Development method of risk assessment model of surface coating quality and risk assessment method

Abstract

The invention relates to the field of steel surface quality assessment, in particular to a surface coating quality risk assessment model development method and a surface coating quality risk assessment method.The surface coating quality risk assessment model development method comprises the steps that image data before and after steel part surface coating is obtained through a hyperspectral imaging system; preprocessing the image data and extracting initial quality data; grid division is carried out on the surface of the steel part coating, pixelization is carried out on the initial quality data, and surface feature data of the surface of the steel part coating are extracted; according to the method, deep learning is performed by using the surface feature data to form the risk assessment model, and the surface feature data is used as a database to obtain the risk assessment model, so that on one hand, a model capable of performing risk assessment on steel surface coating is provided, and the coating quality and the production efficiency are improved; on the other hand, surface feature data in the improved risk assessment model is provided, and the risk assessment model with more accurate data and easier operation is obtained.

Description

Technical Field

[0001] The present invention relates to the field of steel surface quality assessment, and in particular to a method for developing a risk assessment model for surface coating quality and a method for risk assessment of surface coating quality. Background Art

[0002] In the machinery manufacturing industry, the surface coating quality of steel parts is crucial. The quality of the coating process directly affects the appearance, corrosion resistance and service life of steel parts. In order to ensure the quality of coating, a strict quality assessment and risk management system needs to be implemented.

[0003] The existing method of using deep learning models to achieve coating quality assessment usually uses defect images to train the deep learning model. However, due to the large number of defective images on the steel production line, in actual inspection, many servers are often required to work together to improve the efficiency and accuracy of quality assessment. Summary of the invention

[0004] The present invention is made to solve the above-mentioned technical problems, and its purpose is to provide a method for developing a risk assessment model for surface coating quality and a risk assessment method for surface coating quality. The problem of large data volume in subsequent identification can be avoided by acquiring pixel data and forming an assessment model using pixel data.

[0005] In order to achieve the above-mentioned purpose, the present invention provides a method for developing a risk assessment model for surface coating quality, comprising: acquiring image data of the surface of a steel part before and after coating through a hyperspectral imaging system; preprocessing the image data and extracting preliminary quality data; gridding the coating surface of the steel part, pixelating the preliminary quality data, and extracting surface feature data of the coating surface of the steel part; and using the surface feature data for deep learning to form a risk assessment model.

[0006] Preferably, the preliminary quality data includes coating thickness, coating uniformity, coating adhesion and coating surface defects.

[0007] Preferably, the step of gridding the coating surface of the steel part, pixelating the preliminary quality data, and extracting surface characteristic data of the coating surface of the steel part comprises: gridding the coating surface of the steel part and marking it with numbers, and extracting spectral data within each grid; pixelating the preliminary quality data within the grid, and recording the number of pixel data within each grid; and extracting surface characteristic data of the coating surface of the steel part based on the preliminary quality data.

[0008] Preferably, the use of the surface feature data for deep learning to form a risk assessment model includes: standardizing the surface feature data; training the standardized surface feature data using a convolutional neural network; extracting defect data of the surface coating of steel parts through image segmentation, and classifying the defect data to form defect categories; testing and verifying the defect categories through the image data, and obtaining an assessment level based on the test results, and gradually adjusting to form the risk assessment model.

[0009] Preferably, the evaluation level is determined by a preset threshold and the defect category.

[0010] Preferably, it also includes obtaining actual painting test data as a test sample, adjusting the test data according to the test sample to test and verify the defect category; adjusting the painting process parameters during the actual painting according to the test results; and evaluating the surface coating quality of the steel parts with the adjusted painting process parameters.

[0011] The present invention provides a method for risk assessment of surface coating quality, comprising: obtaining a risk assessment model formed by the development method described above, and installing the risk assessment model on a coating production line; obtaining image data and surface feature data of a steel part surface in real time, determining an assessment level, and feeding back the data to a monitoring system; and adjusting the coating process parameters according to the test results.

[0012] Preferably, it also includes real-time acquisition and storage of all surface feature data and assessment data during risk assessment, expansion and updating of the database in the risk assessment model, and optimization of the risk assessment model.

[0013] Preferably, the risk assessment model is integrated into the coating system, and the assessment data acquired by the risk assessment model is summarized in the coating system; an assessment report and a coating report are formed based on the assessment data.

[0014] Preferably, the coating system is communicatively connected to a user terminal; the user terminal can monitor the image data of the steel part surface and the evaluation grade in real time and can receive the evaluation report and the coating report.

