Industrial vision-based forging quality evaluation method, system and device

Abstract

The application discloses a forging and stamping quality evaluation method, system and equipment based on industrial vision, relates to the field of industrial machine vision, and comprises the following steps: establishing a response mapping relationship between a forging and stamping quality evaluation index and a visual parameter, including a direct response relationship and an indirect response relationship; acquiring a multi-source visual detection data set by calling a multi-source image acquisition device based on the visual parameter and performing visual parameter feature extraction; performing response evaluation of the forging and stamping quality evaluation index on the visual feature vector according to the response mapping relationship, obtaining a direct response quality evaluation index and an indirect response quality evaluation index; and generating a quality evaluation result according to the response confidence probability of the direct response quality evaluation index and the indirect response quality evaluation index. The application solves the technical problems of large subjective error, destructive detection, long cycle and isolated quality data in the existing forging and stamping quality evaluation, and achieves the technical effects of reducing subjective error, non-destructive rapid detection and integrated analysis of quality data.

Description

Technical Field

[0001] This application relates to the field of industrial machine vision, and in particular to methods, systems and equipment for quality assessment of forgings based on industrial vision. Background Technology

[0002] As core components in equipment manufacturing, the quality of forgings directly affects the safety and reliability of the entire machine. Therefore, accurate and efficient quality assessment is crucial for industrial production. Currently, the main methods for quality assessment of forgings in the industry rely on manual inspection combined with traditional physicochemical testing. This involves visually inspecting for surface defects and using tensile tests and metallographic analysis to determine mechanical properties and microstructure. These methods are prone to subjective errors due to their over-reliance on human experience. Furthermore, physicochemical testing requires damaging the sample and has a long testing cycle. Additionally, the data from each testing method are isolated, making it difficult to achieve comprehensive correlation and rapid assessment of quality indicators.

[0003] Currently, the quality assessment of forgings faces technical challenges, including large subjective errors, destructive and time-consuming testing, and isolated and difficult-to-correlate quality data. Summary of the Invention

[0004] This application provides a method, system, and equipment for evaluating the quality of forged parts based on industrial vision. It employs techniques such as establishing direct and indirect response mapping relationships between forged part quality evaluation indicators and visual parameters; acquiring detection datasets and extracting visual feature vectors based on visual parameters using multi-source image acquisition equipment; evaluating the feature vectors according to the aforementioned response mapping relationships to obtain direct and indirect response quality evaluation indicators; and generating forged part quality evaluation results by combining the response confidence probabilities of the direct and indirect response indicators. These techniques address the technical problems of large subjective errors, destructive and long detection cycles, and isolated and difficult-to-correlate quality data in existing forged part quality evaluation methods. The application achieves the technical effects of reducing subjective errors, enabling non-destructive and rapid detection, and providing integrated analysis of quality data.

[0005] This application provides a method for quality assessment of forgings based on industrial vision, comprising: establishing a response mapping relationship between forging quality assessment indicators and visual parameters, including direct response relationship and indirect response relationship; calling a multi-source image acquisition device based on visual parameters to obtain a multi-source visual detection dataset, and extracting visual parameter features from the multi-source visual detection dataset; evaluating the response of the extracted visual feature vectors to the forging quality assessment indicators according to the response mapping relationship to obtain direct response quality assessment indicators and indirect response quality assessment indicators; and generating a quality assessment result based on the response confidence probabilities of the direct response quality assessment indicators and indirect response quality assessment indicators.

[0006] In a possible implementation, a response mapping relationship between forging quality assessment indicators and visual parameters is established, and the following processing is performed: Based on the forging quality target, multi-dimensional quality assessment indicators are determined, including internal and surface defects, mechanical properties, chemical structure, and dimensional and surface accuracy; multi-source visual correlation analysis is performed on the multi-dimensional quality assessment indicators to establish correlation relationships, including direct and indirect correlations; based on the correlation relationships, visual factor correlation analysis is performed using the quality assessment indicators as top events to establish event correlation factors and logical relationships; according to the top events, event correlation factors, and logical relationships, a response mapping relationship between the forging quality assessment indicators and visual parameters is established.

[0007] In possible implementations, the following processes are performed: the multi-source vision includes: visible light imaging, three-dimensional topography scanning, infrared thermal imaging, high-speed imaging, and microscopic imaging.

[0008] In a possible implementation, a quality assessment index is used as the top event for visual factor correlation analysis to establish event correlation factors and logical relationships. The following processing is performed: for direct correlation relationships, the related visual sources are decomposed to obtain visual parameter factors; the quality assessment index is used as the top event, and descriptive factor analysis is performed on the top event to obtain evaluation descriptive factors; the evaluation descriptive factors are used as hierarchical targets to perform correlation analysis on the visual parameter factors, identify the response relationship of the visual parameter factors to the evaluation descriptive factors, and determine the event correlation factors based on the response relationship. The event correlation factors are visual parameter factors that have reached the correlation threshold, and their response relationships are used as the logical relationships.

[0009] In a possible implementation, a visual factor association analysis is performed using quality assessment indicators as the top event. Event association factors and logical relationships are established, and the following processing is performed: For indirect association relationships, a reverse causal association analysis is performed on the top event to analyze the direct cause factors that lead to the occurrence of the top event. The cause factors are decomposed layer by layer until the association is a visually measurable factor that can be directly or indirectly observed by at least one visual source, thus constructing a visual factor association tree. Based on the visual factor association tree, a structured response mapping path for the top event is established, event association factors are identified, and logical relationships are established according to the structured response mapping path.

