A valve body internal tap face and appearance detection system and method

Abstract

The application discloses a valve body internal tap surface and appearance detection system and method, aiming at realizing accurate detection of valve body defects and process optimization. The system comprises a detection module, a comparison module and an analysis control module; the detection module collects images of the valve body outer surface, the conical surface and the thread surface through multiple stations, the comparison module stores a standard digital model, the analysis unit of the analysis control module compares the images with the standard model and analyzes defects in association, the prediction unit predicts defect trends based on data, and the control unit dynamically adjusts the detection strategy. The application can improve detection efficiency and reliability, and is suitable for valve body batch detection scenes.

Description

Technical Field

[0001] This invention relates to the field of machine vision inspection technology, and in particular to a system and method for inspecting the internal taper surface and appearance of a valve body. Background Technology

[0002] In the precision manufacturing process of automotive control valve bodies, defect detection of internal thread tapered surfaces, various sealing tapered surfaces such as 84° and 120° tapered surfaces, and complex appearances is a core step in ensuring product sealing performance, operational reliability, and service life. Currently, quality inspection in this field faces multiple technical challenges. Traditional inspection methods mainly rely on manual visual inspection or sampling inspection using simple measuring tools. These methods are not only inefficient and labor-intensive, but also prone to significant subjective judgment bias due to differences in personnel experience. Manual inspection struggles to reliably identify defects such as micron-level scratches, fine burrs, micro-corrosion, and uneven surface textures, and cannot quantitatively assess thread geometric accuracy, leading to frequent missed and false detections. To cover various characteristic areas of the valve body, such as end faces, tapered surfaces, flange faces, side walls, and deep hole threads, existing technologies generally employ multiple independent inspection devices or stations for segmented inspection. However, the lack of data sharing and system integration between inspection units creates information silos. This fragmented inspection process cannot achieve full-process, full-feature closed-loop quality tracking for a single valve body, resulting in fragmented inspection data. Crucially, the judgment results from different workstations are isolated, making it impossible to effectively analyze the symbiotic relationships and causal chains between processing features or defects in adjacent areas. For example, a processing anomaly at one point may trigger a chain reaction between conical step defects and thread burrs, but existing systems cannot comprehensively assess the overall quality by integrating such process correlations, resulting in insufficient reliability of the final judgment. For special structures such as deep-hole threads and non-perpendicular sealing conical surfaces inside valve bodies, conventional machine vision systems struggle to acquire clear images due to physical constraints such as limited space, obstructed viewpoints, and surface reflections, severely limiting algorithm robustness. Existing solutions are mostly customized for single defect types, resulting in weak system adaptability. They cannot dynamically adjust detection strategies to cope with complex correlated defect patterns, nor can they intelligently optimize the detection focus and judgment thresholds of downstream workstations using upstream workstation detection results. Therefore, the industry urgently needs a new detection system that can integrate multi-workstation detection capabilities, break down information silos, and achieve intelligent data correlation and dynamic strategy optimization to achieve high-quality, high-reliability, and fully feature-covered automated defect detection of automotive control valve bodies. Summary of the Invention

[0003] The purpose of this application is to provide a valve body internal taper surface and appearance inspection system and method, which has the advantages of improving inspection efficiency, reducing subjective bias, realizing defect quantitative assessment and intelligent correlation analysis, breaking down information silos, and improving inspection reliability and accuracy.

[0004] To solve the above-mentioned technical problems, the present invention provides a valve body taper surface and appearance inspection system, comprising: The detection module includes multiple detection stations arranged sequentially along the conveying path, used to acquire images of several feature areas of the valve body, wherein the feature areas include at least the outer surface of the valve body, the conical surface of the valve body, and the threaded surface. The comparison module stores a standard digital model containing standard geometric data and reference texture data for each feature region of the valve body; The analysis and control module is communicatively connected to the detection module and the comparison module. The analysis and control module includes an analysis unit and a control unit; The analysis unit is used to compare the acquired real-time images with the standard digital model, identify an initial set of defect features including defect location, type and degree, and perform logical association analysis on defects identified in different inspection stations for the same valve body based on predefined process association rules. The process association rules are established based on the symbiotic or causal relationship between defects in several processing features or adjacent areas, and generate analysis results including association determination. The control unit is used to dynamically adjust the detection strategy of at least one subsequent detection station based on the analysis results; the dynamic adjustment includes adjusting image acquisition parameters, image processing algorithm parameters, or triggering a re-inspection process for defect-related areas.

[0005] By adopting the above technical solutions, the system integrates multiple inspection stations and establishes process association rules to achieve comprehensive automated inspection of various characteristic areas of the valve body. It can not only identify individual defects, but also intelligently associate defects detected at different stations through logical association analysis, thereby revealing the symbiotic or causal relationship between defects and improving the accuracy and comprehensiveness of defect identification. Through the dynamic adjustment function of the analysis and control module, the system can adaptively optimize the inspection strategy of subsequent stations based on the results of previous inspections, such as adjusting the acquisition parameters or triggering re-inspection, thereby achieving flexible response to complex defect patterns and effectively reducing the rate of missed and false detections. In addition, by upgrading the inspection process from isolated station inspections to full-process closed-loop tracking, the system provides a more reliable basis for valve body quality judgment, ultimately improving inspection efficiency while significantly reducing reliance on manual labor and subjective errors.

