Precision manufacturing defect automatic identification method and system based on artificial intelligence

Abstract

The invention provides a precision manufacturing defect automatic identification method and system based on artificial intelligence, and relates to the technical field of precision manufacturing. The method comprises the steps of firstly collecting historical object types and corresponding use scene data, constructing a defect type standard determination model according to defect rating standard labels of different objects in different scenes, collecting image data of a current target precision object and a use scene thereof, and determining the defect type standard of the current target precision object according to the image data. The method comprises the following steps: respectively inputting a pre-trained object type identification model and a pre-trained use scene identification model, obtaining a target precision object type and a current use scene, inputting two results into a defect type standard determination model to obtain a matched defect judgment standard, collecting physical characteristic data of an object through a plurality of sensors, comparing the physical characteristic data with the judgment standard, and determining the target precision object type and the current use scene. Therefore, the defect identification result of the target precision object is determined. By implementing the method, the problem that proper judgment is difficult to make aiming at differences in a traditional method is solved, and a defect identification result is enabled to better meet actual application requirements.

Description

Technical Field

[0001] The present application relates to the field of precision manufacturing technology, and in particular to an artificial intelligence-based method and system for automatically identifying precision manufacturing defects. Background Art

[0002] With the rapid development of the manufacturing industry, precision-manufactured products are widely used in various fields. The quality of these precision-manufactured products directly affects their performance and service life, so identifying product defects is particularly important. Currently, identifying defects in precision-manufactured products has become a key link in manufacturing quality control.

[0003] Existing defect detection for precision-manufactured products primarily relies on a combination of manual visual inspection and fixed-standard testing. Inspectors use visual observation combined with measurement tools to inspect the product's surface and dimensions, following the standardized testing standards specified in the product manual, to identify potential defects. Measuring instruments are also used to quantitatively analyze the product's physical characteristics, comparing the measured data with standard thresholds in the product manual to determine if defects exist.

[0004] However, with the increasing diversification of application scenarios for precision manufacturing products, the tolerance for defects of the same product in different usage scenarios varies significantly. For example, in high-demand applications, some products that should be deemed unqualified may be mistakenly deemed qualified. In ordinary applications, some actually usable products may be overly harshly judged as unqualified, resulting in unnecessary quality loss. This testing method cannot guarantee product quality and safety in high-demand applications and also results in a waste of production resources. Therefore, traditional methods have difficulty making appropriate judgments based on these differences, resulting in defect identification results that may not meet actual application requirements. Summary of the Invention

[0005] This application provides an artificial intelligence-based method and system for automatic identification of precision manufacturing defects, which is used to solve the problem that traditional precision manufacturing defect detection methods are difficult to take into account the differences between different objects in various usage scenarios, resulting in the defect identification results being out of touch with actual application needs.

[0006] In the first aspect, the present application provides an automatic identification method for precision manufacturing defects based on artificial intelligence, which is applied to a defect identification system. The method comprises: after acquiring multiple historical object types and corresponding usage scenario data, constructing a defect type standard determination model according to the defect rating standard annotations of objects of different object types in different usage scenarios; acquiring the current target precision object image data through multiple visual sensors, and simultaneously acquiring the usage scenario image data of the target precision object; inputting the target precision object image data into a preset precision manufacturing object type recognition model to obtain the target precision object type, and inputting the usage scenario image data into a preset usage scenario recognition model. In the scene recognition model, the current usage scenario is obtained. The precision manufacturing object type recognition model is trained in advance based on multiple image data covering various types of precision manufacturing objects. The usage scenario recognition model is trained in advance based on image data of multiple different usage scenarios and corresponding scene labels; the target precision object type and the current usage scenario are input into the defect type standard determination model to obtain the defect judgment standard that matches the target precision object type in the current usage scenario; the physical feature data of the object is collected through multiple sensors, and the physical feature data of the object is compared with the defect standard to determine the defect recognition result of the target precision object.

[0007] By adopting the above technical solution, multiple historical object types and corresponding usage scenario data are obtained, and a defect type standard determination model is constructed. This can integrate the defect judgment criteria of different objects in various scenarios. Multiple visual sensors are used to collect target object and scene image data. The trained model identifies the object type and scene, thereby obtaining an adapted defect judgment standard. Finally, the physical feature data is collected and compared with the standard to determine the defect identification result. This series of operations works closely together, fully considering the diversity of objects and scenarios, changing the limitations of traditional unified standard testing, accurately identifying defects, and effectively solving the problem that traditional methods have difficulty in making appropriate judgments based on differences, making the defect identification results more in line with actual application needs.

[0008] In combination with some embodiments of the first aspect, in some embodiments, after the step of collecting physical feature data of an object through multiple sensors, comparing the physical feature data of the object with the defect standard, and determining the defect identification result of the target precision object, the method further includes: obtaining temperature distribution data and vibration spectrum data in the usage scenario; determining the local thermal stress concentration area of ​​the target precision object based on the temperature distribution data; determining the resonant frequency range of the target precision object based on the vibration spectrum data; spatially mapping the local thermal stress concentration area and the resonant frequency range in the same spatial coordinate system to determine the overlapping area; when the defect in the defect identification result is located in the overlapping area, increasing the defect detection sensitivity of the overlapping area and performing re-detection.

[0009] By adopting the above technical solution, after determining the defect identification results, temperature distribution and vibration spectrum data are obtained, and then the local thermal stress concentration area and resonance frequency range are determined and spatially mapped. Because in the area where thermal stress concentration and resonance overlap, the defects of the object are affected by a variety of complex factors and are more likely to develop and change. When the defect is in this area, increasing the detection sensitivity and re-detection can accurately capture the subtle changes in the defect, avoiding misjudgment due to not considering the impact of this special area on the defect, ensuring that the defect identification results can truly reflect the actual situation of the object under complex working conditions, and effectively solving the problem that the defect identification results of traditional methods do not meet the actual application requirements.

[0010] In combination with some embodiments of the first aspect, in some embodiments, after the step of collecting physical feature data of an object through multiple sensors, comparing the physical feature data of the object with the defect standard, and determining the defect identification result of the target precision object, it also includes: after determining the defect location based on the defect identification result, applying an excitation signal to the defect location; determining the stress wave propagation characteristics corresponding to the defect of the target precision object based on the excitation signal; analyzing the attenuation rate of the stress wave in different usage scenarios based on the stress wave propagation characteristics; when there is an abnormality in the attenuation rate, automatically lowering the defect judgment standard of the defect.