[0010] In a possible implementation, a reverse causal relationship analysis is performed on the top event to analyze the direct causal factors leading to the top event. These causal factors are then decomposed layer by layer downwards until a visually measurable factor that can be directly or indirectly observed by at least one visual source is identified. A visual factor association tree is constructed, and the following processing is performed: The direct causal factors leading to the top event are analyzed, including abnormal microstructure states and deviations in processing parameters. The direct causal factors are then decomposed downwards to analyze the visual state change characteristics of abnormal microstructure states or the visual state change characteristics of the processing process, including static and dynamic measurable factors. Based on the results of the layer-by-layer recursive decomposition, a visual factor association tree is constructed with quality assessment indicators as the root node and visually measurable factors as leaf nodes.

[0011] In a possible implementation, the following processing is performed: the dynamic measurable factor is obtained by analyzing time-series image data, reflecting the rate of change of physical quantities or the structural evolution process of the forging during processing; the static measurable factor is obtained by analyzing image data characterizing the spatial morphology or microstructure of the forging at different times.

[0012] In a possible implementation, the extracted visual feature vectors are used to evaluate the response of forging quality indicators according to the response mapping relationship to obtain direct response quality evaluation indicators and indirect response quality evaluation indicators. The following processing is performed: the extracted visual feature vectors are parsed and labeled with metadata, including geometric features, surface defect features, speckle sequence features, infrared thermal imaging sequence features, and thermal extraction features; according to the metadata parsing and type labeling, the visual feature vectors are imported into the evaluation path corresponding to the response mapping relationship for response evaluation. Among them, feature vectors labeled as geometric features or surface defect features are imported into the direct response relationship evaluation path for image recognition processing, and the direct response quality evaluation indicators are output according to the direct evaluation relationship between the image recognition results and the quality evaluation indicators; features labeled as speckle sequence features, infrared thermal imaging sequence features, or thermal extraction features are imported into the indirect response relationship evaluation path for visual correlation factor extraction, and the evaluation results and confidence scores are aggregated based on the hierarchical correlation relationship of the visual correlation factors to obtain indirect response quality evaluation indicators and label the response confidence probabilities.

[0013] This application also provides a forging quality assessment system based on industrial vision, comprising: a response mapping relationship establishment module, used to establish a response mapping relationship between forging quality assessment indicators and visual parameters, including direct response relationship and indirect response relationship; a visual parameter feature extraction module, used to call a multi-source image acquisition device based on visual parameters to obtain a multi-source visual detection dataset, and to extract visual parameter features from the multi-source visual detection dataset; a response assessment module, used to assess the response of the extracted visual feature vectors to the forging quality assessment indicators according to the response mapping relationship, and obtain direct response quality assessment indicators and indirect response quality assessment indicators; and a quality assessment result generation module, used to generate a quality assessment result based on the response confidence probabilities of the direct response quality assessment indicators and the indirect response quality assessment indicators.

[0014] This application also provides an electronic device, including: a memory for storing executable instructions; and a processor for implementing an industrial vision-based method for evaluating the quality of forgings when executing the executable instructions stored in the memory.

[0015] The proposed method, system, and equipment for quality assessment of forgings based on industrial vision first establishes a response mapping relationship between forging quality assessment indicators and visual parameters, including direct and indirect response relationships. Then, based on the visual parameters, a multi-source image acquisition device is invoked to obtain a multi-source visual inspection dataset. Visual parameter features are extracted from the multi-source visual inspection dataset. The extracted visual feature vectors are then used to evaluate the response of the forging quality assessment indicators according to the aforementioned response mapping relationship, obtaining direct and indirect response quality assessment indicators. Finally, based on the response confidence probabilities of the direct and indirect response quality assessment indicators, a quality assessment result is generated. The proposed method, system, and equipment achieve the technical effects of reducing subjective errors, enabling rapid non-destructive testing, and providing integrated analysis of quality data. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic flowchart of the industrial vision-based forging quality assessment method provided in the embodiments of this application.

[0018] Figure 2A schematic diagram of the structure of the industrial vision-based forging quality assessment system provided in this application embodiment.

[0019] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0020] Figure reference numerals: Response mapping relationship establishment module 10, visual parameter feature extraction module 20, response evaluation module 30, quality evaluation result generation module 40, input device 301, processor 302, memory 303, output device 304. Detailed Implementation

[0021] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In the following description, references to "some embodiments" describe a subset of all possible embodiments; however, it is understood that "some embodiments" may be the same or different subsets of all possible embodiments and may be combined with each other without conflict. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0024] This application provides a method for evaluating the quality of forged parts based on industrial vision, such as... Figure 1 As shown, the method includes:

[0025] Step S100: Establish the response mapping relationship between the quality assessment index of forgings and visual parameters, including direct response relationship and indirect response relationship.

[0026] Specifically, a correspondence is established between forging quality assessment indicators and industrial visual inspection parameters. This correspondence is divided into two categories: direct response relationships and indirect response relationships, providing a basis for subsequent evaluation of forging quality using visual data. The quality objectives of the forgings are determined, thereby identifying quality assessment indicators covering dimensions such as internal and surface defects and mechanical properties. Multi-source visual technologies, including visible light imaging and 3D topography scanning, are used to analyze the correlation between various visual parameters and quality assessment indicators, distinguishing between direct and indirect correlations. The quality assessment indicators are treated as top-level events, and visual parameter factors and evaluation descriptive factors are decomposed. Event correlation factors and logical relationships are constructed for both direct and indirect correlations. The above correlation information is integrated to form a complete response mapping relationship between quality assessment indicators and visual parameters. For example, for the quality assessment indicator of surface cracks, a direct response relationship is established with the edge grayscale difference parameter extracted by visible light imaging, and an indirect response relationship is established with the temperature gradient parameter extracted by infrared thermal imaging.