[0006] The present invention is further configured such that: the plurality of inspection stations include at least one of the following inspection stations: A multi-defect inspection station for conical surfaces is used to inspect at least one of the following defects in 84° and 120° conical surfaces: roughness, steps, mouth defects, electroplating solution residue, tool marks, scratches, bubbles, and corrosion. The thread inspection station, including the internal thread inspection station and the bottom outer ring thread inspection station, is used to inspect the thread surface and the bottom outer thread for burrs, metal residues or machining abnormalities. The valve body appearance and structural damage inspection station includes the cone bottom inspection station, end face inspection station, flange upper and lower surface inspection station, side inspection station, step burr inspection station, and lining end face inspection station, covering the inspection of damage, scratches, pits, corrosion, deformation, foreign object structure and appearance defects of the valve body cone bottom surface, end face, flange, side wall, step, and lining end face; Does the solder have an inspection station for inspecting the solder in the weld area?

[0007] By adopting the above technical solution, the system can cover all key feature areas of the valve body by setting up multiple specialized inspection stations, including conical surfaces at different angles, internal and external thread areas, and various structural surfaces of the exterior. This enables full feature inspection from the inside to the outside. Each station performs high-precision inspection on specific types of defects, such as conical surface roughness, thread burrs, and appearance damage, ensuring that all types of defects can be effectively identified. This clearly defined inspection layout not only improves the professionalism of the system's inspection but also facilitates the adoption of differentiated imaging and processing strategies for different areas. This further enhances the system's ability to detect complex structures and microscopic defects, thereby achieving full coverage and refinement of valve body quality control.

[0008] The present invention is further configured such that each detection station is equipped with an image acquisition device and an illumination device, each image acquisition device is a color area array camera, and the illumination device is configured with one or more combinations of ring light, back light, spherical integrating light, and angled ring light according to the characteristic area to be detected.

[0009] By adopting the above technical solutions, the system is equipped with color area array cameras and various dedicated lighting devices at each inspection station. It can flexibly select lighting methods such as ring light, backlight, and spherical integrating light according to the structural characteristics of the inspection area, thereby improving the image acquisition quality. Especially in areas that are prone to reflection or occlusion, such as deep holes and conical surfaces, high-quality images provide clear and stable input for subsequent defect identification, effectively reducing misjudgments caused by uneven lighting or curved surface reflection. Color imaging can also enhance the ability to distinguish color-related defects such as rust and electroplating solution residue, thereby improving the robustness and accuracy of the system in identifying multiple types of defects and ensuring the stability and reliability of the inspection process.

[0010] The present invention is further configured such that: the process association rules include defect propagation rules and comprehensive judgment rules; The defect propagation rule is used to: when a specific type of defect or a defect located in a defect association area is identified at an earlier detection station, the detection sensitivity or judgment priority of the preset associated defect type is increased in one or more subsequent associated detection stations according to the rule. The comprehensive judgment rule is used to: integrate the identification results of the same valve body at multiple related inspection stations, perform consistency verification or weighted evaluation on defects with process correlation, and output the final defect judgment.

[0011] By adopting the above technical solutions, the system introduces defect propagation rules and comprehensive judgment rules, enabling the detection process to possess process logic reasoning capabilities. The defect propagation rules can automatically improve the detection sensitivity of related stations to related defects when a certain defect is detected, thereby providing early warning of possible associated defect types. The comprehensive judgment rules can perform consistency verification and weighted evaluation of the detection results of the same valve body at multiple stations, avoiding misjudgment of the overall quality due to misjudgment at a single station. The combined effect of these two types of rules enables the system to simulate the judgment logic of process experts, achieving intelligent identification and comprehensive diagnosis of complex defect chains, thereby significantly improving the scientificity and reliability of quality judgment.

[0012] The present invention is further configured such that: the analysis and control module further includes a prediction unit; the prediction unit is used to predict the occurrence trend of the same type of defect in the current production batch and output the risk level based on the analysis results and historical defect data through a machine learning model; the control unit is also configured to adjust the decision priority of the current valve body sorting or re-inspection process according to the risk level. The input features of the machine learning model include the historical defect data.

[0013] By adopting the above technical solution, the system, through the addition of a prediction unit and the introduction of a machine learning model, can predict the occurrence trend and risk level of similar defects based on current detection results and historical defect data. This enables the detection system to have a forward-looking quality early warning capability. The control unit can dynamically adjust the decision priority of sorting or re-inspection according to the risk level, achieving rapid response to high-risk defects and efficient handling of low-risk defects. This not only optimizes the allocation of detection resources but also provides real-time data support for adjusting production line process parameters, thereby improving detection efficiency while enhancing the preventive quality control capability of the production process.

