A method and system for detecting defects in an ultra-precision spring

Abstract

The application relates to the technical field of defect detection, and provides a super-precision spring defect detection method and system, which comprises the following steps: obtaining surface deformation information of a to-be-detected super-precision diaphragm spring; extracting current defect feature data from the surface deformation information; storing the current defect feature data into a defect feature cache area; comparing the current defect feature data with historical defect feature data stored in the defect feature cache area to determine whether the current defect feature data has spatial and temporal constancy, and obtaining a spatial and temporal constancy judgment result; according to the spatial and temporal constancy judgment result, marking a defect type corresponding to the current defect as a system artifact or a real product defect; generating a detection report, and displaying the defect type mark based on preset display rules for distinguishing the system artifact and the real product defect. The application has the advantages of distinguishing the system artifact and the real product defect, improving the accuracy and efficiency of detection, and reducing unnecessary production interruption and verification cost.

Description

Technical Field

[0001] This invention relates to the technical field of defect detection, and specifically to a method and system for detecting defects in ultra-precision springs. Background Technology

[0002] In the field of precision manufacturing, residual stress detection of ultra-precision diaphragm springs is a crucial step in ensuring product quality. Laser speckle interferometry, due to its non-contact and high-sensitivity characteristics, is often used to assess the stress distribution inside springs, thereby identifying potential structural defects.

[0003] However, in practical applications, the optical components of a measurement system, especially the lens surface of an image acquisition camera, may be affected by minute contamination. This contamination may originate from non-standard cleaning agents used in routine maintenance, whose residues can form a thin, unevenly distributed transparent film on the lens surface that is difficult to detect with the naked eye. This film introduces a fixed, position-dependent distortion to the phase information of the measurement laser beam. Existing systems' self-calibration mechanisms typically only compensate for light signal intensity attenuation and cannot identify or correct this phase distortion. Therefore, the system may misinterpret the fixed phase distortion introduced by optical component contamination as a genuine deformation of the spring body, resulting in false defects in the inspection report that do not match the actual situation. Due to the repeatability of its shape and location, this false defect is easily misjudged by experienced technicians as a systemic problem in the manufacturing process, leading to unnecessary production stoppages and costly verification.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a method and system for detecting defects in ultra-precision springs, which has the advantages of improving the accuracy and efficiency of detection and reducing unnecessary production interruptions and verification costs.

[0006] This application provides a method for detecting defects in ultra-precision springs, the key technical points of which are:

[0007] A method for detecting defects in ultra-precision springs includes the following steps:

[0008] Obtain surface deformation information of the ultra-precision diaphragm spring under test;

[0009] Extract current defect feature data from surface deformation information;

[0010] Store the current defect feature data in the defect feature cache area;

[0011] The current defect feature data is compared with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy, and the spatiotemporal constancy judgment result is obtained.

[0012] Based on the spatiotemporal constancy judgment result, the defect type corresponding to the current defect is marked as either a system artifact or a real product defect;

[0013] Generate an inspection report and display defect type markers based on preset display rules that differentiate between system artifacts and real product defects.

[0014] The above solution can effectively distinguish between system artifacts and real product defects, avoid misjudgments, and improve detection accuracy.

[0015] To further address this issue, this application also proposes a step of comparing the current defect feature data with historical defect feature data stored in the defect feature cache to determine whether the current defect feature data possesses spatiotemporal constancy, and obtaining the spatiotemporal constancy determination result, including:

[0016] Obtain the currently detected laser power;

[0017] Adjust the current defect feature data or the historical defect feature data stored in the defect feature cache based on the laser power;

[0018] The adjusted current defect feature data is compared with the adjusted historical defect feature data to determine whether the current defect feature data has spatiotemporal constancy, and the spatiotemporal constancy judgment result is obtained.

[0019] By considering the impact of laser power on defect feature data, the above scheme can further improve the accuracy of spatiotemporal constancy determination.

[0020] To improve the solution, this application also proposes a step of adjusting the current defect feature data or historical defect feature data stored in the defect feature cache according to the laser power, including:

[0021] Based on artifacts in the current defect feature data or historical defect feature data stored in the defect feature cache, the spatial distribution area of ​​the artifacts is identified in the image.

[0022] Determine the local laser power sensitivity at each location within the spatial distribution area of ​​artifacts;

[0023] Based on the laser power and local laser power sensitivity, spatial location-related adjustments are made to the current defect feature data or the historical defect feature data stored in the defect feature cache.

[0024] By introducing local laser power sensitivity through the above scheme, more precise spatial position-related adjustments can be achieved, thereby improving the accuracy of artifact correction.

[0025] To further address the problem, this application also proposes steps for determining the local laser power sensitivity at various locations within the spatial distribution area of ​​artifacts, including:

[0026] Obtain artifact feature data and corresponding laser power data at each location within the spatial distribution area of ​​the artifacts;

[0027] Based on artifact feature data and laser power data, the ratio of the change in artifact features to the change in laser power at each location within the spatial distribution area of ​​the artifact is calculated to obtain the local laser power sensitivity at each location within the spatial distribution area of ​​the artifact.

[0028] Based on the calculated results, the local laser power sensitivity at each location within the spatial distribution area of ​​the artifacts is updated and confirmed.

[0029] The above scheme accurately quantifies the local laser power sensitivity by calculating the ratio of the change in artifact characteristics to the change in laser power, providing a basis for precise adjustment.

[0030] To improve the solution, this application also proposes steps for spatially adjusting the current defect feature data or historical defect feature data stored in the defect feature cache, including:

[0031] The spatial distribution area of ​​artifacts is divided into several tiny spatial units;

[0032] The artifact feature correction amount of the micro spatial unit is calculated based on the local laser power sensitivity and the corresponding laser power of the micro spatial unit.

[0033] The artifact feature correction is applied to the feature value of the corresponding small spatial unit in the current defect feature data or the historical defect feature data stored in the defect feature cache, so as to realize the spatial location-related adjustment of the current defect feature data or the historical defect feature data stored in the defect feature cache.

[0034] The above scheme applies artifact feature correction to small spatial units, enabling refined and localized adjustments and further improving the correction effect.

[0035] To further address the problem, this application also proposes a method for calculating the artifact feature correction amount of a small spatial unit, including:

[0036] Obtain the local laser power sensitivity corresponding to a tiny spatial unit;

[0037] Obtain the current laser power;

[0038] Based on the local laser power sensitivity corresponding to the micro spatial unit and the current laser power, the artifact feature correction amount of the micro spatial unit is determined by consulting a pre-stored correction amount lookup table or applying a pre-calibrated correction curve.

[0039] The above scheme simplifies the process of determining artifact feature correction amounts and improves efficiency by consulting a pre-stored correction amount lookup table or applying a pre-calibrated correction curve.

[0040] To improve the solution, this application also proposes a step for determining whether the current defect feature data has spatiotemporal constancy, and obtaining the spatiotemporal constancy determination result, including:

[0041] Quantify the similarity between the current defect feature data and the historical defect feature data stored in the defect feature cache;

[0042] Based on the relationship between similarity and a preset similarity threshold, it is determined whether the current defect feature data has spatiotemporal constancy, and the spatiotemporal constancy judgment result is obtained.

[0043] The above scheme provides an objective method for judging spatiotemporal constancy by quantifying similarity and comparing it with a threshold, thereby enhancing the reliability of the judgment.

[0044] To further address the problem, this application also proposes a step to determine whether the current defect feature data possesses spatiotemporal constancy based on the relationship between similarity and a preset similarity threshold, and to obtain the spatiotemporal constancy determination result, including:

[0045] Obtain defect feature data of the reference part;

[0046] By comparing the defect feature data of the reference part with the pre-stored reference part standard data, the feature deviation of the reference part is obtained;

[0047] Based on the feature deviation of the reference part, the preset initial threshold is corrected to obtain the corrected preset similarity threshold;

[0048] Based on the relationship between similarity and a preset similarity threshold, it is determined whether the current defect feature data has spatiotemporal constancy, and the spatiotemporal constancy judgment result is obtained.