[0027] In one possible implementation, a response mapping relationship between forging quality assessment indicators and visual parameters is established. Step S100 further includes step S110, which determines multi-dimensional quality assessment indicators based on the forging quality objectives, including internal and surface defects, mechanical properties, chemical structure, and dimensional and surface accuracy. Specifically, the application fields and quality standards of forgings are investigated, for example, for forgings used in engineering machinery, referring to the national standard "General Technical Conditions for Forged Steel Parts". The quality objectives are broken down according to four dimensions: internal and surface defects, mechanical properties, chemical structure, and dimensional and surface accuracy. Among them, the indicators determined by the internal and surface defect dimension include internal cracks, porosity, inclusions, surface scratches, pits, and burrs; the indicators determined by the mechanical property dimension include tensile strength, yield strength, elongation, and impact toughness; the indicators determined by the chemical structure dimension include grain size, phase composition, and carbide distribution; and the indicators determined by the dimensional and surface accuracy dimension include dimensional deviation, hole diameter deviation, flatness, and roughness. Quantitative standards are set for each indicator. For example, the qualified standard for tensile strength is greater than or equal to 800 MPa, and the qualified standard for roughness is less than or equal to 1.6 micrometers.

[0028] Step S120: Perform multi-source visual correlation analysis on the multi-dimensional quality assessment indicators to establish correlation relationships. These relationships include direct and indirect correlations. The multi-source visual technologies include visible light imaging, three-dimensional topography scanning, infrared thermal imaging, high-speed imaging, and microscopic imaging. Specifically, analyze the correlation between each quality assessment indicator and the multi-source visual technology to determine which indicators can be directly detected by visual technology and which need to be indirectly derived through visual technology. Construct a correlation analysis matrix between multi-source visual technologies and quality assessment indicators. Rows in the matrix represent multi-dimensional quality assessment indicators, and columns represent multi-source visual technologies. For each combination of indicator and visual technology, use a correlation analysis algorithm to calculate the correlation degree. For example, use the Pearson correlation coefficient to calculate the correlation coefficient between the surface grayscale features of visible light imaging and the surface scratch indicator. If the absolute value of the correlation coefficient is greater than 0.8, it is considered a direct correlation; if the absolute value of the correlation coefficient is between 0.3 and 0.8, it is considered an indirect correlation. The established direct correlations include: visible light imaging and surface defects and dimensional deviation indicators; three-dimensional morphology scanning and dimensional and surface accuracy indicators; and microscopic imaging and internal defects and chemical structure indicators. The indirect correlations include: infrared thermal imaging and mechanical performance indicators; and high-speed imaging and internal crack propagation trend indicators.

[0029] Step S130: Based on the aforementioned correlation, visual factor correlation analysis is performed using the quality assessment index as the top event to establish event-related factors and logical relationships. Specifically, the quality assessment index is considered as the top event, and the related visual parameter factors are analyzed to determine the logical correspondence between each factor. For direct correlations, the visual source parameters corresponding to the quality assessment index are decomposed to obtain visual parameter factors. Simultaneously, the quality assessment index is decomposed into evaluation descriptive factors. Through correlation analysis, visual parameter factors reaching the threshold are identified as event-related factors, and their correspondence is considered as logical relationships. For indirect correlations, the quality assessment index is decomposed in reverse causal order, and the causal factors causing the index to fail to meet the standards are analyzed layer by layer until they are decomposed to factors observable by visual technology. A visual factor correlation tree is constructed, and then event-related factors and logical relationships are determined based on the correlation tree.

[0030] Step S140: Based on the top event, event correlation factors, and logical relationships, establish a response mapping relationship between the forging quality assessment index and visual parameters. Specifically, integrate the analysis results of step S130 to form a standardized and executable response mapping relationship between quality assessment index and visual parameters. A tree structure is used to construct the response mapping relationship model, with the root node representing the quality assessment index, intermediate nodes representing event correlation factors, and leaf nodes representing visual parameters. For direct response relationships, establish a direct mapping path between the root node and leaf nodes in the model, and label the logical relationship as follows: when the visual parameter factor meets a certain threshold, the quality assessment index meets the standard. For example, when the width of the surface scratch extracted by visible light imaging is less than 0.1 mm, the surface defect index meets the standard. For indirect response relationships, establish a multi-level mapping path from the root node to the leaf nodes in the model, and label the logical relationship at each level as follows: cause factor A leads to quality assessment index B, and visual parameter factor C characterizes cause factor A. For example, coarse grains lead to insufficient tensile strength, and the grain size parameter extracted by microscopic imaging characterizes the degree of grain coarsening. The mapping relationship model is transformed into a rule base that can be recognized by computers for quality assessment calculations.

[0031] In one possible implementation, a quality assessment index is used as the top event for visual factor correlation analysis to establish event correlation factors and logical relationships. Step S130 further includes step S131, whereby for directly correlated relationships, the correlated visual sources are decomposed to obtain visual parameter factors. Specifically, for each directly correlated visual source, parameter decomposition is performed according to its detection principle and output data type. For example, the visual parameter factors decomposed by visible light imaging include edge grayscale difference, defect area ratio, scratch length, and pit depth; the visual parameter factors decomposed by three-dimensional topography scanning include dimensional deviation, aperture deviation, flatness error, and roughness value; and the visual parameter factors decomposed by microscopic imaging include grain size, number of inclusions, and crack width. Each visual parameter factor is quantitatively defined; for example, edge grayscale difference is the difference in grayscale value between the defect area and the normal area, and defect area ratio is the ratio of the defect area to the total area of ​​the detection area.