[0014] The present invention is further configured such that: the control unit is configured to: for defects with a risk level higher than a preset threshold, trigger the marking and alarm of the processing parameters related to the defect type; for defects with a risk level lower than the preset threshold, process them according to the standard procedure.

[0015] By adopting the above technical solution, the system implements differentiated processing strategies based on defect risk levels, achieving intelligent hierarchical management of the inspection process. For high-risk defects, the system automatically triggers process parameter marking and alarm mechanisms, facilitating timely process adjustments and problem tracing, thereby preventing the generation of defects in batches. For low-risk defects, the system is processed according to standard procedures to avoid over-inspection affecting overall efficiency. This hierarchical response mechanism enables the system to maintain both inspection rigor and flexibility, adapting to the quality control needs of different production states, thereby improving inspection accuracy while ensuring stable production rhythm.

[0016] The present invention is further configured as: a method for inspecting the internal taper surface and appearance of a valve body, comprising the following steps: Images of different feature areas of the valve body are acquired by multiple inspection stations arranged sequentially along the conveying path. The feature areas include at least the outer surface, the valve body end face, the internal thread surface, and the deep hole internal taper surface. The acquired real-time images are compared with a pre-stored standard digital model, which includes standard geometric data and reference texture data for each feature region. An initial set of defect features is identified, and based on predefined process association rules, logical association analysis is performed on defects identified in different inspection stations for the same valve body. The process association rules are established based on the symbiotic or causal relationship between defects in several processing features or adjacent areas, generating analysis results that include association determination. Based on the analysis results, the detection strategy of at least one subsequent detection station is dynamically adjusted; the dynamic adjustment includes adjusting image acquisition parameters, image processing algorithm parameters, or triggering a re-inspection process for defect-related areas.

[0017] By adopting the above technical solution, this method achieves full automation from image acquisition to defect analysis by sequentially acquiring images of each feature area of ​​the valve body and comparing them with a standard digital model. Through process association rules, it performs logical association analysis on defects at multiple workstations, enabling the method to identify process connections between defects and thus provide a more comprehensive quality assessment. Based on the analysis results, it dynamically adjusts subsequent detection strategies, giving the detection process adaptive optimization capabilities. It can specifically strengthen the detection of suspicious areas or adjust algorithm parameters, thereby significantly improving the detection rate and identification accuracy of complex defects. At the same time, this method integrates discrete detection steps into a coherent intelligent detection process.

[0018] The present invention is further configured such that, after generating the analysis results, it also includes the steps of: predicting the occurrence trend of similar defects in the current production batch and assessing the risk level based on the analysis results and historical defect data using a machine learning model; and adjusting the decision priority of the current valve body sorting or re-inspection process according to the risk level.

[0019] By adopting the above technical solution, this method further introduces machine learning prediction and risk level assessment after completing defect analysis, thereby upgrading simple defect detection into an intelligent quality control process with trend prediction capabilities. By analyzing historical data and current results to predict batch defect trends, it can identify potential quality risks in advance and adjust the priority of sorting or re-inspection according to the risk level, thereby achieving optimized resource allocation and preventive intervention. This not only improves the processing efficiency of individual valve bodies, but also provides data support for the quality stability of the entire production batch, thus enabling the detection method to have the ability to continuously learn and evolve.

[0020] The present invention is further configured such that the step of dynamically adjusting the detection strategy includes: If, based on the process association rules, it is determined that the defect identified at the current workstation is associated with a subsequent workstation to be inspected, then an association warning message is sent to the subsequent workstation to be inspected. The subsequent inspection station shall perform at least one of the following operations based on the associated early warning information: enable a dedicated image analysis model for the associated defect type, adjust the defect identification threshold of the image processing algorithm, or mark the image area that needs to be reviewed.

[0021] By adopting the above technical solution, this method achieves real-time information linkage and strategy coordination between detection stations by sending related early warning information to subsequent stations after identifying defects at the current station. Subsequent stations can activate dedicated image analysis models, adjust defect identification thresholds, or mark key areas based on the early warning information, thereby automatically focusing the detection attention on possible related defects. This dynamic strategy adjustment mechanism breaks the limitation of isolated operation of stations in traditional detection, making the entire detection process an organic whole, significantly improving the detection sensitivity and response speed of related defect patterns, and further reducing the possibility of missing complex defects.

[0022] The present invention is further configured to include a model update step: collecting historical detection data containing defect features and correlation analysis results; and periodically optimizing and updating the baseline texture data of the standard digital model and the preset process correlation rules based on the historical detection data.

[0023] By adopting the above technical solution, this method enables the detection system to continuously improve itself by regularly collecting historical detection data and using it to optimize the standard model and process association rules. By periodically updating the benchmark data and judgment rules using defect characteristics and association analysis results in actual production, the system's ability to identify and adapt to new defect patterns can be gradually improved, thereby ensuring that the detection accuracy continues to improve with the increase of usage time. This model update mechanism enables the system to keep up with changes in production processes and the evolution of quality requirements, maintain detection efficiency and reliability in the long term, and provide sustainable technical support for realizing a closed loop of quality control in intelligent manufacturing.

