; S2, obtaining historical production data, combining the historical production data with process parameters to screen finished product data, and establishing a finished product database; S3, collecting finished product images of rubber bellows, and performing similarity matching analysis between the finished product images and the finished product database; the present invention avoids errors caused by subjective judgment by conducting analysis and decision-making based on a large amount of production data, and in the abnormal scoring link, assigns weights according to the importance of difference features, quantifies quality problems, provides data support for quality inspection and improvement, and adopts reasonable inspection strategies to give priority to inspecting high-risk products, reduce unnecessary inspection costs, and reduce rework and scrap costs caused by unqualified products.

Description

Technical Field

[0001] The present invention relates to the technical field of rubber bellows processing quality detection, in particular to an effect evaluation method for rubber bellows processing quality detection. Background Art

[0002] In the production and manufacturing of rubber bellows, ensuring product quality is of vital importance. Existing technologies play an important role in the quality inspection of rubber bellows. Its purpose is to identify whether there are quality defects in the product and ensure that the products shipped meet the relevant standards, thereby improving the reliability and safety of the product and meeting the market demand for high-quality rubber bellows.

[0003] At present, the quality inspection of rubber bellows is mostly carried out in the spot inspection link after the product is produced. In this scenario, there are many defects. Since the rubber bellows sampled are randomly selected, they cannot represent the overall rubber bellows. Secondly, if each rubber bellows is inspected in detail, it will not only greatly affect the production efficiency, but also make the production cost uncontrollable. Therefore, a method for evaluating the effect of rubber bellows processing quality inspection is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for evaluating the effect of rubber bellows processing quality detection to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, a method for evaluating the quality of rubber bellows processing is provided, comprising the following steps:

[0006] S1. Obtaining process parameters of rubber bellows production equipment;

[0007] S2. Obtain historical production data, combine the historical production data with process parameters to screen finished product data, and establish a finished product database;

[0008] S3. Collect finished product images of rubber bellows, perform similarity matching analysis on the finished product images and the finished product database, set a similarity threshold, mark the rubber bellows as abnormal based on the similarity threshold, and extract difference features between the abnormally marked finished product images and the standard finished product data;

[0009] S4: Combine the process parameters with the production equipment to simulate the finished product production. Combine the production simulation data with the historical production data to set an abnormal value for each difference feature. Then, combine the difference features extracted in S3 with the set abnormal value to perform an abnormality score.

[0010] S5. Prioritize the processing quality inspection of the rubber bellows based on the abnormality score, and input the inspection results into the historical production data to update the abnormality value setting.

[0011] As a further improvement of the present technical solution, the S1 establishes a data connection with the production management end of the rubber bellows, thereby acquiring in real time the process parameters inputted into the production equipment by the production management end.

[0012] As a further improvement of this technical solution, the steps of S2 are as follows:

[0013] S2.1. Extract historical production data of rubber bellows on the production management side;

[0014] S2.2. Combine historical production data with process parameters to screen finished product data, thereby obtaining finished product data corresponding to each process parameter, and then extract finished product data that is the same as the real-time process parameters to establish a finished product database.

[0015] As a further improvement of this technical solution, the steps of S3 are as follows:

[0016] S3.1. Collect finished product images of the rubber bellows, perform similarity matching analysis on the finished product images and each finished product data in the finished product database, and obtain the similarity between the finished product images and each finished product data;

[0017] S3.2. Set a similarity threshold, then perform an anomaly comparison based on the similarity threshold and the similarity between the finished product image and each finished product data. If the similarity between the finished product image and the finished product data is lower than the similarity threshold, the anomaly is marked. Conversely, if the similarity between the finished product image and the finished product data is higher than the similarity threshold, the anomaly is marked based on the most similar finished product data.

[0018] S3.3. Obtain standard finished product data according to the process parameters, then extract difference features from the abnormally marked finished product image in combination with the standard finished product data to obtain difference features corresponding to the abnormally marked finished product image.

[0019] As a further improvement of the present technical solution, S3.1 flips the rubber bellows produced each time the production equipment produces a rubber bellows, and uses a photographing device to collect appearance images, and merges the collected appearance images into a finished product image of the rubber bellows.