[0032] Step S132: The quality assessment index is taken as the top event, and descriptive factors are analyzed for the top event to obtain evaluation descriptive factors. Specifically, for each quality assessment index as the top event, it is decomposed according to its quantitative standard. For example, the evaluation descriptive factors for internal and surface defect indices include whether cracks exist, whether the number of inclusions exceeds the standard, and whether the scratch length is compliant; the evaluation descriptive factors for mechanical performance indices include whether the tensile strength and yield strength meet the standard; the evaluation descriptive factors for dimensional and surface accuracy indices include whether the dimensional deviation is within the allowable range and whether the roughness meets the requirements. A judgment threshold is set for each evaluation descriptive factor. For example, a scratch length of less than 0.5 mm is an evaluation descriptive factor and judgment standard for surface defects.

[0033] Step S133: Using the evaluation descriptive factor as the hierarchical target, perform correlation analysis on the visual parameter factors to identify the response relationship between the visual parameter factors and the evaluation descriptive factor. Based on the response relationship, determine the event correlation factor. The event correlation factor is the visual parameter factor that reaches the correlation threshold, and its response relationship is used as the logical relationship. Specifically, establish the correspondence between the evaluation descriptive factor and the visual parameter factor, screen out the visual parameter factors that are highly correlated with the evaluation descriptive factor as event correlation factors, and determine the response logic between the two. Construct a correlation analysis model between the evaluation descriptive factor and the visual parameter factor, and use the support vector machine algorithm to train the historical data of both to determine the correlation threshold, for example, 0.85. Input the visual parameter factor into the trained model and calculate its correlation degree with the evaluation descriptive factor. If the correlation degree is greater than or equal to the correlation threshold, then the visual parameter factor is determined as an event correlation factor. At the same time, determine the response relationship as the logical relationship. For example, the logical relationship between the evaluation descriptive factor "scratch length less than 0.5 mm" and the visual parameter factor "scratch length parameter extracted by visible light imaging" is: when the scratch length parameter is less than 0.5 mm, the evaluation descriptive factor is valid.

[0034] Step S134: For indirect correlations, perform reverse causal correlation analysis on the top event, analyze the direct causal factors leading to the top event, and decompose the causal factors layer by layer until the correlation reaches visually measurable factors that can be directly or indirectly observed by at least one visual source, thus constructing a visual factor correlation tree. Specifically, for the quality assessment index of indirect correlations, perform reverse causal decomposition, analyze the causes layer by layer to find factors that can be detected by visual technology, and construct a correlation tree structure. Using the quality assessment index as the top event, use the fault tree analysis method to first analyze the direct causal factors leading to the top event, and then continue to decompose each causal factor downwards until the decomposed factors can be directly or indirectly observed by multi-source visual technology. Construct a visual factor correlation tree according to the causal relationship of the decomposed factors, with the top layer being the top event, the middle layer being the causal factors at each level, and the bottom layer being visually measurable factors.

[0035] Step S135: Based on the visual factor association tree, establish a structured response mapping path for the top event, identify event association factors, and establish logical relationships according to the structured response mapping path. Specifically, based on the visual factor association tree, determine the response mapping path and logical relationships under indirect association relationships. From the top event at the top level of the visual factor association tree to the bottom visually measurable factors, sort out each complete causal path as a structured response mapping path. Identify the visually measurable factors in the path as event association factors, and determine the logical relationship of each path as follows: visually measurable factor X affects causal factor Y, causal factor Y affects top event Z. For example, the visually measurable factor - temperature gradient parameter extracted by infrared thermal imaging affects causal factor - coarse microstructure grains, and coarse microstructure grains affect top event - insufficient tensile strength. The corresponding logical relationship is: when the temperature gradient parameter is greater than a certain degree Celsius per millimeter, the probability of grain coarsening increases, which in turn leads to insufficient tensile strength.

[0036] In one possible implementation, a reverse causal relationship analysis is performed on the top event to analyze the direct causal factors leading to its occurrence. These causal factors are then decomposed layer by layer until a visually measurable factor that can be directly or indirectly observed by at least one visual source is identified, constructing a visual factor association tree. Step S134 further includes step S1341, analyzing the direct causal factors leading to the top event, including abnormal microstructure and deviations in processing parameters. Specifically, the direct causes of indirectly related quality assessment indicators are determined, and the core causal factors leading to non-compliance are identified. Using a causal analysis method, combined with the forging process and material characteristics, the direct causal factors leading to the top event are analyzed. Among these, abnormal microstructure factors include coarse grains, phase composition imbalance, and uneven carbide distribution; deviations in processing parameters include excessively high or low forging temperatures, insufficient forging pressure, and insufficient holding time. By reviewing the forging process database and quality inspection reports, the correlation between the causal factors and the top event was verified. For example, statistical data showed that 80% of forgings with substandard tensile strength had coarse grains in their microstructure.

[0037] Step S1342 involves decomposing the direct cause factors downwards to analyze the visible changes in the microstructure or the processing, including static and dynamic measurable factors. The dynamic measurable factors are obtained by analyzing time-series image data, reflecting the rate of change of physical parameters or the structural evolution of the forging during processing. The static measurable factors are obtained by analyzing image data characterizing the spatial morphology or microstructure of the forging at different times. Specifically, the direct cause factors are further decomposed to analyze their corresponding visible changes, distinguishing between static and dynamic measurable factors. For microstructure abnormality-related cause factors, their static and dynamic measurable factors are decomposed. Static measurable factors include grain size and phase composition ratio observed by microscopic imaging, while dynamic measurable factors include grain growth rate and phase transformation rate observed by high-speed imaging. For factors causing deviations in processing parameters, their static and dynamic measurable factors are decomposed. Static measurable factors include the deformation of the forging part observed by three-dimensional morphological scanning, while dynamic measurable factors include the temperature change rate during the forging process observed by infrared thermography. Static measurable factors are extracted from single-acquired image data, such as grain size parameters extracted from a single image of microscopic imaging, while dynamic measurable factors are extracted from continuously acquired time-series image data, such as temperature change rate parameters extracted from consecutive frames of infrared thermography.