[0024] This invention, by adopting the above technical solutions, has significant technical effects: This application provides a valve body internal taper surface and appearance inspection system and method. It achieves automated and efficient inspection by sequentially arranged inspection stations acquiring images, comparing them with standard models, logically analyzing defects, and dynamically adjusting inspection strategies. By constructing a multi-station integrated intelligent inspection system and method with process correlation analysis and dynamic strategy adjustment capabilities, it completely changes the backward situation of traditional valve body inspection, which is characterized by isolated stations, reliance on manual labor, difficulty in covering all features, and inability to correlate defects. The system improves image quality through color imaging and dedicated lighting, realizes process logic reasoning through defect propagation and comprehensive judgment rules, predicts risks and provides graded responses through machine learning, and achieves self-optimization through model updates. The various claims are progressive and complementary, jointly achieving high-precision, full-coverage, and adaptive inspection of various complex features inside and outside automotive control valve bodies. This significantly improves inspection efficiency, accuracy, and reliability, while providing a complete data loop for real-time optimization of production processes and quality traceability, demonstrating outstanding industrial application value and technological advancement. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of a valve body taper surface and appearance inspection system; Figure 2 This is a flowchart of a method for inspecting the internal taper surface and appearance of a valve body. Detailed Implementation

[0026] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0027] This application proposes a valve body taper surface and appearance inspection system, including: The detection module includes multiple detection stations arranged sequentially along the conveying path, used to acquire images of several feature areas of the valve body, which include at least the outer surface of the valve body, the conical surface of the valve body, and the threaded surface. The comparison module stores a standard digital model containing standard geometric data and reference texture data for each feature area of ​​the valve body. It can store a 3D model as standard geometric data and a set of high-resolution images of a defect-free valve body as reference texture data. This standard digital model can be stored in a local database as static reference data. The analysis and control module is communicatively connected to the detection module and the comparison module. The analysis and control module includes an analysis unit and a control unit. The analysis and control module can access the standard digital model stored in the comparison module. This analysis unit compares the acquired real-time images with the standard digital model to identify an initial set of defect features containing the location, type, and degree of defects. Based on predefined process association rules, it performs logical association analysis on defects identified in different inspection stations for the same valve body. These process association rules are established based on the symbiotic or causal relationship between defects in several processing features or adjacent areas, generating analysis results that include association determination. If an abnormal surface roughness is detected in the valve body at the first station, the analysis unit can determine, according to the rules, that this may be related to electroplating defects detected in subsequent stations. This control unit is used to dynamically adjust the detection strategy of at least one subsequent inspection station based on the analysis results. This dynamic adjustment includes adjusting image acquisition parameters, image processing algorithm parameters, or triggering a re-inspection process for the defect-related area. Each inspection station can be equipped with one or more image acquisition devices, such as industrial cameras, and corresponding lighting devices to ensure clear, high-quality images. As one implementation, each inspection station can use a fixed-focal-length camera with preset lighting conditions to acquire images of the defect-related area. Adjusting the image acquisition parameters of the station, such as increasing the exposure time to capture details more clearly, and adjusting the image processing algorithm parameters, such as lowering the defect recognition threshold to improve sensitivity, are all possible adjustments. Another adjustment method is the re-inspection process, such as instructing a robotic arm to return the valve body to a certain station for secondary inspection, or using different imaging modes for supplementary inspection at the current station.

[0028] The valve body taper surface and appearance inspection system of this application integrates multi-station image acquisition and standard digital model comparison. Based on predefined process association rules, the system can perform logical association analysis on defects identified at different inspection stations, thereby breaking through the limitations of traditional segmented inspection information silos.

[0029] Multiple inspection stations may include one or more of the following: a conical surface multi-defect inspection station, used to inspect at least one of the following defects in 84° and 120° conical surfaces: roughness, steps, mouth defects, electroplating solution residue, tool marks, scratches, bubbles, and corrosion; a thread inspection station, including an internal thread inspection station and a bottom outer ring thread inspection station, used to inspect burrs, metal residue, or machining abnormalities on the thread surface and bottom outer thread; a valve body appearance and structural damage inspection station, including a conical bottom inspection station, an end face inspection station, flange upper and lower surface inspection stations, a side inspection station, a step and burr inspection station, and a liner end face inspection station, covering the inspection of damage, scratches, pits, rust, deformation, foreign object structures, and appearance defects on the bottom surface, end face, flange, side wall, step, and liner end face of the valve body conical surface; and a solder presence / absence inspection station, used to inspect the solder in the weld area.

[0030] The conical surface multi-defect inspection station is specifically designed for the inspection of the valve body's conical surface area. This station is equipped with a high-resolution image acquisition device and an illumination device optimized for the geometric characteristics of the conical surface to ensure clear imaging of the 84° and 120° conical surfaces. Through sophisticated image processing algorithms, this station can identify and quantify a variety of specific defects on the conical surface. This is typically achieved by multi-angle imaging to cover the entire conical surface and by using edge detection, texture analysis, and color recognition technologies to accurately distinguish different types of defects.