[0049] The above scheme introduces a reference part feature deviation correction preset threshold, making the judgment threshold more adaptable and improving the accuracy of the judgment.

[0050] To improve the solution, this application also proposes a step of correcting a preset initial threshold based on the feature deviation of the reference part to obtain a corrected preset similarity threshold, including:

[0051] Obtain the positional deviation, morphological deviation, and stress deviation from the feature deviations of the reference part;

[0052] Based on the combination of position deviation, shape deviation and stress deviation, consult the pre-stored correction mapping table to determine the correction amount for the corresponding position, shape and stress components in the preset initial threshold.

[0053] The determined correction values ​​are applied to the corresponding positions, shapes, and stress components in the initial threshold to obtain the corrected preset similarity threshold.

[0054] By considering the influence of position, shape, and stress deviation on the threshold, the above scheme achieves multi-dimensional and refined threshold correction, further improving the judgment accuracy.

[0055] To further address the problem, this application also proposes an ultra-precision spring defect detection system for performing ultra-precision spring defect detection, comprising:

[0056] The surface deformation acquisition module is used to acquire the surface deformation information of the ultra-precision diaphragm spring under test;

[0057] The defect feature extraction module is used to extract the current defect feature data from the surface deformation information;

[0058] The feature data storage module is used to store the current defect feature data in the defect feature cache area;

[0059] The spatiotemporal constancy judgment module is used to compare the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy and obtain the spatiotemporal constancy judgment result.

[0060] The defect type determination module is used to mark the defect type corresponding to the current defect as a system artifact or a real product defect based on the spatiotemporal constancy determination result.

[0061] The test result display module is used to generate test reports and display defect type markers based on preset display rules that distinguish between system artifacts and real product defects.

[0062] The above scheme provides a system for implementing the aforementioned defect detection method, which is convenient for practical deployment and application, and improves the automation and reliability of detection.

[0063] In summary, the ultra-precision spring defect detection method and system provided in this application compares the spatiotemporal constancy of current defect feature data with historical defect feature data, identifies the defect type as either a system artifact or a real product defect, and displays them separately. This effectively distinguishes between system artifacts and real product defects, avoids misjudgment, improves the accuracy and efficiency of detection, and reduces unnecessary production interruptions and verification costs. Attached Figure Description

[0064] Figure 1This is a flowchart of a method for detecting defects in ultra-precision springs according to one embodiment of the present invention;

[0065] Figure 2 This is one of the flowcharts of a method for detecting defects in ultra-precision springs according to another embodiment of the present invention;

[0066] Figure 3 This is a second flowchart of a method for detecting defects in ultra-precision springs according to another embodiment of the present invention;

[0067] Figure 4 This is the third flowchart of a method for detecting defects in ultra-precision springs according to another embodiment of the present invention;

[0068] Figure 5 This is the fourth flowchart of a method for detecting defects in ultra-precision springs according to another embodiment of the present invention;

[0069] Figure 6 This is the fifth flowchart of a method for detecting defects in ultra-precision springs according to another embodiment of the present invention;

[0070] Figure 7 This is the sixth flowchart of a method for detecting defects in ultra-precision springs according to another embodiment of the present invention;

[0071] Figure 8 This is a system block diagram of an ultra-precision spring defect detection system according to another embodiment of the present invention;

[0072] Explanation of reference numerals in the attached figures:

[0073] 1. Ultra-precision spring defect detection system; 11. Surface deformation acquisition module; 12. Defect feature extraction module; 13. Feature data storage module; 14. Spatiotemporal constancy judgment module; 15. Defect type determination module; 16. Detection result display module. Detailed Implementation

[0074] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0075] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0076] Traditional ultra-precision spring defect detection methods, when using laser speckle interferometry to assess spring stress distribution, suffer from a problem where phase distortion introduced by a non-uniformly distributed transparent film formed on the surface of the image acquisition optical element is misidentified as spring deformation. Existing systems' self-calibration mechanisms typically only compensate for optical signal intensity attenuation and cannot identify or correct such phase distortions. This misidentification leads to false defects in the inspection report that do not match the actual situation. Furthermore, due to the repetitive nature of these false defects in shape and location, they are easily misdiagnosed as systemic problems in the manufacturing process. When the root cause information is blocked by external factors, the troubleshooting path fails, leading to a stalemate in problem diagnosis.

[0077] In response, this application proposes a method for detecting defects in ultra-precision springs, combining... Figure 1 As shown, it includes the following steps:

[0078] S1, Obtain the surface deformation information of the ultra-precision diaphragm spring to be tested;

[0079] S2, extract current defect feature data from surface deformation information;

[0080] S3, store the current defect feature data in the defect feature cache area;

[0081] S4. Compare the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy and obtain the spatiotemporal constancy judgment result.

[0082] S5. Based on the spatiotemporal constancy judgment result, mark the defect type corresponding to the current defect as either a system artifact or a real product defect.

[0083] S6 generates an inspection report and displays defect type markers based on preset display rules that differentiate between system artifacts and real product defects.

[0084] Surface deformation information refers to the data on changes in the surface geometry or microstructure of the ultra-precision diaphragm spring under stress or no stress. This data can be obtained using non-contact optical inspection techniques such as laser speckle interferometry, optical profilometry, or white light interferometry, providing a data foundation for subsequent defect identification and analysis. Current defect characteristic data refers to the quantitative data extracted from the surface deformation information that characterizes the defect attributes. This data may include the defect's location coordinates, geometric dimensions, morphological features, intensity distribution, or stress concentration. Its main purpose is to transform the raw deformation data into structured information for subsequent defect identification and classification. Spatiotemporal constancy refers to the property that the spatial location and morphological characteristics of the defect characteristic data remain consistent or similar across different time points or different inspection batches. This can be determined based on the similarity between current and historical inspection results, aiming to distinguish artifacts caused by system errors (such as optical contamination) from genuine product defects. System artifacts refer to false defect markers that appear in the inspection report due to defects in the detection system itself or interference from the external environment (such as contamination of optical components). They usually have spatiotemporal constancy. True product defects refer to structural or performance abnormalities of the ultra-precision diaphragm spring body caused by materials, processes or usage. They usually do not have spatiotemporal constancy or are random. Their purpose is to accurately classify the detected abnormalities and avoid misjudgment.

[0085] In some embodiments, to further clarify the above working principle, this application is specifically implemented as follows. For example, when acquiring the surface deformation information of the ultra-precision diaphragm spring to be tested, a laser measurement system based on the principles of digital holography or speckle interferometry can be used. A laser beam is emitted to illuminate the spring surface, and a high-resolution CCD camera is used to capture the reflected or transmitted speckle image. Then, the three-dimensional deformation field data of the spring surface is calculated using a phase unwrapping algorithm. When extracting current defect feature data from this surface deformation information, image processing software can be used, for example, by setting thresholds, edge detection, connected component analysis, and other algorithms, to identify abnormal regions in the deformation field and extract the geometric features (such as area, perimeter, shape factor), location information (such as center coordinates), and deformation intensity of these regions as defect feature data. When storing the current defect feature data in the defect feature cache, a structured data storage system (such as an SQL database or NoSQL database) can be used for management, where each defect record contains its feature data, detection timestamp, and corresponding product batch information. When comparing current defect feature data with historical defect feature data stored in the defect feature cache to determine spatiotemporal constancy, a feature-matching algorithm can be used. For example, the similarity of spatial location and morphological features between the current defect feature data and the defect data from the past N detections in the cache can be calculated. If the number of times morphologically similar defects appear at similar locations exceeds a pre-set frequency, it is determined to have spatiotemporal constancy. Based on the spatiotemporal constancy determination result, when classifying the defect type corresponding to the current defect as a system artifact or a real product defect, a rule engine can be set. For example, if it is determined to have spatiotemporal constancy, it is marked as a system artifact; otherwise, it is marked as a real product defect. Finally, when generating the inspection report and displaying the defect type markings, a graphical user interface (GUI) software can be used to visualize the detected defects on the 3D model of the spring, using different colors or icons to distinguish between system artifacts (e.g., marked in blue) and real product defects (e.g., marked in red). The report also provides a defect list and classification statistics.