[0038] Step S1343: Based on the results of the hierarchical recursive decomposition, a visual factor association tree is constructed with quality assessment indicators as the root node and visually measurable factors as the leaf nodes. Specifically, the factors decomposed layer by layer are constructed into a tree structure according to hierarchical relationships, presenting the causal relationship between quality assessment indicators and visually measurable factors. Using a tree structure modeling tool, the quality assessment indicator is set as the root node, the direct cause factor is set as the first-level child node, the next-level decomposition factor of the cause factor is set as the second-level child node, and so on, with visually measurable factors set as leaf nodes. Causal relationships are labeled between each node. For example, the causal relationship between the root node - substandard tensile strength and the first-level child node - coarse grains is: coarse grains lead to substandard tensile strength; the causal relationship between the first-level child node - coarse grains and the second-level child node - abnormal temperature gradient is: abnormal temperature gradient leads to coarse grains; the causal relationship between the second-level child node - abnormal temperature gradient and the leaf node - infrared thermal imaging temperature gradient parameter is: this parameter characterizes the degree of temperature gradient anomaly. Finally, a visualized visual factor association tree is generated for establishing response mapping relationships.

[0039] Step S200: Based on the visual parameters, call the multi-source image acquisition device to obtain the multi-source visual detection dataset, and extract visual parameter features from the multi-source visual detection dataset.

[0040] Specifically, based on the visual parameters involved in the response mapping relationship, corresponding multi-source image acquisition devices are matched in advance, including visible light cameras, 3D topography scanners, infrared thermal imagers, high-speed cameras, and microscopes. Acquisition parameters are set for different devices; for example, the resolution of the visible light camera is set to 24 megapixels, the exposure time to 50 microseconds, the scanning accuracy of the 3D topography scanner to 5 micrometers, and the temperature measurement range of the infrared thermal imager to -20 degrees Celsius to 300 degrees Celsius. The acquired image data is preprocessed, including denoising, enhancement, and registration. Denoising uses a median filtering algorithm, enhancement uses a histogram equalization algorithm, and registration uses a feature point-based SIFT algorithm. Then, visual parameter features are extracted using feature extraction algorithms. For example, the Canny edge detection algorithm is used to extract surface defect edge features from the visible light image, the point cloud registration algorithm is used to extract dimensional deviation features from the 3D topography scan data, and the gray-level gradient algorithm is used to extract temperature distribution features from the infrared thermal imaging data.

[0041] Step S300: The extracted visual feature vectors are evaluated according to the response mapping relationship to obtain the direct response quality evaluation index and the indirect response quality evaluation index.

[0042] Specifically, using the established response mapping relationship, the extracted visual feature vectors are analyzed and calculated to obtain evaluation results directly corresponding to and indirectly corresponding to the quality assessment indicators. First, the extracted visual feature vectors undergo metadata parsing and type labeling, identifying geometric features, surface defect features, speckle sequence features, infrared thermal imaging sequence features, and thermal extraction features. Then, according to the labeled feature type, the feature vectors are imported into the corresponding evaluation path. For geometric features and surface defect features, the direct response relationship evaluation path is used, employing a template matching algorithm to compare the feature vectors with a preset quality standard template. Based on the similarity, the direct response quality assessment indicator is output. For example, comparing the extracted forging part aperture size feature with a standard aperture size template outputs an evaluation result indicating whether the aperture size is qualified or unqualified. For speckle sequence features, infrared thermal imaging sequence features, and thermal extraction features, the indirect response relationship evaluation path is used. A hierarchical association analysis algorithm is employed to extract visual association factors, and the evaluation results are calculated layer by layer according to the logical relationship in the response mapping relationship. Simultaneously, a weighted average algorithm is used to aggregate the evaluation confidence of each level, finally outputting the indirect response quality assessment indicator and labeling the response confidence probability.

[0043] In one possible implementation, the extracted visual feature vectors are evaluated according to the response mapping relationship to obtain direct response quality evaluation indicators and indirect response quality evaluation indicators. Step S300 further includes step S310, which involves parsing and labeling the extracted visual feature vectors with metadata, including geometric features, surface defect features, speckle sequence features, infrared thermal imaging sequence features, and thermal extraction features. Specifically, the visual feature vectors are parsed with metadata to extract metadata information such as the source device, acquisition time, and parameter dimensions. Feature type determination rules were established, and machine learning classification algorithms were used to label the feature vectors. The geometric feature labeling rules were as follows: feature vectors representing the shape and dimensions of the forging, such as length, width, and aperture parameters; surface defect feature labeling rules were as follows: feature vectors representing scratches and pits on the surface of the forging, such as defect area and scratch length parameters; speckle sequence feature labeling rules were as follows: speckle change sequence vectors on the surface of the forging acquired by high-speed imaging; infrared thermal imaging sequence feature labeling rules were as follows: temperature change sequence vectors acquired by an infrared thermal imager; and thermal extraction feature labeling rules were as follows: temperature peak and temperature gradient parameter vectors extracted from infrared thermal imaging data. Finally, type labels were added to each feature vector to generate a labeled feature dataset.