[0031] The internal thread inspection station typically uses an endoscope or a special optical probe, combined with a high-magnification optical system and low-angle ring light or backlight illumination, to highlight the thread contour and internal details, thereby effectively detecting burrs, metal residues or machining abnormalities that may exist on the internal thread surface. The bottom outer ring thread inspection station uses an external camera and specific lighting to image and analyze the outer ring thread at the bottom of the valve body to identify similar defects. The external camera preferentially uses multiple sets of 50mm telephoto lenses.

[0032] The valve body appearance and structural damage inspection station is a comprehensive inspection unit designed to fully cover multiple key areas on the outside of the valve body. This station includes a cone bottom inspection station, an end face inspection station, flange upper and lower surface inspection stations, a side inspection station, a step burr inspection station, and a lining end face inspection station. This station can detect defects in various areas of the valve body cone surface.

[0033] The solder inspection station is specifically used to check the solder filling in the weld area of ​​the valve body. This station usually uses a high-resolution image acquisition device, combined with high-contrast lighting, to clearly identify whether the material is present in the weld area.

[0034] Each inspection station is equipped with an image acquisition device and an illumination device. Each image acquisition device uses a color area array camera. The illumination device is configured with one or more combinations of ring light, back light, spherical integrating light, and angled ring light according to the characteristic area to be inspected. In valve body inspection, color information is of great significance for distinguishing different types of defects, while the area array structure ensures inspection efficiency and complete coverage of complex shape areas.

[0035] The lighting device is configured according to the characteristic area to be detected. This means that for different parts of the valve body, such as conical surfaces, threaded surfaces, end faces, and sides, and for different types of defects such as roughness, scratches, burrs, and pits, the most suitable lighting method is selected. Ring light is usually used to provide uniform shadowless illumination, which is suitable for detecting minor defects or overall appearance in flat areas. Backlight, on the other hand, forms a clear outline image by illuminating from behind the object, and is often used to detect the size, shape, or presence of burrs, deformations, or other protrusions of the object. Integral spherical light can effectively eliminate surface reflection, making the image present a diffuse reflection effect, which is particularly suitable for detecting surface defects on highly reflective or curved surfaces. Angled ring light, through incident light at a specific angle, can highlight surface texture, small bumps or scratches, and is particularly effective for detecting directional defects such as knife marks, scratches, and steps.

[0036] The process association rules include defect propagation rules and comprehensive judgment rules. The defect propagation rules are used to: when a specific type of defect or a defect located in a specific defect association area is identified at an earlier inspection station, the detection sensitivity or judgment priority of the preset associated defect type is increased in one or more subsequent associated inspection stations according to the rule. The comprehensive judgment rules are used to: integrate the identification results of the same valve body at multiple related inspection stations, perform consistency verification or weighted evaluation on defects with process association, and output the final defect judgment.

[0037] Defect propagation rules are used to capture the patterns of defects spreading from one area or process to another in the processing flow. In practice, when a defect that meets the rule conditions is identified at an earlier inspection station, the analysis and control module will send instructions to subsequent associated inspection stations. For example, it may adjust the threshold of the image processing algorithm of the subsequent station to make it more sensitive to the features of the preset associated defects; or it may increase the priority of such defects in the defect list to ensure that operators or automated systems process or review them first.

[0038] The comprehensive judgment rule is used to summarize and finally evaluate all process-related defects identified on the same valve body after all relevant inspection stations have completed their inspections. This helps to avoid misjudgment or omission at a single station and provides a more convincing quality assessment. By refining the process association rule into defect propagation rule and comprehensive judgment rule, this system can manage and utilize the correlation information between defects more precisely. The defect propagation rule enables the system to proactively increase the attention of subsequent related inspection stations to potential related defects when a specific defect is detected early, thereby effectively avoiding the omission of defects.

[0039] The analysis and control module also includes a prediction unit. Based on the analysis results and historical defect data, the prediction unit uses a machine learning model to predict the occurrence trend of similar defects in the current production batch and output the risk level. The prediction unit integrates or calls a machine learning model. This model learns the patterns and rules in historical defect data and combines them with the current analysis results to predict the future occurrence trend of specific types of defects in the current production batch, such as a certain scratch, a certain burr, or a certain surface roughness abnormality. The prediction results are finally output in the form of a quantified risk level.

[0040] The control unit is also configured to adjust the decision priority of the current valve body sorting or re-inspection process based on the risk level. This historical defect data typically includes, but is not limited to, detailed descriptions of defects such as type, location, size, severity, occurrence time, production batch, relevant inspection station, and final quality judgment results. Based on existing real-time detection and local strategy adjustment, machine learning-based predictive capabilities are introduced. The prediction unit can combine real-time analysis results and rich historical defect data to make forward-looking predictions on the occurrence trend of similar defects in the current production batch and output a quantified risk level. This enables the control unit to no longer just make reactive adjustments based on the immediate defect information of a single valve body, but to proactively manage potential risks at the batch level.