[0086] Optional, combined Figure 2 As shown, step S4 compares the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy, and obtains the spatiotemporal constancy determination result. This step includes:

[0087] S41, obtain the currently detected laser power;

[0088] S42, adjust the current defect feature data or the historical defect feature data stored in the defect feature cache according to the laser power;

[0089] S43. Compare the adjusted current defect feature data with the adjusted historical defect feature data to determine whether the current defect feature data has spatiotemporal constancy, and obtain the spatiotemporal constancy judgment result.

[0090] Adjusting the current defect feature data or the historical defect feature data stored in the defect feature cache refers to correcting the defect feature data based on changes in laser power to eliminate or reduce the impact of laser power fluctuations on data features, making them comparable under different power conditions. Specifically, this can be done by establishing a mathematical model or functional relationship between laser power and defect feature data, such as a linear regression model, a multinomial model, or a nonlinear mapping relationship, and then normalizing or compensating the data based on the model. Alternatively, it can be done by pre-calibrating standard defect feature data under different laser powers, constructing a lookup table, and correcting the data based on the current laser power during actual detection. Or, it can be done by using machine learning methods to train a model to learn the nonlinear relationship between laser power and defect feature data, and then using the model for predictive adjustments. The goal is to ensure that the basis for comparing defect feature data remains consistent under changes in laser power, thereby improving the accuracy of spatiotemporal constancy judgment.

[0091] In some preferred embodiments, this application is implemented as follows. When acquiring the currently detected laser power, an optical power sensor integrated in the laser emitter or optical path can be used to monitor and output the power value of the laser beam in real time. This sensor can be a photodiode whose output voltage is proportional to the received optical power, and an analog-to-digital converter converts the analog signal into a digital power value. When adjusting the current defect feature data or historical defect feature data stored in the defect feature buffer based on the laser power, system calibration can be performed beforehand. For example, in a controlled environment, using known and stable artifacts as test objects, multiple detections are performed at multiple different laser power levels (e.g., from the lowest operating power to the highest operating power, increasing in fixed steps), and the artifact feature data detected each time is recorded. Based on these calibration data, a correction model between the laser power and the artifact feature data can be established, for example, by fitting a polynomial function using the least squares method, or by constructing a multidimensional lookup table. When the system acquires the current defect feature data during actual detection, it will simultaneously acquire the current laser power. The system can utilize a pre-established correction model or lookup table to uniformly adjust the current defect feature data or historical defect feature data stored in the defect feature cache to an equivalent feature value under a preset standard laser power, based on their corresponding laser power. For example, if the standard power is set to P_std, the current power is P_curr, and the historical power is P_hist, then the current data F_curr can be adjusted to F_curr_adj = F_curr * (P_std / P_curr_factor), and the historical data F_hist can be adjusted to F_hist * (P_std / P_hist_factor), where factor is a power-related correction factor calculated based on the correction model. When comparing the adjusted current defect feature data with the adjusted historical defect feature data, various similarity or difference measurement methods can be used. For example, the Euclidean distance, cosine similarity, or Pearson correlation coefficient between the adjusted current defect feature data and the adjusted historical defect feature data can be calculated. If the calculated similarity is higher than the preset similarity threshold, or the difference is lower than the preset difference threshold, then the current defect feature data is judged to have spatiotemporal constancy; otherwise, it is not.

[0092] Optional, combined Figure 3 As shown, the step S42, which adjusts the current defect feature data or the historical defect feature data stored in the defect feature cache according to the laser power, includes:

[0093] S421, Based on artifacts in the current defect feature data or historical defect feature data stored in the defect feature cache, identify the spatial distribution area of ​​artifacts in the image;

[0094] S422, determine the local laser power sensitivity at each location within the spatial distribution area of ​​the artifact;

[0095] S423, based on the laser power and local laser power sensitivity, makes spatial location-related adjustments to the current defect feature data or the historical defect feature data stored in the defect feature cache.

[0096] Identifying the spatial distribution of artifacts refers to locating and delineating abnormal regions caused by system artifacts in image data by analyzing defect feature data. The purpose is to clarify the specific range requiring correction, avoiding unnecessary processing of non-artifact areas and ensuring the accuracy of the adjustment. Specifically, this can be achieved through image processing algorithms such as edge detection, region growing, and threshold segmentation to identify artifact regions with specific morphological or intensity characteristics; or by using pattern matching or machine learning methods based on a pre-established artifact pattern library. Local laser power sensitivity refers to the degree to which the feature values ​​of artifacts at different spatial locations in an image change with varying laser power. Its purpose is to reflect the differences in the response of artifacts to changes in laser power in different regions, providing a basis for subsequent refined correction. Specifically, this can be achieved through experimental calibration, collecting artifact data at different laser powers and then calculating the ratio of artifact feature change to laser power change; or by establishing a mathematical model and fitting historical data to predict the sensitivity at different locations. Spatial location-dependent adjustments refer to adjustments made to defect feature data or historical defect feature data. The adjustment amount or method is not globally uniform, but varies according to the specific location of the artifact in the image and its corresponding local laser power sensitivity. The aim is to make targeted adjustments based on the characteristics of artifacts at different spatial locations, avoiding over- or under-adjustment issues that may occur with global adjustments. Specifically, this can involve dividing the artifact region into multiple sub-regions and applying different adjustment parameters to each sub-region; alternatively, it can involve establishing a continuous adjustment function that takes spatial coordinates as input and outputs the corresponding adjustment amount.

[0097] In some preferred embodiments, the specific process of adjusting the current defect feature data or historical defect feature data stored in the defect feature cache according to the laser power can be implemented as follows: First, the spatial distribution region of artifacts is identified in the image. This can be done based on the analysis of the defect feature data. For example, the system can pre-store a series of feature templates for known artifacts, and when a pattern that highly matches these templates is detected in the current defect feature data, it is identified as an artifact. Alternatively, image segmentation algorithms, such as threshold segmentation or region growing, can be used to mark regions in the image with abnormal brightness, texture, or shape features as artifact regions. Furthermore, a deep learning model can be trained to learn and identify specific spatial distribution patterns of artifacts from a large amount of historical data. Further, the local laser power sensitivity at each location within the spatial distribution region of the artifacts is determined. This can be done step by step. For example, the system can acquire artifact feature data and corresponding laser power data at each location within the spatial distribution region of the artifacts. This data can be obtained from historical detection records or through specialized calibration experiments. Then, the system can calculate the ratio of the change in artifact features to the change in laser power at each location within the spatial distribution area of ​​the artifact, thereby obtaining the local laser power sensitivity at each location. For example, regression analysis can be performed on the artifact feature value and laser power for each pixel or each small region to obtain a sensitivity coefficient. Based on the calculated results, the system can update and confirm these local laser power sensitivities, for example, by smoothing or statistical averaging to improve their stability. Finally, based on the laser power and local laser power sensitivity, spatial location-related adjustments are made to the current defect feature data or the historical defect feature data stored in the defect feature cache. Specifically, the spatial distribution area of ​​the artifact can be spatially divided into several small spatial units, for example, dividing the artifact region into a grid-like pixel block or small region. Then, for each small spatial unit, the system can calculate the artifact feature correction amount for that small spatial unit based on its corresponding local laser power sensitivity and corresponding laser power. For example, a pre-stored correction lookup table can be consulted, which records correction values ​​for different sensitivities and laser powers; or a pre-calibrated correction curve can be applied, which describes the functional relationship between the correction amount and sensitivity and laser power. Finally, the calculated artifact feature correction amount is applied to the feature value of the corresponding small spatial unit in the current defect feature data or historical defect feature data stored in the defect feature cache, to achieve spatial location-related adjustments to the data. For example, if the artifact feature value of a small spatial unit is too high due to laser power fluctuations, a negative correction amount is calculated based on its local sensitivity and the current laser power, and subtracted from the feature value to correct the deviation.