[0044] Step S320: According to metadata parsing and type labeling, the visual feature vectors are imported into the evaluation path corresponding to the response mapping relationship for response evaluation. Specifically, feature vectors labeled as geometric features or surface defect features are imported into the direct response relationship evaluation path for image recognition processing. The direct response quality evaluation index is output according to the direct evaluation relationship between the image recognition result and the quality evaluation index. Vectors labeled as speckle sequence features, infrared thermal imaging sequence features, or thermal extraction features are imported into the indirect response relationship evaluation path for visual association factor extraction. Based on the layer-by-layer association relationship of the visual association factors, the evaluation results and confidence scores are aggregated to obtain the indirect response quality evaluation index and label the response confidence probability. Specifically, for vectors labeled as geometric features or surface defect features, the direct response evaluation path is used for processing using a deep learning-based image recognition model. This model includes convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract the spatial dimension information of the features, the pooling layers are used for dimensionality reduction, and the fully connected layers are used for classification. The feature vector is input into the model and compared with a preset quality standard feature template. The matching degree is output. If the matching degree is greater than the preset matching degree threshold, such as 90%, the corresponding quality assessment indicator is considered to meet the standard; otherwise, it is considered not to meet the standard. For example, the aperture geometric feature vector is input into the model and compared with the standard aperture feature template to output a direct response indicator that the aperture size is qualified. For vectors labeled as speckle sequence features, infrared thermal imaging sequence features, or thermal extraction features, an indirect response assessment path is imported, and an association rule mining algorithm is used to extract visual association factors. For example, the temperature change rate association factor is extracted from infrared thermal imaging sequence features. Then, following the hierarchical path of the response mapping relationship, the Bayesian network algorithm is used to calculate the correlation probability of each level factor layer by layer. The weighted summation algorithm is used to aggregate the confidence of each level to obtain the final indirect response quality assessment index and response confidence probability. For example, by calculating that the correlation probability between the temperature change rate correlation factor and the grain coarsening factor is 92%, and the correlation probability between the grain coarsening factor and the tensile strength failure factor is 89%, the confidence probability of the indirect response index of tensile strength failure is obtained after aggregation is 91%.

[0045] Step S400: Generate a quality assessment result based on the response confidence probability of the direct response quality assessment index and the indirect response quality assessment index.

[0046] Specifically, quality assessment rules are pre-defined. For direct response quality assessment indicators, a veto system is adopted; if any direct response indicator is deemed unqualified, the forging is initially judged to be unqualified. For indirect response quality assessment indicators, a confidence probability threshold is set, for example, 90%. If the confidence probability of an indirect response indicator is higher than this threshold, its assessment result is included in the comprehensive judgment; if it is lower than this threshold, the assessment result of that indicator is ignored. The judgment results of direct response indicators and effective indirect response indicators are integrated, and a weighted voting algorithm is used to calculate the comprehensive quality score. For example, the weight of direct response indicators is set to 0.6, and the weight of indirect response indicators is set to 0.4. Based on the comprehensive score, the quality of the forging is divided into four levels: excellent, good, qualified, and unqualified. Finally, a quality assessment report is generated, including the level determination, details of unqualified indicators, and confidence probabilities.

[0047] This application employs techniques such as establishing direct and indirect response mapping relationships between forging quality assessment indicators and visual parameters, using multi-source image acquisition equipment to obtain detection datasets and extract visual feature vectors based on visual parameters, evaluating the feature vectors according to the aforementioned response mapping relationships to obtain direct and indirect response quality assessment indicators, and combining the response confidence probabilities of the direct and indirect response indicators to generate forging quality assessment results. These techniques solve the technical problems of large subjective errors, destructive and long detection cycles, and isolated and difficult-to-correlate quality data in existing forging quality assessments, achieving the technical effects of reducing subjective errors, non-destructive and rapid detection, and integrated analysis of quality data.

[0048] In the above text, refer to Figure 1 A method for evaluating the quality of forged parts based on industrial vision, according to embodiments of the present invention, is described in detail. Next, reference will be made to... Figure 2 A vision-based quality assessment system for forgings according to an embodiment of the present invention is described.

[0049] The forging quality assessment system based on industrial vision according to embodiments of the present invention addresses the technical problems of large subjective errors, destructive and long inspection cycles, and isolated and difficult-to-correlate quality data in existing forging quality assessment methods. It achieves the technical effects of reducing subjective errors, enabling non-destructive and rapid inspection, and providing integrated analysis of quality data. The forging quality assessment system based on industrial vision includes: a response mapping relationship establishment module 10, a visual parameter feature extraction module 20, a response assessment module 30, and a quality assessment result generation module 40.

[0050] The response mapping relationship establishment module 10 is used to establish a response mapping relationship between the forging quality assessment indicators and visual parameters, including direct response relationship and indirect response relationship; the visual parameter feature extraction module 20 is used to call a multi-source image acquisition device based on visual parameters to obtain a multi-source visual detection dataset, and to extract visual parameter features from the multi-source visual detection dataset; the response evaluation module 30 is used to evaluate the response of the extracted visual feature vectors to the forging quality assessment indicators according to the response mapping relationship, and obtain direct response quality assessment indicators and indirect response quality assessment indicators; the quality assessment result generation module 40 is used to generate a quality assessment result based on the response confidence probability of the direct response quality assessment indicators and the indirect response quality assessment indicators.

[0051] The specific configuration of the response mapping relationship establishment module 10 is described in detail below: As mentioned above, the response mapping relationship between the forging quality assessment index and visual parameters is established. The response mapping relationship establishment module 10 may further include: a multi-dimensional quality assessment index determination unit for determining multi-dimensional quality assessment indexes based on the forging quality target, including internal and surface defects, mechanical properties, chemical structure, and dimensional and surface accuracy; a multi-source visual correlation analysis unit for performing multi-source visual correlation analysis on the multi-dimensional quality assessment indexes and establishing correlation relationships, including direct correlation and indirect correlation; a visual factor correlation analysis unit for performing visual factor correlation analysis based on the correlation relationships, using the quality assessment index as the top event, and establishing event correlation factors and logical relationships; and a response mapping relationship establishment unit for establishing the response mapping relationship between the forging quality assessment index and visual parameters according to the top event, event correlation factors, and logical relationships.