[0011] By employing this technical solution, an excitation signal is applied after the defect location is identified, thereby determining the stress wave propagation characteristics and analyzing the attenuation rate of the stress wave in different usage scenarios. Abnormal attenuation rates often indicate that the actual defect is more severe than initially diagnosed. This dynamic adjustment of the defect assessment criteria automatically lowers when an abnormal attenuation rate occurs, overcoming the problem of traditional fixed criteria being unable to adapt to actual defect changes and ensuring a more realistic defect assessment.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after the step of inputting the target precision object type and the current usage scenario into the defect type standard determination model to obtain the defect judgment standard that matches the target precision object type in the current usage scenario, it also includes: obtaining scene records of historical precision object failures, including the ambient temperature and failure consequences of the failure, to form a failure case library; dividing the scenes in the failure case library into high-risk scenarios, medium-risk scenarios and low-risk scenarios according to the severity of the failure consequences; determining the time interval for each defect to evolve into failure under different risk scenarios, and when the failure time interval of the target defect in the high-risk scenario is shorter than that in the non-high-risk scenario, the target defect is marked as a high-risk defect; for the defect type marked as a high-risk defect, the defect judgment standard is automatically tightened in the high-risk scenario; real-time monitoring of the current working scene of the target precision object, and immediately triggering the defect review prompt process when the scene changes from a low-risk scene to a high-risk scene.

[0013] By adopting the above technical solution, historical records of precision object failure scenarios are obtained to build a failure case library, divide risk scenarios, and count the time intervals for defects to evolve into failures. High-risk defects are marked, and the evaluation criteria are tightened in high-risk scenarios. At the same time, scene transitions are monitored in real time and re-inspection prompts are triggered. Due to the different tolerances for defects and the speed of defect development in different scenarios, traditional methods are unable to flexibly respond to such differences. This solution dynamically adjusts the evaluation criteria and detection process according to the scenario risk, so that defect identification can adapt to the actual requirements of different scenarios, avoiding misjudgment of defects in different risk scenarios. It effectively solves the problem of defect identification results in traditional methods not matching actual application requirements, and ensures the reliable operation of objects in different scenarios.

[0014] In combination with some embodiments of the first aspect, in some embodiments, after the step of collecting physical feature data of an object through multiple sensors, comparing the physical feature data of the object with the defect standard, and determining the defect identification result of the target precision object, it also includes: obtaining the physical feature data of the target precision object at preset time periods to establish a time series data set; calculating the rate of change of the physical feature based on the time series data set; when the rate of change of the physical feature exceeds a preset rate threshold, marking the corresponding physical feature as a rapid degradation feature, and increasing the frequency of collecting physical features; determining the feature future trend prediction value of the rapid degradation feature through a feature development prediction model, and the feature development prediction model is obtained in advance based on the physical feature data of multiple target precision objects and the corresponding defect development conditions through machine learning training; if the deviation of the feature future trend prediction value and the current defect judgment standard shows an increasing trend, the warning device is controlled to issue a potential defect warning signal.

[0015] By implementing this technical solution, we achieve dynamic monitoring of an object's physical characteristics and early prediction of potential defects. Traditional methods struggle to effectively monitor dynamic changes in an object's physical characteristics and promptly identify potential problems. This solution, through dynamic monitoring and prediction, can detect potential defects in advance, avoiding misjudgments caused by ignoring changes in physical characteristics. This ensures that defect identification results reflect the actual state of the object, improving the practicality of the results.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of collecting physical feature data of an object through multiple sensors, comparing the physical feature data of the object with the defect standard, and determining the defect identification result of the target precision object, it also includes: constructing a three-dimensional digital model of the target precision object based on the surface contour data, internal structure data and material property data of the target precision object collected in advance; collecting spatial layout data, environmental parameter data and working condition data of the current usage scene, and spatially mapping these scene feature data and the three-dimensional digital model in the same coordinate system to form a 3D display scene; visually annotating the defect identification results and integrating them into the 3D display scene for display.

[0017] By implementing this technical solution, workers can gain a direct and comprehensive understanding of the actual conditions of objects and defects, resolving the issue of unintuitive and incomplete information display in traditional inspections. Traditional methods can lead to misjudgments of defects due to insufficient information. This visualization approach helps workers make quick and accurate judgments, improving the accuracy and efficiency of defect identification. This makes defect identification results more aligned with practical application needs and facilitates the implementation of targeted measures.

[0018] In combination with some embodiments of the first aspect, in some embodiments, after the step of collecting physical feature data of an object through multiple sensors, comparing the physical feature data of the object with the defect standard, and determining the defect identification result of the target precision object, it also includes: collecting stress distribution data around the defect based on the defect identification result to obtain defect feature data; counting the frequency of different defect types in the defect feature data appearing simultaneously on the same object, and calculating the correlation between defect types; when the correlation exceeds a preset correlation threshold, establishing a correlation relationship between defect types; analyzing the order of occurrence of defects with the correlation relationship in the time series to determine the defect evolution path; determining the derivative related defects of the current defect based on the defect evolution path; and increasing the detection frequency of the derived related defects by related equipment.

[0019] By adopting the above technical solution, the problem that traditional methods usually only focus on single defects and ignore the relationship between defects is solved. This solution detects possible derivative defects in advance by exploring the intrinsic connections between defects, avoiding misjudgment of the overall condition of the object due to failure to consider defect correlations, making defect identification more comprehensive and accurate, and effectively solving the problem that traditional defect identification results do not meet actual application requirements, thereby ensuring the overall performance and safety of the object.

[0020] In a second aspect, the present application provides a defect identification system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the defect identification system to execute the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, the present application provides a computer-readable storage medium comprising instructions, which, when executed on a defect identification system, enable the defect identification system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, the present application provides a computer program product, which, when executed on a defect identification system, enables the defect identification system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Due to the adoption of the technical means of collecting multiple historical object types and corresponding usage scenario data to build a defect type standard determination model, using visual sensors to collect data and using the trained model to identify the object type and scene to obtain matching defect judgment standards, and then comparing the object physical feature data with the standard, it effectively solves the technical problem in the existing technology that it is difficult to make appropriate judgments based on the differences between different objects and scenes due to the use of unified standard detection, resulting in defect identification results that do not meet actual application requirements. It then realizes the technical effect of accurately identifying the defects of different precision-manufactured objects in various usage scenarios and significantly improving the accuracy and reliability of defect identification.