[0098] Optional, combined Figure 4As shown, the steps for determining the local laser power sensitivity at various locations within the spatial distribution area of ​​artifacts include:

[0099] S4221, Obtain artifact feature data and corresponding laser power data at each location within the spatial distribution area of ​​the artifact;

[0100] S4222, Based on the artifact feature data and laser power data, calculate the ratio of the change in artifact features to the change in laser power at each location within the spatial distribution area of ​​the artifact, and obtain the local laser power sensitivity at each location within the spatial distribution area of ​​the artifact.

[0101] S4223, based on the calculated results, updates and confirms the local laser power sensitivity at each location within the spatial distribution area of ​​the artifact.

[0102] The artifact feature data at each location within the spatial distribution area of ​​the artifact refers to the artifact-related values ​​exhibited by each specific pixel or micro-region within the artifact region identified in the image, such as brightness, contrast, texture, phase distortion degree, or deformation amplitude. This can be achieved through pixel-level analysis or regional statistics of the artifact region using image processing algorithms, with the aim of quantifying the specific performance of artifacts at different spatial locations. The corresponding laser power data refers to the instantaneous output power value of the laser beam measured synchronously at each artifact location when acquiring the artifact feature data. This can be achieved through real-time monitoring along the laser path using a high-precision optical power meter and synchronous recording with the image acquisition system, with the aim of establishing the correlation between artifact features and the laser energy input that causes the artifacts. The artifact feature variation refers to the numerical difference in artifact feature data at the same spatial location under different laser power conditions. This can be achieved by calculating the difference or ratio between the current artifact feature data and historical or reference artifact feature data, with the aim of reflecting the degree of response of the artifact features to laser power fluctuations. The change in laser power refers to the numerical difference in laser power data at different measurement times or under different test conditions. It can be calculated by comparing the current laser power data with historical or reference laser power data using difference or ratio calculations. The purpose is to quantify the laser energy input fluctuations that cause changes in artifact characteristics. The local laser power sensitivity at each location within the spatial distribution area of ​​the artifact refers to the ratio between the change in artifact characteristics at each specific location within the spatial distribution area of ​​the artifact and the corresponding change in laser power. This ratio characterizes the response intensity of the artifact characteristics at that location to laser power fluctuations. It can be achieved using mathematical model fitting, regression analysis, or direct ratio calculations. The purpose is to accurately quantify the degree of sensitivity of artifacts to laser power differences at different spatial locations, providing a basis for subsequent accurate calibration.

[0103] In some preferred embodiments, determining the local laser power sensitivity at various locations within the spatial distribution area of ​​artifacts can be achieved as follows: First, during system calibration or a specific test mode, the laser power can be actively and incrementally changed within a certain range, for example, from P_min to P_max, and the artifact image corresponding to each power point can be recorded. For each pixel or predefined microgrid unit within the spatial distribution area of ​​artifacts, artifact feature data, such as pixel intensity, phase distortion value, or deformation amplitude, can be extracted from these images. Simultaneously, synchronized with each image acquisition, the current laser power data can be accurately acquired and recorded. Next, for each location within the spatial distribution area of ​​artifacts, at least two sets of artifact feature data and corresponding laser power data under different laser powers can be selected. For example, for a certain pixel, when the laser power changes from P1 to P2, its artifact feature data changes from F1 to F2. At this time, the ratio of the change in artifact feature data (F2-F1) to the change in laser power (P2-P1) at that pixel can be calculated, thereby obtaining the local laser power sensitivity at that location. This process can be repeated multiple times to obtain more stable sensitivity values. Finally, based on these calculations, the local laser power sensitivity can be updated and confirmed. For example, a time-series averaging method can be used, that is, averaging the sensitivity values ​​calculated at the same location at different time points to eliminate random noise; or a spatial smoothing algorithm, such as two-dimensional Gaussian filtering, can be applied to smooth the sensitivity values ​​at adjacent locations, thereby obtaining a more stable and reliable local laser power sensitivity map of each location within the spatial distribution area of ​​artifacts.

[0104] Optional, combined Figure 5 As shown, step S423, which involves making spatial location-related adjustments to the current defect feature data or the historical defect feature data stored in the defect feature cache, includes:

[0105] S4231, spatially divides the spatial distribution area of ​​artifacts into several tiny spatial units;

[0106] S4232, calculate the artifact feature correction amount of the micro spatial unit based on the local laser power sensitivity and the corresponding laser power of the micro spatial unit;

[0107] S4233, apply the artifact feature correction amount to the feature value of the corresponding small spatial unit in the current defect feature data or the historical defect feature data stored in the defect feature cache, so as to realize the spatial location-related adjustment of the current defect feature data or the historical defect feature data stored in the defect feature cache.

[0108] Spatial partitioning refers to dividing a continuous or discrete region into multiple smaller, independent sub-regions according to preset rules or algorithms. Methods such as grid partitioning, feature-based region growing, or clustering can be used. The aim is to perform more detailed localization processing on areas affected by artifacts. A micro-spatial unit is the smallest processing unit with a specific size and location formed through spatial partitioning. It can be a pixel, a pixel block, a predefined small region, or an irregular local region. Its purpose is to achieve refined correction of artifact features. The artifact feature correction amount refers to the numerical value or vector used to eliminate or mitigate the impact of artifacts on defect feature data. It can be a scalar correction value, a scaling factor, or a complex transformation matrix. Its purpose is to correct defect feature data interfered with by artifacts. An eigenvalue refers to the quantized value in the defect feature data corresponding to a specific spatial location or micro-spatial unit. It can be a pixel intensity value, a stress value, a deformation value, or a statistical feature value. Its purpose is to represent the defect feature information at that location.

[0109] In some preferred embodiments, when the system identifies a circular artifact region in an image caused by lens contamination, the system can first spatially divide the circular artifact region to refine its correction. Specifically, the circular region can be divided into several 10x10 pixel micro-spatial units. For each such micro-spatial unit, the system acquires its pre-calibrated local laser power sensitivity, for example, by taking multiple measurements of the unit at different laser powers and recording the ratio of the change in artifact features to the change in laser power. Simultaneously, the system acquires the real-time laser power during defect detection. Subsequently, based on the local laser power sensitivity of the micro-spatial unit and the current laser power, the system calculates the artifact feature correction amount for that micro-spatial unit by consulting a pre-stored correction amount lookup table. For example, if a unit has high sensitivity and the current laser power deviates significantly from the standard value, its correction amount can be larger. Finally, the system applies the calculated artifact feature correction amount to the feature values ​​in the corresponding current defect feature data or historical defect feature data within that micro-spatial unit. For example, if the feature value is pixel intensity, the correction amount can be an increment or scaling factor of the intensity value. By adding or multiplying the correction amount to the original feature value, the defect feature data of that unit is corrected, making it closer to the feature in a true artifact-free state. By performing independent corrections on each tiny spatial unit within the artifact region based on its local sensitivity, the interference of artifacts on defect features can be effectively eliminated, significantly improving the accuracy of defect detection.