[0015] According to the above description and practice, the method for developing a risk assessment model for surface coating quality described in the present invention includes: obtaining image data before and after coating of a steel part through a hyperspectral imaging system; preprocessing the image data and extracting preliminary quality data, the preliminary quality data including but not limited to parameters such as coating thickness, uniformity, adhesion and surface defects, and using it as a tool for preliminary screening and auxiliary analysis; gridding the coating surface of the steel part, pixelating the preliminary quality data, extracting surface feature data of the coating surface of the steel part, taking each grid as a unit, pixelating the preliminary quality data within the unit, forming a unified standard for different types of preliminary quality data, and statistically obtaining surface feature data from the pixel data under the standard; using the surface feature data for deep learning to form a risk assessment model, using the surface feature data as a database to train a deep learning model and then obtaining a risk assessment model, on the one hand, providing a model for risk assessment of steel surface coating, thereby improving coating quality and production efficiency; on the other hand, providing surface feature data within the improved risk assessment model, and obtaining a risk assessment model with more accurate data and easier operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a flowchart of a method for developing a risk assessment model for surface coating quality according to an embodiment of the present invention.

[0017] Figure 2 The present invention is a flowchart of a method for developing a risk assessment model for surface coating quality according to another embodiment of the present invention.

[0018] Figure 3 The present invention is a flowchart of a method for developing a risk assessment model for surface coating quality according to yet another embodiment of the present invention.

[0019] Figure 4 The present invention is a flowchart of a method for risk assessment of surface coating quality in accordance with an embodiment of the present invention.

[0020] Figure 5 The present invention is a flowchart of a method for risk assessment of surface coating quality involved in yet another embodiment of the present invention. DETAILED DESCRIPTION

[0021] The exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and fully convey the concepts of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0022] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, so their repeated description will be omitted. It should be noted that in the present invention, the terms "including", "configured with", and "set in" are used to express the meaning of open-ended inclusion, and mean that there may be other elements / components / etc. in addition to the listed elements / components / etc.; the terms "first", "second", etc. are only used as marks, not to limit the number or order of their objects; the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate the orientation or position relationship based on the orientation or position relationship shown in the accompanying drawings, only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0023] Unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] The present invention discloses a method for developing a risk assessment model for surface coating quality. Figures 1 to 5 In some embodiments, a method for developing a risk assessment model for surface coating quality includes:

[0025] Step S1: Acquire image data of the steel surface before and after painting through a hyperspectral imaging system.

[0026] When developing the risk assessment model, it is necessary to first obtain a certain amount of data on the quality of steel surface coating, install the hyperspectral imaging system at a specific location on the steel coating production line, and obtain image data of the steel surface before and after coating according to the frequency of steel coating, and store it in the database. The hyperspectral imaging system combines imaging technology with spectral technology to obtain the two-dimensional geometric space and one-dimensional spectral information of the steel surface, and obtain continuous, narrow-band image data with high spectral resolution.

[0027] Step S2: Preprocess the image data and extract preliminary quality data.

[0028] When using a hyperspectral imaging system to obtain image data, it is inevitable that the image data will have errors due to interference noise or atmospheric effects. Therefore, before using the above image data, the image data needs to be preprocessed, wherein the preprocessing may include noise reduction, image fusion or dimensionality reduction, etc., to obtain error-free image data. Screening and extraction are performed in the preprocessed image data to obtain data related to the quality of steel surface coating as preliminary quality data.

[0029] Among them, in some application scenarios, preliminary quality data includes but is not limited to parameters such as coating thickness, coating uniformity, coating adhesion and coating surface defects, and is used as a tool for preliminary screening and auxiliary analysis.

[0030] In other application scenarios, in order to obtain more accurate data on the quality of surface coating, for coating thickness, a thickness gauge can be installed near the hyperspectral imaging system to measure and calculate the coating thickness on the steel surface. The thickness gauge includes but is not limited to ultrasonic coating thickness gauge, magnetic thickness gauge and eddy current thickness gauge, etc.; for coating adhesion, grids with a certain width and depth can be divided on the coating surface, and the grids can be pasted and torn off using tape or a pull-off tester, and the adhesion of the coating on the steel surface can be evaluated by the shedding of the coating; for uniformity or surface defects, image processing software can be used to analyze the two-dimensional geometric spatial image of the steel surface in the hyperspectral imaging system, and uniformity or surface defects can be confirmed by features such as color, texture and shape.