[0024] 2. Due to the adoption of technical means to obtain historical records of failure scenarios of precision objects to build a failure case library, divide risk scenarios and count the time intervals for defects to evolve into failures, mark high-risk defects, tighten the evaluation criteria in high-risk scenarios, and monitor scenario conversions in real time and trigger re-inspection prompts, it effectively solves the technical problem that the existing technology cannot dynamically adjust the defect evaluation criteria and detection process according to the risk of the usage scenario, resulting in inaccurate defect identification in different risk scenarios. It then achieves the technical effect of accurately identifying defects in different risk scenarios and ensuring the safe and stable operation of objects in various scenarios.

[0025] 3. Due to the adoption of technical means to collect stress distribution data around defects to obtain defect characteristic data, calculate the correlation between defect types, establish correlation relationships to determine the defect evolution path and derived related defects, and increase the frequency of detection of derived defects, it effectively solves the technical problem of the existing technology that only focuses on a single defect and ignores the correlation between defects, resulting in misjudgment of the overall condition of the object, thereby achieving the technical effect of comprehensively and accurately identifying defects and ensuring the overall performance and safety of the object. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of a method for automatic identification of precision manufacturing defects based on artificial intelligence in an embodiment of the present application; Figure 2 This is another flowchart of the method for automatic identification of precision manufacturing defects based on artificial intelligence in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of a physical device of the defect identification system in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The terms used in the following examples of the present application are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular expressions "a," "an," "said," "above," "the," and "this" are intended to include plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to and encompasses any or all possible combinations of one or more of the listed items.

[0028] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of the technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of this application, unless otherwise specified, "plurality" means two or more.

[0029] For ease of understanding, the following describes the process of the method provided by this implementation. Figure 1 , which is a flow chart of the method for automatic identification of precision manufacturing defects based on artificial intelligence in an embodiment of the present application.

[0030] S101. After obtaining multiple historical object types and corresponding usage scenario data, construct a defect type standard determination model based on defect rating standard annotations for objects of different object types in different usage scenarios; In precision manufacturing, defect rating standards for different object types and usage scenarios vary significantly. Defect recognition systems leverage web crawler technology to capture extensive historical data on multiple object types and corresponding usage scenarios from major manufacturing databases, industry-standard document storage platforms, and internal enterprise production record systems. This defect recognition system can be applied to both production lines and portable devices, without limitation here. Labeling defect rating standards for different objects and scenarios is crucial for model building. For example, in aircraft engine blades, even small cracks can cause serious safety incidents under high-temperature, high-speed operating conditions, resulting in extremely strict crack rating standards. On the other hand, in common industrial fan blades, cracks of the same size may have relatively looser rating standards due to varying usage scenarios. Based on these real-world scenarios, the system categorizes the collected historical data by object type and usage scenario. For each data category, a team of experienced industry experts is invited to annotate defect rating standards. Drawing on their extensive experience and expertise, these experts meticulously categorize different defect types and determine corresponding rating standards, such as defect severity, maximum allowable size, and impact on object performance.

[0031] To build the model, the system uses a neural network algorithm from deep learning, specifically a multilayer perceptron (MLP) model architecture. The cleaned and labeled data is divided into a training set, a validation set, and a test set, with the training set accounting for 70%, the validation set 15%, and the test set 15%. The MLP model is trained using the training set, and by continuously adjusting the model's weights and biases, the model accurately learns the mapping between different object types and usage scenarios and defect rating criteria. During training, a backpropagation algorithm is used to calculate the gradient of the loss function, and the model parameters are updated based on the gradient to minimize the loss function. The validation set is used to evaluate model performance during training and prevent overfitting. Training is terminated when the model's performance on the validation set no longer improves. Finally, the trained model is evaluated on the test set to ensure good generalization to unseen data. After multiple rounds of training and optimization, a defect type determination model is developed that accurately outputs the corresponding defect rating criteria based on the input object type and usage scenario.

[0032] S102, acquiring current target precision object image data through multiple visual sensors, and simultaneously acquiring usage scene image data of the target precision object; The defect recognition system is equipped with a visual sensing system comprised of multiple visual sensors of varying types and functions, ensuring comprehensive image acquisition from multiple angles and dimensions. To capture image data of precision target objects, the system uses high-resolution industrial cameras as the primary sensors. These industrial cameras feature high pixel counts and high frame rates, enabling them to clearly capture even the finest surface features. For example, when inspecting precision mechanical parts, the camera's pixel count can reach millions or even higher, while maintaining a frame rate sufficient for high-speed production line capture, ensuring clear and complete capture of every moving part. Multiple industrial cameras are arranged at various angles around the target object, forming a three-dimensional imaging network. Careful camera positioning and angles ensure that every surface and corner of the object is captured, eliminating blind spots.

[0033] In addition to acquiring image data of the target precision object, the defect recognition device also needs to acquire image data of its usage scenario. To this end, the system utilizes context-aware cameras, installed within the target object's usage scenario, to capture the overall scene in real time. Alternatively, usage scenario information can be obtained through a user terminal connected to the system. This user terminal device must feature an intuitive, easy-to-use interface designed for simplicity and clarity, presenting various options and information in a graphical format. Users can input and query information through simple clicks and swipes. For example, after powering on the device, the user first sees a homepage listing common usage scenario categories, such as "factory production workshop," "outdoor power facilities," and "medical equipment usage environment." Clicking on a category reveals further subcategories. For example, under "factory production workshop," subcategories can be found such as "automotive manufacturing workshop," "electronic component production workshop," and "machining workshop." Users can select the appropriate usage scenario based on their specific needs, and the user terminal then sends the selected scenario data to the system.