[0110] Optionally, the steps for calculating the artifact feature correction for small spatial cells include:

[0111] Obtain the local laser power sensitivity corresponding to a tiny spatial unit;

[0112] Obtain the current laser power;

[0113] Based on the local laser power sensitivity corresponding to the micro spatial unit and the current laser power, the artifact feature correction amount of the micro spatial unit is determined by consulting a pre-stored correction amount lookup table or applying a pre-calibrated correction curve.

[0114] Among them, the local laser power sensitivity corresponding to a micro spatial unit refers to the degree to which the artifact characteristic value in a specific micro spatial region changes with the laser power in a laser speckle interferometry system. It can be characterized by the ratio of the change in artifact characteristics to the change in laser power, and its purpose is to quantify the difference in response to laser power fluctuations at different spatial locations. The current laser power refers to the instantaneous laser energy intensity actually output by the laser during defect detection. It can be measured in real time by a laser power meter or obtained from the system control unit, and its purpose is to provide real-time input parameters for artifact characteristic correction calculation. The pre-stored correction lookup table refers to a data structure that is pre-established and stored in the system memory or database, which contains different local laser power sensitivities and laser power... The artifact feature correction amount corresponding to the rate combination can be stored in the form of a two-dimensional array or hash table, with the aim of achieving efficient and accurate acquisition of the correction amount through quick lookup; the pre-calibrated correction curve refers to the mathematical expression describing the functional relationship between local laser power sensitivity, laser power and artifact feature correction amount through experimental calibration or mathematical modeling. It can take the form of polynomial function, exponential function or piecewise linear function, with the aim of accurately estimating the artifact feature correction amount through function calculation; the artifact feature correction amount refers to the value used to correct or compensate for the deviation of defect feature data introduced by system artifacts. It can take the form of deformation value, stress value or pixel intensity value, with the aim of eliminating the interference of artifacts on defect detection results and improving detection accuracy.

[0115] In some preferred embodiments, calculating the artifact feature correction for micro-spatial units can be implemented as follows. First, during system initialization or periodic calibration, a series of experimental calibrations can be performed. For example, under controlled conditions, the laser power is varied, and artifact feature data for different micro-spatial units at different laser powers are measured. This data can be used to construct a correction lookup table or correction curve.

[0116] Specifically, artifact feature measurements can be performed on each micro-spatial unit (e.g., unit A, unit B, unit C, etc.) at multiple discrete laser power points (e.g., P1, P2, P3, etc.), and the corresponding artifact feature correction amount can be calculated by combining this with its local laser power sensitivity. These correction amounts can be stored in a two-dimensional lookup table, where rows represent different micro-spatial units or their local laser power sensitivity ranges, columns represent different laser power ranges, and the values ​​in the table are the corresponding artifact feature correction amounts. During actual detection, the system obtains the local laser power sensitivity of the current micro-spatial unit and the current laser power. Then, the system can quickly retrieve the corresponding artifact feature correction amount from the lookup table based on the obtained local laser power sensitivity and the current laser power, using interpolation or direct matching.

[0117] As another specific implementation method, the artifact feature correction amount can also be determined through a pre-calibrated correction curve. For example, experimental data can be fitted to a mathematical function curve through regression analysis or curve fitting, which describes the relationship between the artifact feature correction amount and local laser power sensitivity and laser power. For example, a multivariate function F(S,P)=M can be established, where S represents local laser power sensitivity, P represents laser power, and M represents the artifact feature correction amount. During actual detection, the system obtains the current local laser power sensitivity S and laser power P, and then substitutes these values ​​into the pre-calibrated correction curve function F to calculate the accurate artifact feature correction amount M. Through these two methods, the system can ensure accurate correction of artifact features under different operating conditions, thereby improving the reliability of defect detection.

[0118] Optional, combined Figure 6 As shown, the steps to determine whether the current defect feature data has spatiotemporal constancy and to obtain the spatiotemporal constancy determination result include:

[0119] A1 quantifies the similarity between the current defect feature data and the historical defect feature data stored in the defect feature cache;

[0120] A2. Based on the relationship between similarity and preset similarity threshold, determine whether the current defect feature data has spatiotemporal constancy and obtain the spatiotemporal constancy judgment result.

[0121] Quantifying the similarity between current defect feature data and historical defect feature data stored in the defect feature cache refers to converting the degree of similarity or difference between two or more data objects into a numerical value through mathematical methods. Various measurement methods can be used, such as calculating Euclidean distance, Manhattan distance, cosine similarity, Pearson correlation coefficient, or structural similarity index (SSIM). The purpose is to transform complex differences in defect feature data into comparable numerical values, providing a basis for subsequent automated judgment.

[0122] The preset similarity threshold refers to a threshold value that is set in advance to distinguish between "similar" and "dissimilar" when making similarity judgments. Specifically, it can be obtained based on historical data analysis, expert experience, or training through machine learning models. For example, it can be set as a certain percentage value or a specific numerical range. Its purpose is to provide an objective judgment standard, avoid errors caused by subjective judgment, and allow the system to adjust the strictness of the judgment according to actual needs.

[0123] In some preferred embodiments, this application is implemented as follows. Assume that both the current defect feature data and the historical defect feature data stored in the defect feature cache are represented as multi-dimensional vectors. For example, each vector contains the defect's location coordinates, morphological parameters (such as area, perimeter, and aspect ratio), and stress distribution characteristics (such as maximum stress value and stress gradient). When determining whether the current defect feature data possesses spatiotemporal constancy, firstly, Euclidean distance can be used to quantify the similarity between the current defect feature data and the historical defect feature data stored in the defect feature cache. Specifically, if the current defect feature data is represented as vector A and the historical defect feature data as vector B, then the Euclidean distance between them can be calculated as the square root of the sum of the squares of the differences in each corresponding dimension. The smaller the Euclidean distance, the more similar the two defect feature data are. Then, the system determines whether the current defect feature data possesses spatiotemporal constancy based on the relationship between the calculated Euclidean distance and a preset similarity threshold. For example, a preset Euclidean distance threshold can be set. If the calculated Euclidean distance is less than or equal to this threshold, the current defect feature data is considered to have spatiotemporal constancy, meaning the defect is a recurring system artifact. If the Euclidean distance is greater than this threshold, the current defect feature data is considered not to have spatiotemporal constancy, and may be a real product defect or random noise. This preset similarity threshold can be determined through statistical analysis of the Euclidean distance distribution obtained by measuring artifact feature data at the same location multiple times in a defect-free state. For example, it can be set to a certain percentile of this distribution to ensure effective differentiation between artifacts and real defects.

[0124] Optional, combined Figure 7As shown, A2 determines whether the current defect feature data has spatiotemporal constancy based on the relationship between similarity and a preset similarity threshold. The steps to obtain the spatiotemporal constancy determination result include:

[0125] A21, Obtain defect feature data of the reference part;

[0126] A22, compare the defect feature data of the reference part with the pre-stored reference part standard data to obtain the feature deviation of the reference part;

[0127] A23, Based on the feature deviation of the reference part, the preset initial threshold is corrected to obtain the corrected preset similarity threshold;

[0128] A24. Based on the relationship between similarity and preset similarity threshold, determine whether the current defect feature data has spatiotemporal constancy and obtain the spatiotemporal constancy judgment result.