[0020] As a further improvement of the present technical solution, the finished product data in S3.2 is divided into qualified rubber bellows and damaged rubber bellows. When the most similar finished product data corresponding to the finished product image is a damaged rubber bellows, an abnormal mark is performed;

[0021] When the most similar finished product data corresponding to the finished product image is a qualified rubber bellows, no abnormality mark is performed.

[0022] As a further improvement of this technical solution, S3.1 further includes the following steps:

[0023] S3.1.1. Detect cracks in finished product images;

[0024] S3.1.2. When the test result shows that the finished product image of the rubber bellows contains cracks, the rubber bellows is judged to be unqualified. Conversely, when the test result shows that the finished product image of the rubber bellows does not contain cracks, a similarity analysis is performed.

[0025] As a further improvement of this technical solution, the steps of S4 are as follows:

[0026] S4.1. Simulate the production of finished rubber bellows products by combining process parameters with production equipment, thereby obtaining production simulation data containing multiple finished products;

[0027] S4.2. Extract features of damaged rubber bellows using the production simulation data in combination with historical production data. Then, set abnormal numerical values ​​based on the extracted features of damaged rubber bellows in combination with each difference feature.

[0028] S4.3. Calculate an anomaly score by combining the difference features corresponding to the finished product image with the anomaly value to obtain an anomaly score for each anomaly-marked finished product image.

[0029] As a further improvement of the present technical solution, the S5 performs a priority inspection on the processing quality of the rubber bellows according to the abnormality score, and prioritizes the processing quality inspection on the rubber bellows with the highest abnormality score.

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

[0031] 1. In this method for evaluating the effectiveness of rubber bellows processing quality inspection, a data connection is established with the production management end to obtain process parameters in real time, and a finished product database is established in combination with historical production data to provide an accurate reference for subsequent inspections. In the image inspection link, not only crack detection is performed, but also quality problems of rubber bellows can be accurately identified by setting similarity thresholds and extracting difference features, such as slight differences from standard finished product data, effectively preventing unqualified products from entering the market and improving the overall quality of the product.

[0032] 2. In this method for evaluating the effect of rubber bellows processing quality inspection, by combining process parameters with production equipment to simulate finished product production, potential quality problems can be discovered in advance. By analyzing simulation data and historical production data, reasonable abnormal values ​​are set to monitor and adjust the production process. When products with high abnormality scores are detected, they are given priority for inspection and processing, and production deviations are corrected in a timely manner to avoid the production of a large number of unqualified products, optimize the production process, and improve production efficiency.

[0033] 3. This method for evaluating the effectiveness of rubber bellows processing quality inspection avoids errors caused by subjective judgment by analyzing and making decisions based on a large amount of production data. In the abnormal scoring link, weights are assigned according to the importance of difference features, quality problems are quantified, and data support is provided for quality inspection and improvement. A reasonable inspection strategy prioritizes the inspection of high-risk products, reduces unnecessary inspection costs, and at the same time reduces the rework and scrap costs caused by unqualified products. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is the overall flow chart of the present invention;

[0035] Figure 2 This is a flowchart of the process of extracting historical production data of rubber bellows at the production management end of the present invention;

[0036] Figure 3 A flowchart of collecting images of finished rubber bellows according to the present invention;

[0037] Figure 4 A flowchart of obtaining an abnormality score for each abnormally marked finished product image according to the present invention. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0039] See also Figures 1-4 As shown, the purpose of this embodiment is to provide a method for evaluating the effect of rubber bellows processing quality detection, including the following steps:

[0040] S1. Obtaining process parameters of rubber bellows production equipment;

[0041] The S1 establishes a data connection with the production management end of the rubber bellows to obtain the process parameters input to the production equipment by the production management end in real time. The specific steps are as follows:

[0042] Preparation: Determine the communication method between the production management terminal and the device used to obtain parameters (such as a computer). Common communication methods include Ethernet communication and serial communication.