[0034] S103: Input the target precision object image data into a preset precision manufacturing object type recognition model to obtain the target precision object type, and input the usage scene image data into a preset usage scene recognition model to obtain the current usage scene. The precision manufacturing object type recognition model is pre-trained based on a plurality of image data covering various types of precision manufacturing objects, and the usage scene recognition model is pre-trained based on image data of a plurality of different usage scenes and corresponding scene labels. The system employs a convolutional neural network (CNN) architecture for its precision-manufactured object type recognition model. During the training phase, the system collects image data of various precision-manufactured objects from a massive image database. This data covers a wide range of scenarios, including the same type of object produced by different manufacturers, variations between different production batches of the same object, and changes in the object's appearance under different conditions of use. For example, for precision-manufactured objects like electronic chips, the collected data includes not only images of chips of different brands and models, but also images of chips at different stages of the production process, as well as images of chips subjected to wear, oxidation, and other conditions that may occur during actual use. Once the target precision object image data is acquired, it is input into the trained precision-manufactured object type recognition model. The image data first passes through a series of convolutional layers, where convolution kernels slide across the image to extract local features. These convolution kernels are trained to learn different characteristic patterns, such as lines and corners. Next, the pooling layer performs dimensionality reduction on the feature maps output by the convolutional layers, reducing the data volume while preserving key features and improving computational efficiency. Finally, the features are integrated and classified through the fully connected layer to output the type of the target precision object, such as "gear", "bearing", "integrated circuit board", etc.

[0035] For the usage scene recognition model, the system also uses a deep learning model based on convolutional neural networks. During the training phase, the system collects a large amount of image data from different usage scenarios and annotates each image with accurate scene labels, such as "factory production workshop," "outdoor construction site," and "laboratory environment." In order to enable the model to better learn the key features in the scene images, in addition to the conventional convolutional and pooling layers, the system introduces an attention mechanism. For example, when identifying the "factory production workshop" scene, the model will focus on key elements such as the production equipment and production line layout in the workshop, while ignoring some irrelevant background information. After sufficient training, the acquired usage scene image data is input into the usage scene recognition model. The model extracts and analyzes the image data, captures the key features in the scene through the convolutional layer and attention mechanism, and then performs classification through the pooling layer and fully connected layer. It outputs the category of the current usage scene, providing accurate scene information for the subsequent determination of defect assessment criteria.

[0036] S104: inputting the target precision object type and the current usage scenario into the defect type standard determination model to obtain a defect assessment standard that matches the target precision object type in the current usage scenario; After obtaining the target precision object type and the current usage scenario, the defect recognition system inputs these two key pieces of information into the defect type standard determination model to obtain the matching defect evaluation criteria. The defect type standard determination model uses a neural network algorithm in deep learning, specifically a multi-layer perceptron (MLP) model structure. Before building the model, the system deeply processes the large amount of historical data collected. This historical data covers detailed information on different object types in various usage scenarios, including the physical characteristics of the object, defect type, defect rating standards, and corresponding usage scenario parameters. For example, for aircraft engine blades, in high-temperature, high-speed flight scenarios, their defect rating standards will have strict restrictions on the size, location, and number of defects such as cracks and wear; while for ordinary household fan blades, in low-speed, normal-temperature usage scenarios, the allowable range of defects of the same type will be much looser.

[0037] During model training, the system divides the preprocessed data into training, validation, and test sets according to preset ratios. The MLP model is trained using the training set, and by continuously adjusting the model's weights and biases, it accurately learns the complex mapping between different object types and usage scenarios and defect rating criteria. When the target precision object type and the current usage scenario are input into the defect type standard determination model, the model first extracts and encodes the input information, converting it into a vector form that the model can process. Then, through computations and nonlinear transformations using multiple layers of neurons, it gradually learns the complex relationship between the input information and the defect assessment criteria. For example, the model comprehensively considers factors such as the object's material properties, the operating environment's temperature, humidity, and vibration, and the impact of different defect types on object performance to determine the defect assessment criteria. Finally, the model outputs defect assessment criteria that match the target precision object type in the current usage scenario. This includes detailed information such as the defect type, maximum allowable size, defect severity level, and criteria for assessing the impact on object performance. For example, if the input target precision object type is "automobile engine piston" and the current usage scenario is "engine high-speed operation, high temperature and high pressure environment", the model may output the following defect judgment criteria: for scratches on the piston surface, if the length exceeds 0.5 mm, it is judged as a serious defect; for wear on the piston ring groove, if the wear exceeds 0.1 mm, it is judged as unqualified; for pore defects inside the piston, if the pore diameter exceeds 0.2 mm, it is judged as a defective product. These defect judgment criteria will provide an accurate basis for the subsequent collection of physical feature data of the object through sensors and defect identification.

[0038] S105 , collecting physical feature data of the object through multiple sensors, comparing the physical feature data of the object with the defect standard, and determining a defect recognition result of the target precision object.

[0039] The system is equipped with various types of sensors to comprehensively collect physical characteristic data from objects, including high-precision displacement sensors to measure dimensional accuracy and shape deviation. Displacement sensors can accurately measure dimensions such as apertures and shaft diameters of precision mechanical parts down to the micron level, ensuring accurate measurement data. To measure surface roughness, the system uses a surface roughness meter. This instrument uses a stylus to trace across the surface, acquiring microscopic surface profile information and calculating surface roughness parameters. Ultrasonic sensors play a key role in detecting internal defects in metal objects. They exploit the principle that ultrasonic waves, when propagating within an object, are reflected, refracted, and scattered upon encountering defects. By analyzing the signal characteristics of the reflected waves, they determine the presence of defects within the object, as well as their location, size, and shape. During the physical characteristic data collection process, the system regularly calibrates the sensors to ensure measurement accuracy. Furthermore, to improve data collection efficiency, the system utilizes multi-sensor synchronous acquisition technology, enabling multiple sensors to operate simultaneously to rapidly acquire data on different aspects of an object's physical characteristics.

[0040] After receiving the physical feature data, the system compares it with the previously acquired defect assessment criteria. During the comparison process, the system uses precise algorithms to determine whether the data meets the standards. For example, for dimensional data, the system will compare the measured value with the allowable size range specified in the defect assessment standard. If the measured value exceeds the range, it is determined that the dimension is defective. For surface roughness data, the system will compare the measured roughness parameters with the thresholds in the standard. If the parameters exceed the thresholds, the surface roughness is considered to be unsatisfactory. For defect signal data obtained by ultrasonic testing, the system will analyze the signal's amplitude, phase and other characteristics based on the preset defect feature model to determine whether there is a defect and the severity of the defect.