[0129] Among them, the defect feature data of the reference part refers to the defect-related information obtained by measuring a reference part in a known state. Specifically, it can be the surface deformation data, stress distribution data, or their derived geometric feature data of the reference part, the purpose of which is to provide a benchmark for subsequent threshold correction. The reference part standard data refers to the feature information of the reference part that is stored in advance and represents an ideal or known qualified state. Specifically, it can be theoretical design values, precisely calibrated gold sample data, or statistical average data, the purpose of which is to serve as a benchmark for comparison to quantify the actual deviation of the reference part. The reference part feature deviation refers to the difference between the defect feature data of the reference part and the reference part standard data, specifically it can be positional deviation, morphological deviation, stress deviation, or a combination thereof. The purpose of the initial similarity threshold is to reflect the influence of the manufacturing tolerances of the reference part itself, measurement system errors, or environmental factors on the measurement results. The preset initial threshold refers to a fixed similarity limit used to judge the spatiotemporal constancy of defects before any correction is made. Specifically, it can be an empirical value, a statistical average, or a fixed value set based on a theoretical model. Its purpose is to provide a preliminary judgment standard. The corrected preset similarity threshold refers to a new similarity limit obtained by adjusting the preset initial threshold after considering the characteristic deviation of the reference part. Specifically, it can be a dynamic value obtained through linear adjustment, nonlinear mapping, or table lookup. Its purpose is to make the similarity judgment standard more adaptable to the actual detection environment and the characteristics of the reference part, and improve the accuracy of the judgment.

[0130] In some preferred embodiments, this application is implemented as follows: First, a precision-calibrated, known defect-free ultra-precision diaphragm spring can be selected as a reference. Surface deformation information of the reference spring is acquired using a laser speckle interferometry system, and its defect feature data is extracted from it. For example, stress distribution characteristics, geometric contour characteristics, or texture characteristics of specific areas in the surface deformation map can be extracted. Next, the acquired reference spring defect feature data is compared with reference spring standard data pre-stored in the system database. The reference spring standard data can be the theoretical design deformation data of that spring model, or the average deformation characteristics obtained through statistical analysis of a large number of qualified products. The comparison process can employ various algorithms, such as calculating the Euclidean distance, correlation coefficient, or structural similarity index between the two sets of data, thereby quantifying the reference spring feature deviation. This deviation can specifically manifest as the difference between the actual deformation of the reference spring and the standard deformation in terms of position, shape, or stress value. Subsequently, based on the calculated reference spring feature deviation, a preset initial similarity threshold is corrected. For example, if the reference part's feature deviation is large in a certain direction (such as deformation amplitude), a preset correction function, such as a linear function or a piecewise function, can be used to appropriately amplify the initial threshold; conversely, if the deviation is small, it can be fine-tuned or left unchanged. The corrected preset similarity threshold will better reflect the actual performance of the current detection system and the inherent characteristics of the reference part. Finally, when actually detecting the spring under test, the system calculates the similarity between the current defect feature data and the historical defect feature data stored in the defect feature cache. Then, this similarity is compared with the preset similarity threshold just obtained. If the similarity is higher than the corrected threshold, the current defect is judged to have spatiotemporal constancy; otherwise, it is judged not to have spatiotemporal constancy, thus obtaining the final spatiotemporal constancy judgment result.

[0131] Optionally, the step of correcting the preset initial threshold based on the feature deviation of the reference part to obtain the corrected preset similarity threshold includes:

[0132] Obtain the positional deviation, morphological deviation, and stress deviation from the feature deviations of the reference part;

[0133] Based on the combination of position deviation, shape deviation and stress deviation, consult the pre-stored correction mapping table to determine the correction amount for the corresponding position, shape and stress components in the preset initial threshold.

[0134] The determined correction values ​​are applied to the corresponding positions, shapes, and stress components in the initial threshold to obtain the corrected preset similarity threshold.

[0135] The correction mapping table refers to a pre-established lookup table that stores the correspondence between different types of reference component feature deviation combinations and corresponding threshold correction amounts. It can be implemented using a database, lookup matrix, or multidimensional array. Its purpose is to quickly and accurately determine the required threshold adjustment parameters based on the actual deviation of the reference component. The correction amount refers to the value calculated or obtained from a table based on the specific feature deviation of the reference component (such as position deviation, shape deviation, stress deviation), used to adjust the corresponding part of the preset initial threshold. Specifically, it can be a positive value, a negative value, or zero. Its purpose is to refine and dynamically adjust the initial threshold to adapt to the actual state of the reference component, thereby improving the accuracy of spatiotemporal constancy judgment.

[0136] In some preferred embodiments, to correct the preset initial similarity threshold, surface deformation data of the reference part can first be acquired using a high-precision sensor array. Combined with 3D scanning technology, the minute spatial offsets (i.e., positional deviations) of the reference part, as well as the differences in its geometry compared to the standard model (i.e., morphological deviations), can be extracted. Simultaneously, laser speckle interferometry can be used to measure the residual stress distribution of the reference part and compare it with a standard stress distribution, thereby quantifying its internal stress deviations. These deviation data can be digitized and stored. Subsequently, a processing unit can receive combined data of these positional deviations, morphological deviations, and stress deviations. This processing unit can have a pre-stored correction mapping table built-in. This table can be a multi-dimensional lookup table, where the index can be a combination of deviations within different ranges, and the corresponding value is the correction amount for the corresponding part of the preset initial threshold. For example, if the positional deviation is in a certain range, the morphological deviation is in another range, and the stress deviation is in a third range, the processing unit can consult the table to determine the specific correction values ​​for the position-sensitive, morphological-sensitive, and stress-sensitive parts of the initial threshold. Finally, the processing unit applies these determined correction values ​​to the corresponding position, shape, and stress components of the preset initial threshold using either a weighted average or direct summation, thereby generating a dynamically adjusted preset similarity threshold. For example, if the initial threshold is defined as multiple components (such as a position threshold, shape threshold, and stress threshold), each component can be added to or subtracted from its corresponding correction value to obtain the final corrected threshold. This corrected threshold will be used for subsequent defect feature data similarity judgment to improve the accuracy of the judgment.

[0137] It should be noted that in the ultra-precision spring defect detection method, the threshold is dynamically adjusted to improve the accuracy and reliability of defect judgment, especially when the reference part itself may have manufacturing deviations or measurement errors.

[0138] A specific implementation example of dynamically adjusting the threshold can be as follows:

[0139] First, it is necessary to obtain defect characteristic data of a reference component. This reference component can be a standard spring known to be of good quality, or a specific sample with known and stable deviations. The reference component is then inspected using a laser speckle interferometry system to obtain its surface deformation information, and data representing its characteristics are extracted from it. This characteristic data may include, but is not limited to, the deformation amplitude, stress concentration, and geometric profile parameters of the reference component at specific locations.

[0140] Next, the acquired defect feature data of the reference part is compared with pre-stored reference part standard data. The reference part standard data consists of feature values ​​of a reference part under ideal conditions or after rigorous calibration; for example, it can be established through theoretical calculations, CAD model data, or statistical averages of a large number of historical qualified products. By comparing these values, the differences between the actual features of the reference part and the standard features can be quantified, thus obtaining the reference part feature deviation. This deviation can be decomposed into multiple dimensions, such as positional deviation (the offset between the actual position of the reference part and the standard position), morphological deviation (the difference between the actual shape of the reference part and the standard shape), and stress deviation (the difference between the actual stress distribution of the reference part and the standard stress distribution).

[0141] Then, based on the deviations of these decomposed reference part features, the preset initial similarity threshold is corrected. The initial similarity threshold is a fixed standard used by the system to determine the spatiotemporal constancy of defects before any dynamic adjustments are made. To perform this correction, a correction mapping table can be pre-established. This correction mapping table is a multi-dimensional lookup table that contains the threshold correction amounts corresponding to different combinations of positional deviations, morphological deviations, and stress deviations. For example, if the reference part has a large positional deviation, a small morphological deviation, and a moderate stress deviation, the system will look up the necessary adjustment values ​​for the "position-sensitive part," "morphological-sensitive part," and "stress-sensitive part" of the initial threshold in the correction mapping table based on this specific combination of deviations.