[0043] Establish a connection: If Ethernet communication is used, set the network parameters on the device where the parameters are obtained, such as IP address, subnet mask, gateway, etc., according to the network interface specifications provided by the production management end. Use the corresponding network programming interface or tool to send a connection request to the production management end to establish a network connection. For example, achieve communication connection with the production management end through socket programming.

[0044] If serial communication is used, correctly connect the serial cable between the production management end and the serial port interface of the device from which the parameters are to be obtained. Configure the serial port parameters on the device, such as baud rate, data bits, stop bits, and parity bit, to ensure that they are consistent with those on the production management end. Then, use serial communication software or a programming interface to open the serial port and establish a connection.

[0045] Send parameter acquisition request: After establishing a connection and passing authentication (if necessary), send a process parameter acquisition request to the production management end according to the data interaction protocol of the production management end. After receiving the parameter acquisition request, the production management end will return the process parameters currently input to the production equipment in the agreed format.

[0046] S2. Obtain historical production data, combine the historical production data with process parameters to screen finished product data, and establish a finished product database;

[0047] The steps of S2 are as follows:

[0048] S2.1. Extract historical production data of rubber bellows on the production management side;

[0049] The historical production data of rubber bellows is extracted from the database of the production management end through query statements.

[0050] S2.2. Combine historical production data with process parameters to screen finished product data, thereby obtaining finished product data corresponding to each process parameter. Then, extract finished product data that has the same real-time process parameters to establish a finished product database. The specific steps are as follows:

[0051] Filter finished product data based on process parameters: For each finished product data record, compare its corresponding process parameters with the pre-set screening conditions, classify the filtered finished product data according to different process parameters, and for each process parameter group, compile a list of finished product data under that process parameter;

[0052] Extract finished product data that are identical to the real-time process parameters and establish a finished product database: Match the real-time process parameters with the finished product data corresponding to each process parameter obtained previously. Matching can be achieved by comparing the real-time process parameters with the process parameter values ​​in each group, finding finished product data records that are exactly the same as the real-time process parameters, extracting these matched finished product data, organizing these data according to a certain data structure (such as a table form), and storing them in a database, thereby establishing a finished product database.

[0053] S3. Collect finished product images of rubber bellows, perform similarity matching analysis on the finished product images and the finished product database, set a similarity threshold, mark the rubber bellows as abnormal based on the similarity threshold, and extract difference features between the abnormally marked finished product images and the standard finished product data;

[0054] The steps of S3 are as follows:

[0055] S3.1. Collect finished product images of rubber bellows and perform similarity matching analysis on the finished product images and each finished product data in the finished product database to obtain the similarity between the finished product images and each finished product data. The formula is as follows:

[0056] ;

[0057] Among them, S j is the similarity between the finished product image and each finished product data in the finished product database, n is the dimension of the feature vector, m is the number of finished product data in the finished product database, F image，i is the i-th element of the finished image feature vector, F database，i，j is the i-th element of the feature vector of the j-th finished product data in the finished product database.

[0058] In the step S3.1, each time a rubber bellows is produced by the production equipment, the produced rubber bellows is turned over, and at the same time, appearance images are collected by a photographing device, and the collected appearance images are merged into a finished product image of the rubber bellows.

[0059] Use a high-resolution shooting device (such as an industrial camera) to collect appearance images of the rubber bellows from multiple angles, and merge the appearance images collected from different angles to form a complete finished product image of the rubber bellows. An image stitching algorithm can be used to match and stitch the images based on overlapping areas and feature points.

[0060] The S3.1 further comprises the following steps:

[0061] S3.1.1. Detect cracks in finished product images;

[0062] Analyze the image after edge detection to extract possible crack features, connect the broken edges through morphological operations (such as dilation and erosion), and remove small noise points;

[0063] Calculate edge features such as length, width, and continuity. For example, use a contour detection algorithm to find the edge contour and then calculate the perimeter of the contour as the length of the crack.

[0064] Based on the extracted crack features, judgment rules are set to determine whether there are cracks in the image. For example, if the detected edge length exceeds a certain threshold and the edge continuity is good, it is determined that a crack exists.