[0041] After the comparison is completed, the system will generate a defect identification report based on the comparison results. The report records in detail the various physical feature data of the object, the comparison with the defect evaluation criteria, and the final defect identification results. If the object has defects, the report will clearly indicate the type, location, severity and other information of the defects. For example, the report may show that "the target precision object is a car engine piston. A scratch with a length of 0.6 mm was found on the surface of the piston. According to the defect evaluation criteria in the current usage scenario, this scratch is a serious defect; the wear of the piston ring groove is 0.12 mm, which is judged to be unqualified."

[0042] In an embodiment of the present application, a defect type standard determination model is constructed by collecting data from multiple historical object types and corresponding usage scenarios, using visual sensors to collect data and using trained models to identify object types and scenarios to obtain matching defect judgment standards, and then collecting physical feature data of the object and comparing it with the standard to determine the defect identification result, thereby achieving accurate identification of defects of different precision-manufactured objects in various usage scenarios, and effectively solving the technical problem in the prior art that due to the use of unified standard detection, it is difficult to make appropriate judgments based on the differences between different objects and scenarios, resulting in defect identification results that do not meet actual application requirements.

[0043] In some embodiments, after completing the critical step of collecting physical feature data of an object through multiple sensors and comparing it to defect standards to determine the defect identification results for the target precision object, to further improve the efficiency of understanding and managing the object and its defects, the system uses specialized 3D modeling software and algorithms to construct a 3D digital model of the target precision object based on the previously collected surface contour data, internal structure data, and material property data. This model accurately reproduces the object's true form and characteristics, from subtle surface textures to complex internal structures and the distribution of different materials. Simultaneously, the system also collects spatial layout data, environmental parameter data, and operating condition data for the current usage scenario. This includes spatial layout information such as the placement of devices in the scenario and the size of the space; environmental parameters such as temperature, humidity, and light intensity; and operating condition data such as the device's operating speed and power. Then, using spatial mapping technology, this scene feature data is integrated with the 3D digital model in the same coordinate system to form a realistic 3D display scene. In this 3D display scene, the object is closely integrated with its environment, accurately reflecting the object's state in the actual usage scenario. Finally, the system visually annotates the previously determined defect identification results, clearly marking the defect's location, shape, size, and other information on the 3D model through specific graphics and colors, and integrating this information into the 3D display scene for display. This allows workers to intuitively and comprehensively observe the overall condition of the target precision object in actual use scenarios, including the object's structure, environment, and existing defects. This greatly facilitates defect analysis and processing, improves the accuracy and efficiency of defect identification, and enables subsequent repairs and improvements to be carried out more targeted.

[0044] In some embodiments, after using multiple sensors to collect physical characteristic data from an object and comparing it to defect standards to determine the defect identification results for the target precision object, the system continuously acquires physical characteristic data from the target precision object at preset intervals, such as every hour or every day, and arranges this data in an orderly manner to create a time series dataset. Using this dataset, the system can calculate the rate of change of physical characteristics, such as the speed of change of physical characteristics such as object size and surface roughness over time. Once the rate of change of a physical characteristic exceeds a preset rate threshold, the system quickly marks the corresponding physical characteristic as a rapid degradation characteristic, indicating that the physical characteristic has undergone an abnormal change, which may indicate a rapid decline in the object's performance. At this point, the system automatically increases the frequency of physical characteristic acquisition and monitors the characteristic changes more intensively to keep abreast of its dynamics. The system uses a feature development prediction model, previously trained through machine learning based on a large number of physical characteristic data from target precision objects and corresponding defect development annotations, to determine the future trend prediction value of the rapid degradation characteristic, thereby predicting the future change direction of the physical characteristic. If the deviation between the predicted future trend value of the feature and the current defect judgment standard shows an increasing trend, that is, the predicted change in the physical feature may lead to a more serious defect risk in the object, the system will immediately control the early warning equipment to issue a potential defect early warning signal, reminding relevant personnel to pay attention and take measures in time to avoid serious problems caused by potential defects and ensure that the object can continue to operate stably.

[0045] In some embodiments, after determining the defect identification results, stress distribution data around the defect can be collected based on the results to obtain defect signature data that reflects the defect's characteristics. The system then conducts in-depth analysis of this defect signature data, counting the frequency of different defect types occurring simultaneously on the same object and calculating the correlation between the defect types using a specific algorithm. Once the correlation exceeds a pre-set threshold, the system establishes a correlation between the defect types, indicating a potential inherent connection between the different defect types. The system then conducts a detailed analysis of the chronological order of the associated defects, teasing out the evolution of the defects from a wealth of historical data and current cases. This analysis then identifies the defect evolution path, clearly demonstrating how one defect leads to other related defects. Based on this determined defect evolution path, the system can accurately identify related defects that may arise from the current defect. To promptly detect these potential related defects, the system automatically increases the frequency of inspections on relevant equipment, enabling early detection and effective action before the resulting defects cause more serious consequences. This ensures the overall performance and safety of precision objects, avoids misjudgments of the object's overall condition due to ignoring the correlations between defects, and improves the comprehensiveness and accuracy of defect identification.

[0046] In some embodiments, after completing the acquisition of physical characteristic data from multiple sensors and comparing it to defect standards to determine the defect identification results for the target precision object, the system acquires temperature distribution data and vibration spectrum data within the usage scenario to further accurately monitor the object's defect status. Leveraging a high-precision temperature sensor network, the system can comprehensively and accurately acquire temperature information at every location within the usage scenario. Using a thermal stress analysis algorithm, the system identifies localized thermal stress concentration areas within the target precision object based on the temperature distribution data. Uneven temperature distribution can cause thermal stress within the object, and certain areas may form thermal stress concentration points due to factors such as structure and material. Simultaneously, the system uses vibration sensors to acquire vibration spectrum data and, through spectrum analysis techniques, determines the resonant frequency range of the target precision object. Resonant frequency is closely related to the object's structure and mass distribution. The system then uses spatial mapping technology to integrate the localized thermal stress concentration areas and the resonant frequency range within the same spatial coordinate system to identify overlapping regions. This is because the structural stability of the object itself is already compromised in these areas of thermal stress concentration, and resonance further exacerbates the object's vibration and stress changes. The overlapping regions have a more complex and critical impact on the development of defects in the object. When the defect in the defect identification result happens to be located in the overlapping area, considering the special impact of this area on the development of the defect, in order to avoid missed detection or misjudgment, the system will increase the defect detection sensitivity of the overlapping area and perform re-detection, such as using higher-resolution detection equipment, more advanced detection algorithms, etc., to more accurately capture subtle changes in the defect, ensure more accurate judgment of the defect, and provide a reliable basis for subsequent targeted repair or preventive measures, ensuring the safe and stable operation of precision objects under complex working conditions.