[0041] The control unit is configured to: for defects with a risk level higher than a preset threshold, trigger the marking and alarm of the processing parameters related to the defect type; for defects with a risk level lower than the preset threshold, process them according to the standard procedure; when the risk level of a defect reaches or exceeds this threshold, it indicates that the defect has a high potential hazard and the system needs to take higher-level attention and handling measures.

[0042] Triggering the marking and alarming of machining parameters related to the defect type means that when the control unit in the analysis and control module receives defect information with a risk level higher than a preset threshold from the prediction unit, the system will automatically identify and locate upstream machining parameters that may have a causal relationship with the defect type. These parameters may include, but are not limited to, machine tool feed rate, cutting depth, coolant concentration, electroplating current, heating temperature, pressure setting, etc. Subsequently, the system will mark these identified relevant process parameters and send real-time alarm information to the production management system, operators, or quality engineers as soon as possible. The alarm methods can be diversified, such as audible and visual alarms, SMS notifications, email reminders, or displaying prominent prompts on the production control interface.

[0043] For defects with a risk level below the preset threshold, they are handled according to the standard procedure. This graded handling mechanism helps to distinguish the priority of defects and avoid unnecessary and frequent intervention in the production process.

[0044] In some embodiments described above in this application, a method for inspecting the taper surface and appearance inside a valve body is proposed. However, in its implementation, traditional inspection methods struggle to perform logical correlation analysis on defects identified at different inspection stations for the same valve body, resulting in the inability to identify composite defects caused by process correlations, thus affecting the reliability of overall quality assessment. Therefore, this application further proposes a method for inspecting the taper surface and appearance inside a valve body, including the following steps: Multiple inspection stations arranged sequentially along the conveying path acquire images of different feature areas of the valve body. These feature areas include at least the outer surface, valve body end face, internal thread surface, and deep hole internal taper surface. The acquired real-time images are compared with a pre-stored standard digital model, which contains standard geometric data and reference texture data for each feature area. This standard digital model serves as a reference for defect identification, identifying an initial set of defect features. Based on predefined process association rules, logical association analysis is performed on defects identified in different inspection stations for the same valve body. These process association rules are established based on the symbiotic or causal relationship between defects in several processing features or adjacent areas, thereby generating analysis results that include association determination.

[0045] Based on the analysis results, the detection strategy of at least one subsequent detection station is dynamically adjusted. This dynamic adjustment includes adjusting the image acquisition parameters, image processing algorithm parameters, or triggering a re-inspection process for the defect-related area. When the initial defect feature set shows that there is micro-corrosion on the tap surface inside the deep hole, the control unit dynamically increases the exposure time of the subsequent thread detection station and reduces the noise filtering intensity of the image processing algorithm to improve the detection rate of metal residual defects.

[0046] After generating the analysis results, the process also includes using machine learning models to predict the occurrence trend of similar defects in the current production batch and assess the risk level based on the analysis results and historical defect data. According to the risk level, the decision priority of the current valve body sorting or re-inspection process is adjusted. These data together constitute the basis for the machine learning model to learn and predict, enabling the model to extract batch-specific and trend-based patterns from individual defect information.

[0047] The model uses the current defect information analysis results of the valve body and historical defect data as input features. By learning from historical data, it identifies the potential correlation patterns between different defect types, defect combinations and future defect occurrence trends, and predicts the occurrence trend of similar defects in the current production batch. This means that the model outputs a quantitative indicator that shows the changing trend of the probability or severity of a specific defect type in the current batch or in the future. The risk level assessment is based on the predicted occurrence trend, mapping the prediction results to a preset risk level system.

[0048] When a machine learning model determines that a valve body or a current batch has a high risk level, the control unit can increase the priority of that valve body in the sorting process, such as sending it to the manual re-inspection station first, or marking it as requiring stricter quality control. At the same time, it can also adjust the decision priority of the re-inspection process. This adjustment ensures that limited inspection and processing resources can be more effectively allocated to the valve bodies or batches that require the most attention.

[0049] In valve body inspection methods, although the strategies of subsequent inspection stations can be dynamically adjusted based on the defect analysis results, if the process correlation between defects in different stations is not clearly identified and utilized, the adjustment of subsequent stations may lack specificity. This may lead to insufficient sensitivity in detecting potentially related defects, or difficulty in efficiently focusing on high-risk areas when processing a large amount of inspection data, thereby affecting the overall accuracy and efficiency of the inspection.

[0050] In response, this application further proposes a step for dynamically adjusting the detection strategy, which specifically includes: if, based on predefined process association rules, it is determined that the defect identified at the current workstation is associated with a subsequent workstation to be inspected, the analysis and control module will send association warning information to the subsequent workstation to be inspected. After receiving the association warning information, the subsequent workstation to be inspected will perform at least one of the following operations according to the warning content: enable a dedicated image analysis model for the associated defect type, adjust the defect identification threshold of the image processing algorithm, or mark the image area that needs to be reviewed. Once the analysis unit determines that the defect identified at the current workstation is associated with a subsequent workstation to be inspected, for example, a certain surface roughness defect found in an early inspection workstation may indicate the risk of bubbles or poor adhesion in the subsequent electroplating workstation, the control unit will immediately send association warning information containing the warning type, associated defect information, and suggested adjustment direction to the corresponding subsequent workstation to be inspected.