[0065] S3.1.2. When the test result shows that the finished product image of the rubber bellows contains cracks, the rubber bellows is judged to be unqualified. Conversely, when the test result shows that the finished product image of the rubber bellows does not contain cracks, a similarity analysis is performed.

[0066] S3.2. Set a similarity threshold (determine an appropriate threshold based on the quality requirements and actual production conditions for rubber bellows). Analyze data from past production processes, including the characteristics of qualified and unqualified products and the distribution of their similarities with the standard sample. If the similarities between qualified products and the standard sample are mostly concentrated within a certain range, set the threshold near the lower limit of this range. Also, consider the practical experience of production and quality control personnel, and the acceptable degree of product variation in actual production, to comprehensively determine the similarity threshold.

[0067] Then, the similarity threshold is combined with the similarity between the finished product image and each finished product data to perform an anomaly comparison. When the similarity between the finished product image and the finished product data is lower than the similarity threshold, an anomaly mark is performed. Conversely, when the similarity between the finished product image and the finished product data is higher than the similarity threshold, an anomaly mark is performed based on the most similar finished product data.

[0068] In S3.2, the finished product data is divided into qualified rubber bellows and damaged rubber bellows. When the most similar finished product data corresponding to the finished product image is a damaged rubber bellows, an abnormal mark is performed;

[0069] If the most similar finished product data is a damaged rubber bellows, even if the similarity is higher than the threshold, the finished product image is marked as abnormal because its most similar sample is damaged, indicating that the finished product may have potential problems.

[0070] When the most similar finished product data corresponding to the finished product image is a qualified rubber bellows, no abnormality mark is performed;

[0071] If the most similar finished product data is a qualified rubber bellows, no abnormality mark is performed, and the rubber bellows corresponding to the finished product image is considered to be of qualified quality.

[0072] S3.3. Obtain standard finished product data based on process parameters (based on current production process parameters, such as temperature, pressure, speed, raw material formula, etc.) and perform a query. Filter out standard finished product data records that match the current process parameters from the database. Then, extract differential features from the abnormally marked finished product image combined with the standard finished product data to obtain differential features corresponding to the abnormally marked finished product image.

[0073] For the finished product images with abnormal marks, image preprocessing is first performed. A suitable feature extraction algorithm is used to extract features of the finished product images with abnormal marks, such as edge features. The same feature extraction operation is also performed on the images in the standard finished product data. The features of the finished product images with abnormal marks are compared with the features of the standard finished product data to find out the differences between the two. For example, the differences in edge shape, length, position, etc., or the differences in texture distribution, density, etc.

[0074] S4: Combine the process parameters with the production equipment to simulate the finished product production. Combine the production simulation data with the historical production data to set an abnormal value for each difference feature. Then, combine the difference features extracted in S3 with the set abnormal value to perform an abnormality score.

[0075] The steps of S4 are as follows:

[0076] S4.1. Combine the process parameters with the production equipment to simulate the production of rubber bellows finished products, thereby obtaining production simulation data containing multiple finished products. The specific steps are as follows:

[0077] Simulation of finished rubber bellows production: Conduct in-depth analysis of the rubber bellows production process, clarify the working principles and relationships between process parameters (such as temperature, pressure, time, and raw material properties) and production equipment (such as extruders and vulcanizers), and use physical, mathematical, or machine learning models to describe the production process. For example, use finite element analysis software to establish a mechanical property model of rubber materials under different process parameters.

[0078] Input the current process parameters into the constructed production model. The process parameters can be obtained in real time from the production management system. Start the simulation program to let the model simulate the production process of rubber bellows based on the input process parameters and the characteristics of the production equipment, and generate a variety of possible finished product data, including the size, shape, physical properties and other information of the finished product.

[0079] S4.2. Combine the production simulation data with the historical production data to extract the features of the damaged rubber bellows. Then, combine the extracted features of the damaged rubber bellows with each difference feature to set the abnormal value. The specific steps are as follows:

[0080] Feature extraction of damaged rubber bellows: The generated production simulation data is integrated with historical production data. The historical production data contains information about actual damaged and qualified rubber bellows that occurred in the past production process. Features related to damage are selected from the integrated data. Common features include crack length, hole size, surface defect density, and physical performance indicators (such as abnormal changes in hardness and tensile strength). Image processing algorithms (such as edge detection and morphological operations) can be used to extract crack and hole features.