[0142] Finally, these correction values ​​determined from the correction mapping table are applied to the corresponding position, shape, and stress components of the preset initial threshold. For example, if the initial threshold is defined to include multiple components such as position threshold, shape threshold, and stress threshold, then each component will be adjusted according to its corresponding correction value. In this way, a finely adjusted preset similarity threshold that better reflects the current state of the detection system and the actual characteristics of the reference part can be obtained. This adjusted threshold will be used for subsequent defect spatiotemporal constancy determination; that is, when comparing the similarity between the defect feature data of the current spring under test and historical defect feature data, it will be compared with this dynamically adjusted threshold to obtain a more accurate spatiotemporal constancy determination result.

[0143] In the defect detection method for ultra-precision springs, the initial setting of the threshold is a crucial step in ensuring detection accuracy. According to those skilled in the art, the initial setting of the threshold can be based on the following implementation methods:

[0144] One implementation method is based on historical data statistical analysis. In actual production environments, a large amount of inspection data on known qualified and known defective ultra-precision springs can be collected. From this data, corresponding defect feature data can be extracted, and the similarity between current defect feature data and historical defect feature data can be calculated. By performing statistical analysis on these similarity data, such as calculating the mean, standard deviation, or performing distribution fitting, a similarity threshold that can effectively distinguish between qualified and defective products can be determined. For example, a statistical confidence interval can be set, and similarity values ​​falling outside this interval can be used as a preliminary basis for defect judgment.

[0145] Another approach is to set an initial similarity threshold based on expert experience. In the absence of extensive historical data, experienced quality control engineers or technical experts can directly set an initial similarity threshold based on their long-term observation and understanding of defects in ultra-precision springs, combined with the tolerance for defects in actual production. For example, experts can provide a preliminary similarity range based on the visual characteristics of the defect, its impact on product performance, and past testing experience, serving as a basis for judging spatiotemporal constancy.

[0146] Alternatively, a reference calibration method can be used for initial setup. This involves selecting one or more reference parts in a known state (e.g., a standard defect-free spring or a spring with typical artifacts) and performing multiple tests under controlled conditions. The defect feature data of these reference parts are recorded, and their similarity to their own standard data is calculated. Based on these similarity values, an initial threshold can be set that reflects the inherent fluctuations of the system under ideal or known deviation conditions. For example, if the similarity between the defect feature data of a standard defect-free reference part and the standard data remains at a relatively high level across multiple tests, then a value near that level can be used as the initial similarity threshold.

[0147] In practice, these initially set thresholds may need to be verified and adjusted through small-batch trial production or actual operation to ensure their effectiveness and robustness in a real production environment. For example, after initially setting the thresholds, a small number of known defects and qualified products can be tested, and the thresholds can be fine-tuned based on the test results to achieve optimal detection performance.

[0148] A high-precision spring defect detection system, combined with Figure 8 As shown, the structure of the ultra-precision spring defect detection system 1 includes:

[0149] The surface deformation acquisition module 11 is used to acquire the surface deformation information of the ultra-precision diaphragm spring to be tested.

[0150] Defect feature extraction module 12 is used to extract current defect feature data from surface deformation information;

[0151] Feature data storage module 13 is used to store the current defect feature data in the defect feature cache area;

[0152] The spatiotemporal constancy judgment module 14 is used to compare the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy and obtain the spatiotemporal constancy judgment result.

[0153] The defect type determination module 15 is used to mark the defect type corresponding to the current defect as a system artifact or a real product defect based on the spatiotemporal constancy determination result.

[0154] The test result display module 16 is used to generate a test report and display defect type markers based on preset display rules that distinguish between system artifacts and real product defects.

[0155] The system comprises several modules: Surface Deformation Acquisition Module, Dedicated to Capturing Microscopic Deformation of the Spring Surface; and Surface Deformation Analysis Module. The Surface Deformation Acquisition Module utilizes optical measurement equipment such as laser speckle interferometers, white light interferometers, confocal microscopes, or 3D scanners to acquire high-precision surface deformation data as raw input for defect analysis. Defect Feature Extraction Module identifies and quantifies defect-related features from the raw deformation information. It employs image processing algorithms, signal processing techniques, or machine learning models to transform the raw deformation data into structured feature data for subsequent analysis. Feature Data Storage Module persistently or temporarily stores defect feature data. It utilizes high-speed caching, solid-state drives, database systems, or distributed storage systems to establish a historical record of defect features, supporting temporal analysis and comparison. Spatiotemporal Constancy Judgment Module assesses the consistency of defect features across different time points and spatial locations. It employs statistical methods, pattern recognition algorithms, or rule-based logical judgment to distinguish artifacts caused by inherent system characteristics from genuine defects in the product itself. The defect type determination module is a unit used to classify and label identified defects based on the spatiotemporal constancy judgment results. It can employ classification algorithms such as decision trees, support vector machines, and neural networks, or be implemented using an expert system based on preset logical rules. Its purpose is to clearly distinguish between system artifacts and real product defects, providing an accurate basis for subsequent processing. The detection result display module is a unit used to present the defect detection results to the user in a visual manner. It can be implemented using graphical user interface software, data visualization tools, or report generation engines. Its purpose is to intuitively display the detection results, facilitating user understanding and decision-making, and highlighting the difference between artifacts and real defects.

[0156] This application's solution, through a modular system design, achieves automation and intelligence in the detection of defects in ultra-precision springs, particularly providing an effective technical means to distinguish between system artifacts and genuine product defects. Specifically, the surface deformation acquisition module first acquires the surface deformation information of the ultra-precision diaphragm spring under test, providing raw data input for the entire detection process. Subsequently, the defect feature extraction module processes this raw deformation information, identifying and extracting the feature data of the currently detected defects, transforming unstructured deformation information into quantifiable feature representations. These extracted current defect feature data are received by the feature data storage module and stored in the defect feature cache. This cache not only stores current data but also accumulates historical defect feature data, laying the foundation for subsequent temporal analysis. The spatiotemporal constancy judgment module is the core of this system. It compares the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data is constant in time and space. This comparison mechanism can identify "defects" that repeatedly appear in the same location and have similar shapes in multiple detections; these are usually artifacts caused by optical contamination of the system itself or environmental interference. In this way, the system can obtain a spatiotemporal constancy judgment result. Next, the defect type determination module classifies the current defect based on the spatiotemporal constancy judgment result. If the judgment result shows that the defect has spatiotemporal constancy, it is marked as a system artifact; otherwise, it is marked as a real product defect. This spatiotemporal constancy-based judgment mechanism enables the system to automatically and accurately distinguish between false signals introduced by the detection equipment itself and real defects in the product itself, avoiding misjudgment. Finally, the detection result display module generates a detailed detection report based on the output of the defect type determination module. This module also distinguishes between system artifacts and real product defects based on preset display rules; for example, different colors, icons, or marks can be used to visually present these two different types of defects. This allows users to clearly identify which product quality issues require attention and which are interferences from the system itself, thereby avoiding production stoppages and resource waste caused by misjudged artifacts. Through the above system structure, this application elevates the original ultra-precision spring defect detection method from purely algorithmic logic to a solution with physical support and automation capabilities. This system can efficiently execute each step of the method, and in particular, through the cooperation of the spatiotemporal constancy judgment module and the defect type judgment module, it solves the problem that traditional methods are difficult to automatically and accurately distinguish between system artifacts and real product defects in practical applications. This systematic implementation makes the detection process more reliable, reduces the risk of human intervention and misjudgment, and thus significantly improves the efficiency and accuracy of detection.