[0051] Upon receiving this associated warning information, the image processing system of the subsequent workstations will no longer simply execute the general inspection process, but will take targeted measures based on the warning content. In addition, when the system cannot fully automate the process or requires manual intervention for confirmation, the image processing system can also clearly indicate specific areas that are judged to be high-risk or require special attention based on the associated warning information on the image interface by highlighting, bordering, color coding, etc., thereby guiding operators or more advanced algorithms to prioritize these areas.

[0052] When inspecting the valve body, images are acquired through multiple inspection stations and compared with a pre-stored standard digital model to identify defects and perform logical correlation analysis. This allows for dynamic adjustment of subsequent inspection strategies. However, in actual production, the manufacturing process, material properties, or environmental factors of the valve body may undergo subtle changes, causing the baseline texture data or predefined process correlation rules in the original standard digital model to gradually deviate from the actual situation. This affects the accuracy of defect identification and the effectiveness of correlation analysis, potentially leading to missed detections or misjudgments.

[0053] In response, this application further proposes a model update step, which includes collecting historical inspection data containing defect features and correlation analysis results, and periodically optimizing and updating the baseline texture data or predefined process correlation rules in the standard digital model using the historical inspection data. For the predefined process correlation rules, these rules are established based on the symbiotic or causal relationship between defects in processing features or adjacent areas. By deeply mining the historical inspection data, the system can discover new defect correlation patterns, or verify, strengthen, and correct existing rules. The optimization and update is not a one-time operation, but is performed periodically, for example, it can be set to be performed weekly, monthly, or when production batches are switched. This periodicity ensures that the system can continuously adapt to changes in the production environment and maintain the timeliness and accuracy of its detection capabilities.

[0054] The following example will provide a more detailed explanation of the above technical solution: On an automated production line for automotive control valve bodies, the valve bodies to be inspected pass through multiple inspection stations sequentially via a conveyor path. This inspection system aims to perform comprehensive and accurate defect detection on the internal thread tapered surface, various sealing tapered surfaces, and complex appearance of the valve bodies.

[0055] The valve body enters the detection module, which has multiple detection stations arranged sequentially along the conveying path. At each detection station, the image acquisition device acquires real-time images of the corresponding feature areas of the valve body. These real-time images are transmitted to the analysis and control module. The analysis unit in the analysis and control module receives these images and compares them with the standard digital model stored in the comparison module. The analysis unit uses image processing algorithms to identify areas in the real-time images that do not conform to the standard model, thereby initially determining the location, type, and degree of defects and generating an initial defect feature set.

[0056] A key innovation of this system lies in its intelligent correlation analysis capability. Based on these rules, the analysis unit performs logical correlation analysis on the defects identified in the valve body at different inspection stations.

[0057] The control unit in the analysis and control module dynamically adjusts the detection strategy of subsequent detection stations based on the analysis results. This dynamic adjustment mechanism is significantly better than the existing technology's independent detection and information silo mode, avoiding the risk of downstream missed detection due to upstream defects not being fully identified or associated.

[0058] In addition, the analysis and control module also includes a prediction unit. Based on the current analysis results and historical defect data, the prediction unit uses a machine learning model to predict the occurrence trend of similar defects in the current production batch and outputs the risk level. For example, if multiple valve bodies have consecutive defects with side wall scratches and conical tool marks, the prediction unit will determine that the current batch has a high risk of processing abnormality. The control unit will adjust the decision priority of the current valve body sorting or re-inspection process according to the risk level. For defects with a risk level higher than the preset threshold, the control unit will trigger the marking and alarm of the processing parameters related to the defect type, and promptly remind the production line to adjust the process.

[0059] Throughout the inspection process, the system continuously collects historical inspection data containing defect features and correlation analysis results. This data is used to periodically optimize and update the baseline texture data in the standard digital model or the predefined process correlation rules, enabling the system to continuously learn and adapt to new defect patterns and process changes, further improving the accuracy and robustness of the inspection. This adaptive and learning capability is not available in existing single defect detection systems.

Claims

1. A valve body internal taper surface and appearance inspection system, characterized in that, include: The detection module includes multiple detection stations arranged sequentially along the conveying path, used to acquire images of several feature areas of the valve body, wherein the feature areas include at least the outer surface of the valve body, the conical surface of the valve body, and the threaded surface. The comparison module stores a standard digital model containing standard geometric data and reference texture data for each feature region of the valve body; The analysis and control module is communicatively connected to the detection module and the comparison module. The analysis and control module includes an analysis unit and a control unit; The analysis unit is used to compare the acquired real-time images with the standard digital model, identify an initial set of defect features including defect location, type and degree, and perform logical association analysis on defects identified in different inspection stations for the same valve body based on predefined process association rules. The process association rules are established based on the symbiotic or causal relationship between defects in several processing features or adjacent areas, and generate analysis results including association determination. The control unit is used to dynamically adjust the detection strategy of at least one subsequent detection station based on the analysis results; the dynamic adjustment includes adjusting image acquisition parameters, image processing algorithm parameters, or triggering a re-inspection process for defect-related areas.