[0031] Step S3: meshing the steel component coating surface, pixelating the preliminary quality data, and extracting surface feature data of the steel component coating surface.

[0032] The coating surface of steel parts obtained by the hyperspectral imaging system is divided into grids one by one, and each grid is taken as a unit. The preliminary quality data within the unit is pixelated, and different types of preliminary quality data are formed into a unified standard. The pixel data under this standard is statistically analyzed to obtain surface feature data.

[0033] Step S4: Use surface feature data for deep learning to form a risk assessment model.

[0034] The surface feature data collected, acquired and processed are used as a database for model training. Specifically, a convolutional neural network structure can be selected. On the one hand, it provides a model for risk assessment of steel surface coating, which improves coating quality and production efficiency; on the other hand, it provides surface feature data within the improved risk assessment model, obtaining a risk assessment model with more accurate data and easier operation.

[0035] Furthermore, in order to ensure that the acquired surface feature data can have a certain unified standard and facilitate subsequent analysis, in some application scenarios, the steel coating surface is gridded, the preliminary quality data is pixelated, and the surface feature data of the steel coating surface is extracted, including:

[0036] Step S31: Divide the steel coating surface into grids and mark them with numbers, and extract the spectral data in each grid.

[0037] The surface of the steel layer is divided into multiple grids, and a complete image is divided into multiple areas. The divided grids are numbered and marked for easy subsequent correspondence. The spectral data in each grid is extracted to obtain the effect of the surface quality of the layer in the grid, store it in the database, and classify it according to the corresponding number.

[0038] Step S32: pixelate the preliminary quality data in the grids and record the number of pixel data in each grid.

[0039] When there are coating surface defects in the corresponding grid, the spectral data in the grid is identified and stored at the pixel level, and the number of defective pixels in the grid is recorded in the database.

[0040] Step S33: extracting surface feature data of the steel component coating surface based on the preliminary quality data.

[0041] Among them, surface feature data include color histogram, texture features and shape features, etc. The color histogram can determine the color distribution map of the layer according to the color distribution information; the texture features can be obtained by texture analysis methods such as local binary pattern or gray level co-occurrence matrix; the shape features can be obtained according to image data, such as area, perimeter or convex hull, etc. According to the above dimensions, the surface feature data of the steel coating surface is obtained by classification. The surface feature data can be used as the evaluation dimension for subsequent model training or surface coating quality.

[0042] Furthermore, using surface feature data to train the risk assessment model specifically includes:

[0043] Step S41: Standardizing the surface feature data.

[0044] Among them, the surface feature data are summarized and the mean is removed to achieve standardization. When the model is trained, the distribution of each dimension of the output surface feature data is similar, ensuring that the same learning rate and activation function can be set when training with the deep learning model, which simplifies the training difficulty of the deep learning model.

[0045] Step S42: using a convolutional neural network to train the surface feature data after the normalization process.

[0046] Among them, the convolutional neural network includes but is not limited to multi-layer convolutional layers, pooling layers, fully connected layers, and Dropout layers. The convolutional layer is responsible for extracting the features of the input data. The multi-layer convolutional layers can extract different types of features and send the extracted features to the pooling layer. The pooling layer reduces the spatial dimension of the output of the convolutional layer, reduces the number of parameters and the amount of calculation, and keeps the features unchanged. The fully connected layer integrates the features of the convolutional layer and the pooling layer. The Dropout layer randomly resets the output values ​​of some neurons to zero during the training process in the entire convolutional neural network, reducing the complex adaptation relationship between multiple neurons. The convolutional neural network is trained using standardized surface feature data to avoid the dependency of continuous surface feature data during training, thereby forming a more accurate and adaptable risk assessment model.

[0047] Step S43: extracting defect data of the surface coating of the steel part by image segmentation, and classifying the defect data into defect categories.

[0048] The surface feature data extracted from the image is classified to obtain defect data in the surface feature data. Specifically, the defect data is classified into defect categories using different colors, textures, and shapes, and different types of defect data are weighted using a weighted loss function to distinguish the costs of different types of defects and classify them in the convolutional neural network.

[0049] Step S44: testing and verifying the defect category through image data, obtaining an assessment level based on the test results, and gradually adjusting to form a risk assessment model.