[0047] In some embodiments, after obtaining defect judgment criteria that match the target precision object type and the current usage scenario, the defect recognition system will obtain historical records of precision object failure scenarios to form a failure case library, divide the scenarios into high, medium, and low risk according to the severity of the failure consequences, and count the time intervals for each defect to evolve into failure under different risk scenarios. Target defects with a shorter time interval for failure in high-risk scenarios than in non-high-risk scenarios will be marked as high-risk defects, and the judgment criteria for such defects will be tightened in high-risk scenarios. At the same time, the system monitors the working scene of the target precision object in real time. Once the scene changes from low risk to high risk, it will immediately trigger the defect re-inspection prompt process to accurately control defect risks and ensure the safe and stable operation of the object. The following is a further and more specific process description of the method provided by this implementation. Please refer to Figure 2 , is another flow chart of the method for automatic identification of precision manufacturing defects based on artificial intelligence in an embodiment of the present application.

[0048] S201. Obtain historical records of failure scenarios of precision objects, including the ambient temperature and consequences of the failure, to form a failure case library. Precision components have varying tolerances for defects at different ambient temperatures. Collecting ambient temperature data from a large number of failure cases can provide a reference for developing more precise defect assessment criteria. Precision components operating in high-temperature environments may have a lower tolerance for certain defects (such as microcracks caused by concentrated thermal stress). Meanwhile, the acceptable range for the same defect may vary at room or low temperatures. The temperature data in the failure case library can help defect recognition systems fully consider ambient temperature when determining defect assessment criteria, ensuring that the criteria better meet actual application needs.

[0049] In the data collection phase, the system uses a variety of data collection methods. On the one hand, by establishing a data interface with the company's production management system and quality monitoring system, it directly obtains historical data accumulated within the company. For example, in an automobile manufacturing company, it connects to the quality inspection database of its engine production workshop to obtain failure records of engine parts during production testing and actual use, including the failure time of key components such as pistons and crankshafts, the operating conditions at the time, and the corresponding ambient temperature data. On the other hand, web crawler technology is used to widely collect relevant data from industry data platforms, professional technical forums, and academic databases. For the electronic chip industry, failure cases of different types of chips are obtained from professional chip data platforms, including the failure conditions of chips in different application scenarios (such as communication base stations and consumer electronic devices).

[0050] S202. Classify the scenarios in the failure case library into high-risk scenarios, medium-risk scenarios, and low-risk scenarios based on the severity of the failure consequences; The system obtains standard values ​​for key performance parameters from product design data and simultaneously monitors actual values ​​in real time through sensors or records them at the moment of failure. For example, during engine operation, onboard sensors collect real-time data such as power output and fuel consumption. For semiconductor chips, actual values ​​such as signal transmission rate and power consumption are collected at monitoring nodes during production testing or during use. The deviation of key performance parameters is calculated using a specific formula: Deviation = (|actual value - standard value| ÷ standard value) × 100%. For example, if the standard power output of an automobile engine is 150 horsepower and the actual power output at the time of failure is 120 horsepower, then the power output deviation = (|120 - 150| ÷ 150) × 100% = 20%. Based on the operating requirements, safety standards, and historical failure data of precision instruments, the system sets different deviation thresholds to categorize risk levels. For aircraft engines with extremely high performance requirements, a deviation of more than 10% in a key performance parameter is considered high-risk; a deviation between 5% and 10% is considered medium-risk; and a deviation below 5% is considered low-risk. For general industrial equipment, these thresholds may be adjusted to 15%, 8%-15%, and below 8% respectively.

[0051] The system compares the calculated deviations of key performance parameters with the set thresholds. If the deviation exceeds the high-risk threshold, the corresponding failure scenario is classified as high-risk; if it is within the medium-risk threshold, it is classified as medium-risk; if it is below the low-risk threshold, it is classified as low-risk. After classification, the system labels each scenario with a corresponding risk level in the failure case library, allowing for subsequent differentiated defect analysis and management strategy development for scenarios with different risk levels.

[0052] S203. Determine the time interval for each defect to evolve into a failure under different risk scenarios. If the time interval for the target defect to fail under a high-risk scenario is shorter than that under a non-high-risk scenario, mark the target defect as a high-risk defect. Based on the risk scenarios that have been divided, the system filters out the relevant data of each defect in different risk scenarios from the failure case library. Using SQL query statements, combined with information such as defect type and risk scenario label, the required data can be accurately located. For example, for the "piston ring wear" defect of automobile engines, all relevant failure case data under high-risk scenarios (such as long-term high-speed, heavy-load driving conditions), medium-risk scenarios (such as frequent start-stop conditions on urban roads) and low-risk scenarios (such as short-distance low-speed driving conditions) are screened out. These data cover the time record from the first detection of the defect to the time information that ultimately led to the failure of the object. In the time interval calculation and analysis link, the system calculates the time interval from the first occurrence of the defect to the failure of the object for each screened defect case, uses timestamp calculation technology to accurately obtain the difference between the two time points, and converts it into a unified time unit for storage and analysis.

[0053] For each defect in a large number of cases under different risk scenarios, the system calculates time interval statistics and plots time interval distribution charts (such as bar charts and box plots) to intuitively display the distribution characteristics of the defect evolution time intervals under different risk scenarios. Through these charts, the system can clearly observe the differences in the evolution time intervals of the same defect in high-risk scenarios, medium-risk scenarios, and low-risk scenarios. During the high-risk defect identification and marking stage, the system sets the determination threshold and judgment rules. The average value of the time interval is used as the primary reference indicator, combined with statistics such as standard deviation to determine a reasonable determination threshold. For example, when the average failure time interval of the target defect in the high-risk scenario is more than 30% shorter than the average failure time interval in the medium-risk and low-risk scenarios, and the difference is confirmed to be statistically significant through hypothesis testing, the target defect is determined to be a high-risk defect.