[0081] Abnormal value setting: Match the extracted damaged rubber bellows features with each difference feature obtained in the previous step to determine which difference features are associated with the damage features. Based on the relationship between the damage features and the difference features, set an abnormal value for each difference feature. The formula is as follows:

[0082] ;

[0083] ;

[0084] Among them, β1 is the slope, β0 is the intercept, k is the number of samples, x e is the damage feature of the e-th sample, y e is the difference eigenvalue of the e-th sample, is the sample mean of the damaged feature, is the sample mean of the difference feature. Based on the estimated β1 and β0, the value of the difference feature y corresponding to different damage features x can be predicted, and the abnormal value can be set according to the predicted value.

[0085] S4.3. Calculate anomaly scores by combining the difference features corresponding to the finished product images with the anomaly values ​​to obtain an anomaly score for each anomaly-marked finished product image. The specific steps are as follows:

[0086] Determination of feature weight: According to the quality requirements of the rubber bellows and the importance of each difference feature to the product quality, a corresponding weight is assigned to each difference feature. The weight value range is usually between 0 and 1, and the sum of all feature weights is 1;

[0087] Matching abnormal values ​​with difference features: Compare and analyze the difference feature value of each finished product image with the pre-set abnormal value;

[0088] Anomaly score calculation: For each difference feature, a score is calculated based on its difference degree and weight. Then, the scores of all difference features are combined to obtain the anomaly score of each abnormally marked finished image.

[0089] Scoring result recording: The calculated anomaly score is associated with the corresponding finished product image, and the anomaly score result of each finished product image is recorded to facilitate subsequent evaluation and analysis of the rubber bellows quality.

[0090] S5. Prioritize the processing quality inspection of the rubber bellows based on the abnormality score, and input the inspection results into the historical production data to update the abnormality value setting.

[0091] S5 performs a priority inspection on the processing quality of the rubber bellows according to the abnormality score, and prioritizes the processing quality inspection of the rubber bellows with the highest abnormality score. The specific steps are as follows:

[0092] Sort by anomaly score: Sort all images of rubber bellows marked as abnormal by their anomaly score from high to low. This step can use a sorting algorithm such as quick sort or bubble sort to ensure that rubber bellows with high anomaly scores are ranked first for priority inspection.

[0093] Priority Inspection: From the sorted list, rubber bellows with the highest abnormality scores are selected for actual processing quality inspection. Inspection content may include inner wall inspection (for cracks, holes, etc.), physical property testing (such as hardness, tensile strength, etc.), dimensional measurement, etc. Using professional testing equipment and methods, detailed quality information of the rubber bellows is obtained;

[0094] Test result record: The test results of the rubber bellows will be recorded in detail, including the values ​​of various test indicators, whether they are qualified, etc. These records will serve as an important basis for subsequent updates to abnormal value settings;

[0095] Update historical production data: input the test results into the historical production database and integrate them with the original historical production data;

[0096] Update of abnormal value settings: Based on the comprehensive analysis of test results and historical production data, the abnormal value settings are adjusted. Analyze the cases where the abnormality score is high but the test result is qualified, and the cases where the abnormality score is low but the test result is unqualified, to find out the possible problems with the abnormal value settings. For example, if it is found that the abnormality threshold of some difference features is set too low, resulting in many actually qualified rubber bellows being marked as abnormal, then the abnormality threshold of the feature is appropriately increased; conversely, if it is found that the abnormality threshold of some features is too high, resulting in some unqualified rubber bellows not being detected as abnormal in time, then the abnormality threshold of the feature is lowered;

[0097] Cyclic detection and update: Repeat the above steps to detect the next abnormal rubber bellows after sorting, and continuously update the historical production data and abnormal value settings according to the detection results until all abnormally marked rubber bellows are detected and processed.