2. The valve body taper surface and appearance inspection system according to claim 1, characterized in that, Multiple inspection stations include at least one of the following inspection stations: A multi-defect inspection station for conical surfaces is used to inspect at least one of the following defects in 84° and 120° conical surfaces: roughness, steps, mouth defects, electroplating solution residue, tool marks, scratches, bubbles, and corrosion. The thread inspection station, including the internal thread inspection station and the bottom outer ring thread inspection station, is used to inspect the thread surface and the bottom outer thread for burrs, metal residues or machining abnormalities. The valve body appearance and structural damage inspection station includes the cone bottom inspection station, end face inspection station, flange upper and lower surface inspection station, side inspection station, step burr inspection station, and lining end face inspection station, covering the inspection of damage, scratches, pits, corrosion, deformation, foreign object structure and appearance defects of the valve body cone bottom surface, end face, flange, side wall, step, and lining end face; Does the solder have an inspection station for inspecting the solder in the weld area? 3. The valve body taper surface and appearance inspection system according to claim 2, characterized in that, Each inspection station is equipped with an image acquisition device and a lighting device. Each image acquisition device uses a color area array camera. The lighting device is configured with one or more combinations of ring light, back light, spherical integrating light, and angled ring light according to the characteristic area to be inspected.

4. The valve body taper surface and appearance inspection system according to claim 1, characterized in that, The process association rules include defect propagation rules and comprehensive judgment rules; The defect propagation rule is used to: when a specific type of defect or a defect located in a defect association area is identified at an earlier detection station, the detection sensitivity or judgment priority of the preset associated defect type is increased in one or more subsequent associated detection stations according to the rule. The comprehensive judgment rule is used to: integrate the identification results of the same valve body at multiple related inspection stations, perform consistency verification or weighted evaluation on defects with process correlation, and output the final defect judgment.

5. The valve body taper surface and appearance inspection system according to claim 1, characterized in that, The analysis and control module also includes a prediction unit; the prediction unit is used to predict the occurrence trend of similar defects in the current production batch and output the risk level based on the analysis results and historical defect data through a machine learning model; the control unit is also configured to adjust the decision priority of the current valve body sorting or re-inspection process according to the risk level. The input features of the machine learning model include the historical defect data.

6. The valve body taper surface and appearance inspection system according to claim 5, characterized in that, The control unit is configured to: for defects with a risk level higher than a preset threshold, trigger the marking and alarm of the processing parameters related to the defect type; for defects with a risk level lower than the preset threshold, process them according to the standard procedure.

7. A method for inspecting the internal taper surface and appearance of a valve body, characterized in that, Includes the following steps: Images of different feature areas of the valve body are acquired by multiple inspection stations arranged sequentially along the conveying path. The feature areas include at least the outer surface, the valve body end face, the internal thread surface, and the deep hole internal taper surface. The acquired real-time images are compared with a pre-stored standard digital model, which includes standard geometric data and reference texture data for each feature region. An initial set of defect features is identified, and based on predefined process association rules, logical association analysis is performed on defects identified in different inspection stations for the same valve body. The process association rules are established based on the symbiotic or causal relationship between defects in several processing features or adjacent areas, generating analysis results that include association determination. Based on the analysis results, the detection strategy of at least one subsequent detection station is dynamically adjusted; the dynamic adjustment includes adjusting image acquisition parameters, image processing algorithm parameters, or triggering a re-inspection process for defect-related areas.

8. The method for inspecting the taper surface and appearance inside the valve body according to claim 7, characterized in that, After generating the analysis results, the process also includes the following steps: based on the analysis results and historical defect data, using a machine learning model to predict the occurrence trend of similar defects in the current production batch and assess the risk level. Based on the risk level, adjust the decision priority of the current valve body sorting or re-inspection process.

9. The method for inspecting the taper surface and appearance inside the valve body according to claim 7, characterized in that, The steps for dynamically adjusting the detection strategy include: If, based on the process association rules, it is determined that the defect identified at the current workstation is associated with a subsequent workstation to be inspected, then an association warning message is sent to the subsequent workstation to be inspected. The subsequent inspection station shall perform at least one of the following operations based on the associated early warning information: enable a dedicated image analysis model for the associated defect type, adjust the defect identification threshold of the image processing algorithm, or mark the image area that needs to be reviewed.

10. The method for inspecting the taper surface and appearance inside the valve body according to claim 7, characterized in that, It also includes a model update step: collecting historical inspection data containing defect features and correlation analysis results; based on this historical inspection data, periodically optimizing and updating the baseline texture data of the standard digital model and the preset process correlation rules.

Images