[0054] S204. For defect types marked as high-risk defects, automatically tighten defect assessment criteria in high-risk scenarios; During the strategy formulation phase for adjusting the evaluation criteria, the system determines the direction and magnitude of the evaluation criteria adjustments for different high-risk defect types based on historical failure data, industry standards, and expert experience. For mechanical high-risk defects, such as crack defects in aircraft engine blades, adjustments are made in high-risk, high-temperature, and high-speed flight scenarios by reducing the tolerance for defect size and increasing the limit on the number of defects. Originally, tiny cracks up to 0.5 mm in length were allowed on the blade surface, but this limit has been tightened to no more than 0.3 mm in high-risk scenarios. For electronic high-risk defects, such as chip leakage defects, the detection cycle is shortened and the detection accuracy is improved in high-risk, high-power operation scenarios. For example, the chip leakage detection cycle is shortened from once a month to once a week, while the sensitivity of the leakage detection equipment is increased and the detectable leakage current threshold is lowered.

[0055] The system uses machine learning algorithms to conduct in-depth analysis of large amounts of historical failure data, uncovering the key factors and patterns that influence the performance of objects in high-risk scenarios due to different high-risk defects. Taking crack defects in aircraft engine blades as an example, analysis revealed that factors such as crack length, depth, and location, as well as the temperature and stress environment in which the blades are exposed, are highly correlated with blade failure. Based on these analysis results, the system determines specific adjustment parameters for the evaluation criteria to ensure that the adjusted evaluation criteria better meet actual application requirements and can more accurately assess the degree of harm posed by high-risk defects to objects. During the standard update and implementation phase, the system will update the adjusted evaluation criteria into the defect type standard determination model.

[0056] S205. Monitor the current working scene of the target precision object in real time. When the scene changes from a low-risk scene to a high-risk scene, immediately trigger the defect re-inspection prompt process.

[0057] During the real-time scene data collection phase, the system utilizes a variety of sensors to conduct comprehensive, real-time monitoring of the target precision object's operating environment. Regarding the environment, temperature, humidity, and air pressure sensors are used to collect data such as temperature, humidity, and air pressure. Regarding operating conditions, current, voltage, and speed sensors are used to obtain parameters such as the equipment's operating current, voltage, and speed. Regarding spatial layout, LiDAR, cameras, and other devices are used to collect information about the scene's spatial structure. These sensors transmit the collected data in real time to the defect recognition system at a specific frequency (e.g., once per second or higher).

[0058] The system uses edge computing technology to perform preliminary processing on sensor data, reducing data transmission volume and processing pressure. For example, in industrial production workshops, edge computing devices perform real-time analysis of sensor data, extracting key features such as trends in equipment operating status and abnormal fluctuations in environmental parameters. This processed data is then transmitted to the defect identification system. During the intelligent scenario transition determination phase, the system analyzes the collected scenario data using a pre-trained scenario classification model. This model employs deep learning algorithms (such as convolutional neural networks and recurrent neural networks). During training, it utilizes a large amount of historical data from diverse risk scenarios to accurately identify scenario types. The system sets a scenario transition determination threshold. When the probability of a high-risk scenario output by the scenario classification model exceeds a threshold (e.g., 0.8), the scenario transition is determined to have occurred from low risk to high risk. For example, in a power equipment operation scenario, if a temperature sensor detects a continuous increase in temperature and the scenario classification model determines that the current scenario is a high-risk overheating scenario with a probability of 0.85, the system determines a scenario transition has occurred. Once a scenario transition is determined, the system immediately triggers the defect re-inspection prompt process. A push notification mechanism sends a re-inspection reminder to relevant personnel's devices (e.g., mobile phones and computers). The message includes information such as the target precision object's name and number, the fact that the current scenario has changed to a high-risk scenario, and the recommended re-inspection time. A re-inspection work order is generated in the production management system and automatically assigned to the corresponding inspector. The re-inspection task status is updated to "pending." Upon receiving the reminder, the inspector re-inspects the target precision object according to the prescribed inspection process and standards, ensuring timely detection of potential defects and ensuring the safe operation of the object in high-risk scenarios.

[0059] In the embodiment of the present application, the technical means of obtaining historical failure scenario records of precision objects in the embodiment of the present application to build a failure case library, divide risk scenarios and count the time intervals for defects to evolve into failures, mark high-risk defects, tighten the evaluation criteria in high-risk scenarios, and monitor scenario conversions in real time and trigger re-inspection prompts are adopted. By systematically collecting, analyzing and utilizing historical failure data, the evaluation criteria and detection processes are dynamically adjusted according to risk scenarios, and accurate identification and control of precision object defects in different risk scenarios are achieved. This not only effectively solves the technical problem in the prior art that defect evaluation criteria and detection processes cannot be dynamically adjusted according to the risk of usage scenarios, resulting in inaccurate defect identification in different risk scenarios, but also ensures the safe and stable operation of objects in various scenarios, and significantly improves the quality reliability of precision manufacturing products.

[0060] The following describes the defect recognition system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , is a schematic diagram of the physical device structure of the defect identification system in an embodiment of the present application.

[0061] It should be noted that Figure 3 The structure of the defect identification system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0062] like Figure 3 As shown, the defect identification system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes, such as the methods described in the above embodiments, based on programs stored in a read-only memory (ROM) 302 or programs loaded from a storage unit 308 into a random access memory (RAM) 303. RAM 303 also stores various programs and data required for system operation. CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to bus 304.

[0063] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, push button switches, and the like; an output section 307 including a liquid crystal display (LCD), an audio output device, indicator lights, and the like; a storage section 308 including a hard disk and the like; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card or a modem. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. Removable media 311, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 310 as needed, so that computer programs read from the removable media can be installed in the storage section 308 as needed.

[0064] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for executing the methods illustrated in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 309 and / or installed from removable media 311. When executed by the central processing unit (CPU) 301, the computer program performs the various functions defined in the present invention.

[0065] It should be noted that specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0066] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram can represent a module, program segment, or part of the code, and the above-mentioned module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings.

[0067] Specifically, the defect recognition system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the automatic recognition method for precision manufacturing defects based on artificial intelligence provided by the above embodiment is implemented.

[0068] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the defect recognition system described in the above embodiments, or may exist independently and not be incorporated into the defect recognition system. The storage medium carries one or more computer programs, which, when executed by a processor of the defect recognition system, enable the defect recognition system to implement the artificial intelligence-based automatic identification method for precision manufacturing defects provided in the above embodiments.