[0050] After obtaining the defect category in step S43, the image data of the steel surface after painting is obtained by using the hyperspectral imaging system for testing and verification, and the evaluation level is obtained according to the test results, including: no risk, low risk, medium risk and high risk, and different measures are taken for the steel surface according to the corresponding risk level, including but not limited to re-spraying or directional repainting. After obtaining the evaluation level, human assistance is used to determine whether the evaluation level is accurate and consistent with the actual situation. If not, the classification value is gradually adjusted, and the classification standard is formulated until a risk assessment model that meets the needs is formed.

[0051] In some application scenarios, the assessment level is determined by preset thresholds and defect categories, where the preset thresholds can be well-known standards, such as national standards, and are graded according to test results, data in the database, and the quality of surface coating on steel parts. The specific assessment level is set based on actual conditions to ensure the accuracy and effectiveness of the entire risk assessment model.

[0052] It is understandable that in the process of forming a risk assessment model using the image data before and after steel surface coating, the image data is subjected to a variety of processing, including but not limited to splitting, noise reduction, specific feature extraction and other operations. After a variety of processing, the recognition data in the risk assessment model inevitably has a certain error with the actual image data. In order to ensure that the risk assessment model can still recognize the unprocessed initial image data, in some application scenarios, the development method of the risk assessment model of surface coating quality also includes:

[0053] Step S5: Acquire actual coating test data as test samples, and adjust the test data according to the test samples to test and verify the defect category.

[0054] The image data without any processing is reused, which is the actual painting test data. The actual painting test data is used as a test sample, and the test data is adjusted according to the test sample and sent to the risk assessment model for identification.

[0055] Step S6: Adjust the coating process parameters during actual coating according to the test results.

[0056] It is understandable that the standards for surface coating quality are different for different steel materials or coating processes. In order to enable the risk assessment model to adapt to different steel materials or coating processes, the coating process parameters are gradually adjusted according to the test results so that the risk assessment model can adapt to different steel materials or coating processes, thereby ensuring the accuracy of the risk assessment model.

[0057] Step S7: Evaluate the coating quality of the steel part surface after the coating process parameters are adjusted.

[0058] Observe and adjust the coating process parameters to see if the coating quality meets the requirements of the corresponding steel or coating process. Once it is determined to be compliant, the risk assessment model can be applied to the conditions of the steel type and coating process, further improving the accuracy of risk assessment and expanding the application scenarios of the risk assessment model.

[0059] The present invention provides a method for risk assessment of surface coating quality. The above risk assessment model is used to implement risk assessment of steel surface coating quality. In some application scenarios, the method for developing the risk assessment model for surface coating quality also includes:

[0060] Step S101: Obtain the risk assessment model formed by the above development method, and install the risk assessment model on the coating production line.

[0061] The adjusted risk assessment model adapted to the type of steel and coating process is installed on the actual coating production line to monitor the coating quality of the steel surface on the coating production line in real time. It can also decide whether to rework or perform subsequent repainting operations based on the actual monitoring results, greatly improving the quality of the coating on the steel surface, avoiding subsequent rework, and reducing the waste of manpower and material resources.

[0062] Step S102: Acquire image data and surface feature data of the steel part surface in real time, determine the evaluation level, and feed back to the monitoring system.

[0063] When evaluating the quality of steel surface coating, the risk assessment model obtains real-time image data of the steel surface through the hyperspectral imaging system and processes the image data within the risk assessment model to obtain surface feature data. The grade of steel surface coating quality is determined based on the surface feature data, and the grade is fed back to the monitoring system for subsequent processing.

[0064] Among them, the monitoring system is set up near the steel surface coating production line, including a display screen and an alarm structure. On the one hand, it is used to receive and view the data fed back by the risk assessment model in real time, and correspond the preset standards to the assessment levels one by one, so as to timely warn of coating defects on the steel surface and select different treatment methods according to the preset standards.

[0065] Step S103: Adjust coating process parameters according to the test results.

[0066] When the monitoring system receives a poor evaluation grade and the surface coating defects are numerous and continue to occur, the coating process parameters of the monitored coating production line are adjusted according to the test results to reduce the coating defects on the steel surface.

[0067] Understandably, since monitoring of coating production lines often requires long-term and intensive work, during the entire life cycle of the equipment, due to equipment wear or changes in the external environment, the data in the risk assessment model often needs to be updated and calibrated to adapt to changes in the equipment and the external environment, and to achieve accurate identification of the quality of steel surface coating. Therefore, the risk assessment method also includes:

[0068] Step S104: Acquire and save all surface feature data and assessment data during risk assessment in real time, expand and update the database in the risk assessment model, and optimize the risk assessment model.