[0098] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the quality of rubber bellows processing, characterized by: The following steps are involved: S1. Obtaining process parameters of rubber bellows production equipment; S2. Obtain historical production data, combine the historical production data with process parameters to screen finished product data, and establish a finished product database; S3. Collect finished product images of rubber bellows, perform similarity matching analysis on the finished product images and the finished product database, set a similarity threshold, mark the rubber bellows as abnormal based on the similarity threshold, and extract difference features between the abnormally marked finished product images and the standard finished product data; The steps of S3 are as follows: S3.

1. Collect finished product images of the rubber bellows, perform similarity matching analysis on the finished product images and each finished product data in the finished product database, and obtain the similarity between the finished product images and each finished product data; The S3.1 further comprises the following steps: S3.1.

1. Detect cracks in finished product images; S3.1.

2. If the test results show that the finished product image of the rubber bellows contains cracks, the rubber bellows is judged to be unqualified. Conversely, if the test results show that the finished product image of the rubber bellows does not contain cracks, similarity analysis is performed; S3.

2. Set a similarity threshold, then perform an anomaly comparison based on the similarity threshold and the similarity between the finished product image and each finished product data. If the similarity between the finished product image and the finished product data is lower than the similarity threshold, the anomaly is marked. Conversely, if the similarity between the finished product image and the finished product data is higher than the similarity threshold, the anomaly is marked based on the most similar finished product data. S3.

3. Obtain standard finished product data based on process parameters, then extract differential features from the abnormally marked finished product image and the standard finished product data to obtain differential features corresponding to the abnormally marked finished product image; S4: Combine the process parameters with the production equipment to simulate the finished product production. Combine the production simulation data with the historical production data to set an abnormal value for each difference feature. Then, combine the difference features extracted in S3 with the set abnormal value to perform an abnormality score. The steps of S4 are as follows: S4.

1. Simulate the production of finished rubber bellows products by combining process parameters with production equipment, thereby obtaining production simulation data containing multiple finished products; S4.

2. Extract features of damaged rubber bellows using the production simulation data in combination with historical production data. Then, set abnormal numerical values ​​based on the extracted features of damaged rubber bellows in combination with each difference feature. S4.

3. Calculate an anomaly score by combining the difference features corresponding to the finished product images with the anomaly values ​​to obtain an anomaly score for each anomaly-marked finished product image; S5. Prioritize the processing quality inspection of the rubber bellows based on the abnormality score, and input the inspection results into the historical production data to update the abnormality value setting.

2. The method for evaluating the processing quality of a rubber bellows according to claim 1, wherein: The S1 establishes a data connection with the production management end of the rubber bellows, thereby obtaining in real time the process parameters inputted into the production equipment by the production management end.

3. The method for evaluating the processing quality of a rubber bellows according to claim 1, wherein: The steps of S2 are as follows: S2.

1. Extract historical production data of rubber bellows on the production management side; S2.

2. Combine historical production data with process parameters to screen finished product data, thereby obtaining finished product data corresponding to each process parameter, and then extract finished product data that is the same as the real-time process parameters to establish a finished product database.

4. The method for evaluating the processing quality of a rubber bellows according to claim 1, wherein: In the step S3.1, each time a rubber bellows is produced by the production equipment, the produced rubber bellows is turned over, and at the same time, appearance images are collected by a photographing device, and the collected appearance images are merged into a finished product image of the rubber bellows.

5. The method for evaluating the processing quality of a rubber bellows according to claim 1, wherein: In S3.2, the finished product data is divided into qualified rubber bellows and damaged rubber bellows. When the most similar finished product data corresponding to the finished product image is a damaged rubber bellows, an abnormal mark is performed; When the most similar finished product data corresponding to the finished product image is a qualified rubber bellows, no abnormality mark is performed.

6. The method for evaluating the processing quality of a rubber bellows according to claim 1, wherein: The S5 performs a priority inspection on the processing quality of the rubber bellows according to the abnormality score, and prioritizes the processing quality inspection on the rubber bellows with the highest abnormality score.

Images