[0069] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0070] As used in the above embodiments, the term “when” may be interpreted to mean “if” or “after” or “in response to determining that” or “in response to detecting that”, depending on the context. Similarly, the phrases “upon determining that” or “if (stated condition or event) is detected” may be interpreted to mean “if determining that” or “in response to determining that” or “upon detecting (stated condition or event)” or “in response to detecting (stated condition or event)”, depending on the context.

[0071] Those skilled in the art will appreciate that all or part of the process steps in the above-described method embodiments can be implemented by a computer program instructing the relevant hardware. The program can be stored in a computer-readable storage medium, and when executed, the program can include the process steps in the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. An artificial intelligence-based automatic identification method for precision manufacturing defects, applied to a defect identification system, characterized in that: The method comprises: After obtaining multiple historical object types and corresponding usage scenario data, a defect type standard determination model is constructed based on the defect rating standard annotations of objects of different object types in different usage scenarios; Acquire the current image data of the target precision object through multiple visual sensors, and simultaneously acquire the image data of the usage scene of the target precision object; Inputting the target precision object image data into a preset precision manufacturing object type recognition model to obtain the target precision object type, and inputting the usage scene image data into a preset usage scene recognition model to obtain the current usage scene, wherein the precision manufacturing object type recognition model is pre-trained based on a plurality of image data covering various types of precision manufacturing objects, and the usage scene recognition model is pre-trained based on image data of a plurality of different usage scenes and corresponding scene labels; Inputting the target precision object type and the current usage scenario into the defect type standard determination model to obtain a defect assessment standard that matches the target precision object type in the current usage scenario; The physical feature data of the object is collected by multiple sensors, and the physical feature data of the object is compared with the defect standard to determine the defect recognition result of the target precision object.

2. The method according to claim 1, characterized in that After the steps of collecting physical feature data of the object by using a plurality of sensors, comparing the physical feature data of the object with the defect standard, and determining the defect recognition result of the target precision object, the method further includes: Obtain temperature distribution data and vibration spectrum data in the usage scenario; determining a local thermal stress concentration area of ​​the target precision object based on the temperature distribution data; determining a resonance frequency range of a target precision object based on the vibration spectrum data; Performing spatial mapping on the local thermal stress concentration area and the resonance frequency interval in the same spatial coordinate system to determine an overlapping area; When the defect in the defect recognition result is located in the overlapping area, the defect detection sensitivity of the overlapping area is increased and detection is performed again.

3. The method according to claim 1, characterized in that After the steps of collecting physical feature data of the object by using a plurality of sensors, comparing the physical feature data of the object with the defect standard, and determining the defect recognition result of the target precision object, the method further includes: After determining the defect location according to the defect identification result, applying an excitation signal to the defect location; determining stress wave propagation characteristics corresponding to defects in the target precision object based on the excitation signal; Analyzing the attenuation rate of stress waves in different usage scenarios based on the stress wave propagation characteristics; When there is an abnormality in the decay rate, the defect judgment standard is automatically lowered.

4. The method according to claim 1, wherein After inputting the target precision object type and the current usage scenario into the defect type standard determination model to obtain a defect assessment standard that matches the target precision object type in the current usage scenario, the method further includes: Obtain historical records of failure scenarios of precision objects, including the ambient temperature and consequences of failure, to form a failure case library; Classifying the scenarios in the failure case library into high-risk scenarios, medium-risk scenarios, and low-risk scenarios according to the severity of the failure consequences; Determine the time interval for each defect to evolve into a failure under different risk scenarios. If the time interval for the target defect to fail under a high-risk scenario is shorter than that under a non-high-risk scenario, mark the target defect as a high-risk defect. For defect types marked as high-risk defects, the defect assessment criteria are automatically tightened in high-risk scenarios; Monitor the current working scene of the target precision object in real time. When the scene changes from a low-risk scene to a high-risk scene, the defect review prompt process is immediately triggered.

5. The method according to claim 1, wherein After the steps of collecting physical feature data of the object through a plurality of sensors, comparing the physical feature data of the object with the defect standard, and determining the defect recognition result of the target precision object, the method further includes: Obtain physical feature data of the target precision object at preset time intervals to establish a time series data set; Calculating a rate of change of a physical characteristic based on the time series data set; When the rate of change of the physical feature exceeds a preset rate threshold, the corresponding physical feature is marked as a rapidly degrading feature, and the frequency of collecting the physical feature is increased; Determining a predicted future trend value of the rapid degradation feature using a feature development prediction model, wherein the feature development prediction model is previously trained by machine learning based on physical feature data of multiple target precision objects and corresponding annotations of defect development conditions; If the deviation between the future trend prediction value of the feature and the current defect judgment standard shows an increasing trend, the early warning device is controlled to issue a potential defect early warning signal.

6. The method according to claim 1, wherein After the steps of collecting physical feature data of the object by using a plurality of sensors, comparing the physical feature data of the object with the defect standard, and determining the defect recognition result of the target precision object, the method further includes: Construct a three-dimensional digital model of the target precision object based on the surface contour data, internal structure data and material property data of the target precision object collected in advance; Collecting spatial layout data, environmental parameter data, and working condition data of the current usage scene, and spatially mapping these scene feature data with the three-dimensional digital model in the same coordinate system to form a 3D display scene; The defect recognition results are visually annotated and integrated into the 3D display scene for display.

7. The method according to claim 1, characterized in that After the steps of collecting physical feature data of the object by using a plurality of sensors, comparing the physical feature data of the object with the defect standard, and determining the defect recognition result of the target precision object, the method further includes: Collecting stress distribution data around the defect according to the defect identification result to obtain defect characteristic data; Counting the frequencies of different defect types in the defect feature data appearing simultaneously on the same object, and calculating the correlation between the defect types; When the correlation exceeds a preset correlation threshold, establishing a correlation relationship between the defect types; Analyze the order of occurrence of defects with the aforementioned correlation in a time series to determine the defect evolution path; Determine the related defects derived from the current defect according to the defect evolution path; And improve the detection frequency of related equipment for the said derivative related defects.

8. A defect identification system, characterized in that: The defect identification system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the defect identification system to execute the method described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on a defect recognition system, the defect recognition system is caused to perform the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a defect recognition system, the defect recognition system is caused to perform the method according to any one of claims 1 to 7.

Images