[0052] The multi-source visual correlation analysis unit may further include: the multi-source visuals include: visible light imaging, three-dimensional topography scanning, infrared thermal imaging, high-speed imaging, and microscopic imaging.

[0053] The process involves using quality assessment indicators as top-level events for visual factor correlation analysis, establishing event correlation factors and logical relationships. The visual factor correlation analysis unit may further include: a parameter decomposition subunit for performing parameter decomposition on the correlated visual sources for direct correlation relationships to obtain visual parameter factors; a descriptive factor analysis subunit for using the quality assessment indicators as top-level events to perform descriptive factor analysis on the top-level events to obtain evaluation descriptive factors; and a correlation analysis subunit for using the evaluation descriptive factors as hierarchical targets to perform correlation analysis on the visual parameter factors, identifying the response relationship between the visual parameter factors and the evaluation descriptive factors, and determining event correlation factors based on the response relationship. The event correlation factors are visual parameter factors that reach a correlation threshold, and their response relationships are used as the logical relationships.

[0054] The process involves using quality assessment indicators as the top event for visual factor correlation analysis, establishing event correlation factors and logical relationships. The visual factor correlation analysis unit may further include: a reverse causal correlation analysis subunit for performing reverse causal correlation analysis on the top event for indirect correlation relationships, analyzing the direct cause factors leading to the occurrence of the top event, and decomposing the cause factors layer by layer downwards until the correlation reaches visually measurable factors that can be directly or indirectly observed by at least one visual source, thus constructing a visual factor correlation tree; and a structured response mapping relationship path establishment subunit for establishing a structured response mapping relationship path for the top event based on the visual factor correlation tree, identifying event correlation factors, and establishing logical relationships according to the structured response mapping relationship path.

[0055] The process includes: performing reverse causal correlation analysis on the top event, analyzing the direct causal factors leading to the top event, and decomposing the causal factors layer by layer until a visually measurable factor that can be directly or indirectly observed by at least one visual source is identified, thus constructing a visual factor correlation tree. The reverse causal correlation analysis subunit may further include: a direct causal factor analysis component for analyzing the direct causal factors leading to the top event, including microscopic organizational state abnormalities and deviations in processing parameters; a visual state change feature analysis component for decomposing the direct causal factors downwards and analyzing the visual state change features of microscopic organizational state abnormalities or the visual state change features of processing, including static measurable factors and dynamic measurable factors; and a visual factor correlation tree construction component for constructing a visual factor correlation tree with quality assessment indicators as the root node and visually measurable factors as leaf nodes based on the results of the layer-by-layer recursive decomposition.

[0056] The visible state change feature analysis component may further include: the dynamic measurable factor is obtained by analyzing time-series image data, reflecting the parameter change rate or structural evolution process of physical quantities of the forging during processing; the static measurable factor is obtained by analyzing image data characterizing the spatial morphology or microstructure of the forging at different times.

[0057] The specific configuration of the response evaluation module 30 is described in detail below: As mentioned above, the extracted visual feature vectors are evaluated according to the response mapping relationship to obtain the response evaluation index of the forging part quality evaluation index, and the direct response quality evaluation index and the indirect response quality evaluation index are obtained. The response evaluation module 30 may further include: a metadata parsing unit for parsing and labeling the extracted visual feature vectors with metadata, including geometric features, surface defect features, speckle sequence features, infrared thermal imaging sequence features, and thermal extraction features; a response evaluation unit for importing the visual feature vectors into the evaluation path corresponding to the response mapping relationship for response evaluation according to the metadata parsing and type labeling, wherein the feature vectors labeled as geometric features or surface defect features are imported into the direct response relationship evaluation path for image recognition processing, and the direct response quality evaluation index is output according to the direct evaluation relationship between the image recognition result and the quality evaluation index; the feature vectors labeled as speckle sequence features, infrared thermal imaging sequence features, or thermal extraction features are imported into the indirect response relationship evaluation path for visual correlation factor extraction, and the evaluation results and confidence scores are aggregated based on the hierarchical correlation relationship of the visual correlation factors to obtain the indirect response quality evaluation index and label the response confidence probability.

[0058] The industrial vision-based forging quality assessment system provided in this embodiment of the invention can execute the industrial vision-based forging quality assessment method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0059] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0060] Based on the foregoing embodiments, this application also provides an electronic device. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 3The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention. This electronic device is in the form of a general-purpose computing device, and its components may include, but are not limited to, an input device 301, a processor 302, a memory 303, and an output device 304. The processor 302 may be one or more; the memory 303 may include a computer-readable medium and at least one program product having a set (at least one) of program modules configured to perform the functions of the embodiments of this application.

[0061] The memory 303 shown in this embodiment of the invention can be any combination of one or more computer-readable media. The computer-readable storage medium can be, but is not limited to, infrared, semiconductor systems, devices or components, or any combination thereof, used to store software programs, computer-executable programs and modules, such as the program instructions / modules corresponding to the industrial vision-based forging quality assessment method in this embodiment of the invention. The processor 302 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 303, thereby realizing the above-mentioned industrial vision-based forging quality assessment method.

[0062] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for quality assessment of forged parts based on industrial vision, characterized in that, include: Establish the response mapping relationship between forging quality assessment indicators and visual parameters, including direct response relationship and indirect response relationship; Based on visual parameters, a multi-source image acquisition device is invoked to obtain a multi-source visual detection dataset, and visual parameter features are extracted from the multi-source visual detection dataset. The extracted visual feature vectors are evaluated according to the response mapping relationship to obtain the response evaluation index of the forging part quality evaluation index, and the direct response quality evaluation index and the indirect response quality evaluation index are obtained. The quality assessment results are generated based on the response confidence probabilities of the direct response quality assessment indicators and the indirect response quality assessment indicators.