[0157] In some preferred embodiments, this application is implemented as follows: The surface deformation acquisition module can employ a measurement device based on the principle of laser speckle interferometry. For example, a high-power laser can be configured as the light source, and a beam splitter can divide the laser into a reference beam and an object beam. The object beam illuminates the surface of the ultra-precision diaphragm spring under test, and after reflection, it forms an interference speckle pattern with the reference beam on a CCD or CMOS image sensor. The image sensor can be a high-resolution industrial camera; for example, a camera with high frame rate and low noise characteristics can be selected to capture minute deformation information of the spring surface. The defect feature extraction module can be an embedded processor or an industrial computer running image processing software. This software can first preprocess the acquired speckle pattern, for example, by performing noise filtering and image enhancement. Subsequently, a digital image correlation algorithm or a phase unwrapping algorithm can be applied to calculate the full-field deformation data of the spring surface from the speckle pattern. Then, a feature extraction algorithm, such as a gradient, curvature, or local texture analysis method, can be used to identify abnormal regions from the deformation data and extract their geometric and deformation features. The feature data storage module can be a high-speed solid-state drive or a dedicated data server configured with a database management system. The current defect feature data extracted from each detection, including the defect's location coordinates, geometric features, deformation features, and detection timestamp, can be structured and stored in the database. The database can maintain a defect feature cache to store historical defect feature data from the most recent period or batch of products, facilitating rapid retrieval and comparison. The spatiotemporal constancy judgment module can be implemented by a high-performance central processing unit or graphics processing unit, on which a comparison algorithm runs. This algorithm can retrieve historical defect feature data from the defect feature cache that is close to the current defect location. During comparison, the similarity between the current defect feature data and historical defect feature data can be calculated, for example, using Euclidean distance, cosine similarity, or structural similarity index. If the similarity exceeds a preset threshold, and the defects highly overlap in spatial location, it can be determined that it possesses spatiotemporal constancy. The defect type determination module can be a separate software module that receives the spatiotemporal constancy judgment result. This module can incorporate a decision logic. For example, if the spatiotemporal constancy judgment module outputs "constant," the current defect is marked as a "system artifact"; if it outputs "non-constant," it is marked as a "real product defect." This logic can be adjusted according to the actual application scenario. For instance, confidence assessment can be introduced to improve the robustness of the judgment. The detection result display module can be a monitor connected to an industrial computer, running a user-friendly graphical interface program. This program can display the detection progress and results in real time. When generating a detection report, the program can use preset display rules, such as marking system artifacts with blue or dashed boxes and real product defects with red or solid boxes, and providing a detailed defect list and statistical information in the report.Users can view images, feature data, and judgment results of defects through the interface, and can export reports for archiving or further analysis.

[0158] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for detecting defects in ultra-precision springs, characterized in that, Includes the following steps: Obtain surface deformation information of the ultra-precision diaphragm spring under test; Extract current defect feature data from the surface deformation information; Store the current defect feature data in the defect feature cache area; The current defect feature data is compared with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy, and the spatiotemporal constancy determination result is obtained. The step of comparing the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy and obtaining a spatiotemporal constancy determination result includes: obtaining the currently detected laser power; adjusting the current defect feature data or the historical defect feature data stored in the defect feature cache according to the laser power; comparing the adjusted current defect feature data with the adjusted historical defect feature data to determine whether the current defect feature data has spatiotemporal constancy and obtaining a spatiotemporal constancy determination result; The step of adjusting the current defect feature data or the historical defect feature data stored in the defect feature cache based on the laser power includes: identifying the spatial distribution area of ​​artifacts in the image based on the artifacts in the current defect feature data or the historical defect feature data stored in the defect feature cache; determining the local laser power sensitivity at each location within the spatial distribution area of ​​the artifacts; and making spatial position-related adjustments to the current defect feature data or the historical defect feature data stored in the defect feature cache based on the laser power and the local laser power sensitivity. The step of determining whether the current defect feature data has spatiotemporal constancy and obtaining the spatiotemporal constancy determination result includes: Quantify the similarity between the current defect feature data and the historical defect feature data stored in the defect feature cache; Based on the relationship between the similarity and the preset similarity threshold, it is determined whether the current defect feature data has spatiotemporal constancy, and a spatiotemporal constancy judgment result is obtained; Based on the spatiotemporal constancy judgment result, the defect type corresponding to the current defect is marked as either a system artifact or a real product defect. A detection report is generated, and the defect type marker is displayed based on a preset display rule that distinguishes between system artifacts and real product defects.

2. The method for detecting defects in ultra-precision springs according to claim 1, characterized in that, The step of determining the local laser power sensitivity at each location within the spatial distribution area of ​​the artifact includes: Obtain artifact feature data and corresponding laser power data at each location within the spatial distribution area of ​​the artifacts; Based on the artifact feature data and the laser power data, the ratio of the change in artifact features to the change in laser power at each location within the spatial distribution area of ​​the artifact is calculated to obtain the local laser power sensitivity at each location within the spatial distribution area of ​​the artifact. Based on the calculated results, the local laser power sensitivity at each location within the spatial distribution area of ​​the artifacts is updated and confirmed.

3. The method for detecting defects in ultra-precision springs according to claim 1, characterized in that, The step of making spatial location-related adjustments to the current defect feature data or the historical defect feature data stored in the defect feature cache includes: The spatial distribution area of ​​artifacts is divided into several tiny spatial units; Based on the local laser power sensitivity and corresponding laser power of the micro spatial unit, the artifact feature correction amount of the micro spatial unit is calculated; The artifact feature correction amount is applied to the feature value of the corresponding micro spatial unit in the current defect feature data or the historical defect feature data stored in the defect feature cache, so as to realize the spatial location-related adjustment of the current defect feature data or the historical defect feature data stored in the defect feature cache.

4. The method for detecting defects in ultra-precision springs according to claim 3, characterized in that, The step of calculating the artifact feature correction amount of the tiny spatial unit includes: Obtain the local laser power sensitivity corresponding to the micro spatial unit; Obtain the current laser power; Based on the local laser power sensitivity corresponding to the micro spatial unit and the current laser power, the artifact feature correction amount of the micro spatial unit is determined by consulting a pre-stored correction amount lookup table or applying a pre-calibrated correction curve.

5. The method for detecting defects in ultra-precision springs according to claim 1, characterized in that, The step of determining whether the current defect feature data has spatiotemporal constancy based on the relationship between the similarity and a preset similarity threshold, and obtaining the spatiotemporal constancy determination result, includes: Obtain defect feature data of the reference part; By comparing the defect feature data of the reference part with the pre-stored reference part standard data, the feature deviation of the reference part is obtained; Based on the feature deviation of the reference part, the preset initial threshold is corrected to obtain the corrected preset similarity threshold; Based on the relationship between the similarity and the preset similarity threshold, it is determined whether the current defect feature data has spatiotemporal constancy, and a spatiotemporal constancy judgment result is obtained.

6. The method for detecting defects in ultra-precision springs according to claim 5, characterized in that, The step of correcting the preset initial threshold based on the feature deviation of the reference part to obtain the corrected preset similarity threshold includes: Obtain the positional deviation, morphological deviation, and stress deviation from the characteristic deviations of the reference part; Based on the combination of position deviation, shape deviation and stress deviation, consult the pre-stored correction mapping table to determine the correction amount for the corresponding position, shape and stress portion in the preset initial threshold. The determined correction values ​​are applied to the corresponding positions, shapes, and stress components in the initial threshold to obtain the corrected preset similarity threshold.

7. A system for detecting defects in ultra-precision springs, used to perform the method for detecting defects in ultra-precision springs as described in claim 6, characterized in that, include: The surface deformation acquisition module is used to acquire the surface deformation information of the ultra-precision diaphragm spring under test; The defect feature extraction module is used to extract current defect feature data from the surface deformation information; A feature data storage module is used to store the current defect feature data in a defect feature cache area; The spatiotemporal constancy determination module is used to compare the current defect feature data with the historical defect feature data stored in the defect feature cache to determine whether the current defect feature data has spatiotemporal constancy and obtain the spatiotemporal constancy determination result. The defect type determination module is used to determine whether the defect type corresponding to the current defect is a system artifact or a real product defect based on the spatiotemporal constancy determination result. The detection result display module is used to generate a detection report and display the defect type marker based on a preset display rule that distinguishes between system artifacts and real product defects.

Images