[0069] In some application scenarios, in order not to impose a large operating burden on the risk assessment model and the servers in the monitoring system, the risk assessment model is calibrated at a preset interval based on the surface feature data and assessment data obtained by the risk assessment model. After the calibration is completed, all assessment data within this preset period of time is uploaded to a cloud platform such as a database, which not only realizes the real-time calibration of the risk assessment model, but also reduces the burden on the servers in the system.

[0070] Furthermore, in some application scenarios, the risk assessment model is integrated into the coating system to timely identify and predict defects and improve the surface coating quality. Specifically, the coating system includes systems such as HMI (Human Machine Interface) and SCADA (Supervisory Control And Data Acquisition). The corresponding coating system is improved and the corresponding interface of the risk assessment model is added to facilitate the aggregation of the assessment data obtained by the risk assessment model into the coating system, and the formation of an assessment report and a coating report based on the assessment data. Operators can use the assessment report and the coating report to intuitively analyze the steel surface coating production line, determine the areas that need to be optimized or adjusted in the entire production line, and further ensure the quality of the steel surface coating.

[0071] In addition, in order to facilitate operators and analysts to remotely monitor and interact with the risk assessment model, in some application scenarios, the coating system is connected to the user end for communication. The user end can monitor the image data and assessment level of the steel surface in real time, and can receive assessment reports and coating reports, thus realizing remote data visualization. The coating effect of the steel surface coating production line can be quickly obtained through the summarized reports, which more intuitively displays the coating surface quality problems and their corresponding assessment levels.

[0072] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A method for developing a risk assessment model for surface coating quality, characterized in that: include: The image data of the steel surface before and after painting is obtained through the hyperspectral imaging system; Preprocessing the image data and extracting preliminary quality data; Meshing the steel coating surface, pixelating the preliminary quality data, and extracting surface feature data of the steel coating surface; The surface feature data is used for deep learning to form a risk assessment model.

2. The development method according to claim 1, characterized in that: The preliminary quality data includes coating thickness, coating uniformity, coating adhesion and coating surface defects.

3. The development method according to claim 1, characterized in that: The step of meshing the steel coating surface, pixelating the preliminary quality data, and extracting surface feature data of the steel coating surface includes: The steel coating surface is divided into grids and numbered, and the spectral data in each grid is extracted; Pixelating the preliminary quality data in the grids, and recording the number of pixel data in each grid; Surface feature data of the coating surface of the steel part is extracted based on the preliminary quality data.

4. The development method according to claim 3, characterized in that: The method of using the surface feature data for deep learning to form a risk assessment model includes: Performing standardization on the surface feature data; Using a convolutional neural network to train the surface feature data after the standardized processing; Extracting defect data of the surface coating of steel parts by image segmentation, and classifying the defect data to form defect categories; The defect category is tested and verified by the image data, and an assessment level is obtained according to the test result, and is gradually adjusted to form the risk assessment model.

5. The development method according to claim 4, characterized in that: The evaluation level is determined by a preset threshold and the defect category.

6. The development method according to claim 4, characterized in that: Also includes: Acquire actual coating test data as a test sample, and adjust the test data according to the test sample to test and verify the defect category; Adjusting the coating process parameters during the actual coating according to the test results; The coating quality of steel parts surface after adjusting the coating process parameters was evaluated.

7. A method for risk assessment of surface coating quality, characterized in that: include: Obtain a risk assessment model formed by the development method described in any one of claims 1 to 6, and install the risk assessment model on a coating production line; Acquire image data and surface feature data of steel parts in real time, determine the evaluation level, and feed it back to the monitoring system; The coating process parameters are adjusted according to the test results.

8. The risk assessment method according to claim 7, characterized in that: Also includes: All surface feature data and assessment data during risk assessment are acquired and saved in real time, the database in the risk assessment model is expanded and updated, and the risk assessment model is optimized.

9. The risk assessment method according to claim 7, characterized in that: The risk assessment model is integrated into the coating system, and the assessment data acquired by the risk assessment model is aggregated into the coating system; An evaluation report and a coating report are generated based on the evaluation data.

10. The risk assessment method according to claim 9, characterized in that: The coating system is communicatively connected to the user end; The user terminal can monitor the image data of the steel part surface and the evaluation grade in real time and can receive the evaluation report and the coating report.

Images