2. The method for evaluating the quality of forged parts based on industrial vision according to claim 1, characterized in that, Establish a response mapping relationship between forging quality assessment indicators and visual parameters, including: Based on the quality objectives of forgings, multi-dimensional quality assessment indicators are determined, including internal and surface defects, mechanical properties, chemical structure, and dimensional and surface accuracy. Multi-source visual correlation analysis is performed on the multi-dimensional quality assessment indicators to establish correlation relationships, including direct and indirect correlations. Based on the aforementioned correlation, visual factor correlation analysis is performed using quality assessment indicators as top events to establish event correlation factors and logical relationships. Based on the top events, event correlation factors, and logical relationships, a response mapping relationship between the forging quality assessment indicators and visual parameters is established.

3. The method for evaluating the quality of forged parts based on industrial vision according to claim 2, characterized in that, The multi-source vision includes: visible light imaging, three-dimensional topography scanning, infrared thermal imaging, high-speed imaging, and microscopic imaging.

4. The method for evaluating the quality of forged parts based on industrial vision according to claim 3, characterized in that, Using quality assessment indicators as the top event, visual factor correlation analysis is performed to establish event correlation factors and logical relationships, including: For direct relationships, the associated visual sources are decomposed into parameters to obtain visual parameter factors. The quality assessment index is taken as the top event, and the top event is analyzed by descriptive factor analysis to obtain the evaluation descriptive factor; Using the evaluation descriptive factor as the hierarchical target, correlation analysis is performed on the visual parameter factor to identify the response relationship between the visual parameter factor and the evaluation descriptive factor. Based on the response relationship, event correlation factors are determined. The event correlation factors are visual parameter factors that have reached the correlation threshold, and their response relationship is used as the logical relationship.

5. The method for evaluating the quality of forged parts based on industrial vision according to claim 3, characterized in that, Using quality assessment indicators as the top event, visual factor correlation analysis is performed to establish event correlation factors and logical relationships, including: For indirect correlations, reverse causal correlation analysis is performed on the top event to analyze the direct cause factors that lead to the occurrence of the top event. The cause factors are decomposed layer by layer until the correlation is reached by visually measurable factors that can be directly or indirectly observed by at least one visual source, and a visual factor correlation tree is constructed. Based on the visual factor association tree, a structured response mapping path for the top event is established, event association factors are identified, and logical relationships are established according to the structured response mapping path.

6. The method for evaluating the quality of forged parts based on industrial vision according to claim 5, characterized in that, A reverse causal relationship analysis is performed on the top event to analyze the direct causal factors leading to its occurrence. These causal factors are then decomposed layer by layer downwards until a visually measurable factor that can be directly or indirectly observed from at least one visual source is identified. A visual factor association tree is then constructed, including: The direct causal factors leading to the top event were analyzed, including abnormal microstructure and deviations in processing parameters. The direct cause factors are decomposed downwards to analyze the visible state change characteristics of abnormal micro-organism state or the visible state change characteristics of the processing process, including static measurable factors and dynamic measurable factors. Based on the results of the hierarchical recursive decomposition, a visual factor association tree is constructed with quality assessment indicators as the root node and visually measurable factors as the leaf nodes.

7. The method for evaluating the quality of forged parts based on industrial vision according to claim 6, characterized in that, The dynamic measurable factor is obtained by analyzing time-series image data and reflects the rate of change of physical parameters or the structural evolution process of the forging during the processing. The static measurable factor is obtained by analyzing image data that characterizes the spatial morphology or microstructure of the forging at different times.

8. The method for evaluating the quality of forged parts based on industrial vision according to claim 5, characterized in that, The extracted visual feature vectors are used to evaluate the response of forging quality indicators according to the aforementioned response mapping relationship, resulting in direct response quality evaluation indicators and indirect response quality evaluation indicators, including: The extracted visual feature vectors are parsed and labeled with metadata, including geometric features, surface defect features, speckle sequence features, infrared thermal imaging sequence features, and thermal extraction features. Based on metadata parsing and type labeling, visual feature vectors are imported into the evaluation path corresponding to the response mapping relationship for response evaluation. Feature vectors labeled as geometric features or surface defect features are imported into the direct response relationship evaluation path for image recognition processing. The direct response quality evaluation index is output according to the direct evaluation relationship between the image recognition result and the quality evaluation index. Feature vectors labeled as speckle sequence features, infrared thermal imaging sequence features, or thermal extraction features are imported into the indirect response relationship evaluation path for visual correlation factor extraction. The evaluation results and confidence scores are aggregated based on the hierarchical correlation relationship of the visual correlation factors to obtain the indirect response quality evaluation index and label the response confidence probability.

9. A forging quality assessment system based on industrial vision, characterized in that, The system is used to implement the industrial vision-based forging quality assessment method according to any one of claims 1-8, and the system comprises: The response mapping relationship establishment module is used to establish the response mapping relationship between the quality evaluation index of forgings and visual parameters, including direct response relationship and indirect response relationship; The visual parameter feature extraction module is used to call a multi-source image acquisition device based on visual parameters, obtain a multi-source visual detection dataset, and extract visual parameter features from the multi-source visual detection dataset. The response evaluation module is used to evaluate the extracted visual feature vectors according to the response mapping relationship to obtain the direct response quality evaluation index and the indirect response quality evaluation index. The quality assessment result generation module is used to generate quality assessment results based on the response confidence probabilities of the direct response quality assessment indicators and the indirect response quality assessment indicators.

10. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the industrial vision-based forging quality assessment method according to any one of claims 1 to 8.

Images