An aero-engine single-crystal turbine blade coating thickness uniformity detection method based on AI and three-dimensional reconstruction

Abstract

The application discloses an aero-engine single-crystal turbine blade coating thickness uniformity detection method based on AI and three-dimensional reconstruction, and relates to the technical field of aero-engine component detection. The method collects the structure data of the turbine blade before and after coating, performs three-dimensional reconstruction, realizes accurate alignment and differential processing of double models by using an AI registration model, and extracts a three-dimensional coating model. Thickness data of the whole region and key regions are obtained through voxelization processing and an AI thickness analysis model, multi-dimensional evaluation is performed in combination with uniformity coefficients, and finally, the coating qualification is determined and a comprehensive report is output. The application can realize accurate detection of the whole region and key analysis of the key region of the coating thickness, significantly improve the detection efficiency and accuracy, and meet the high-precision quality control requirements of the aero-engine turbine blade coating.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aero-engine component detection, and in particular to an aero-engine single-crystal turbine blade coating thickness uniformity detection method based on AI and three-dimensional reconstruction. BACKGROUND

[0002] The aero-engine single-crystal turbine blade is a core component of the engine, and works in an extreme environment of high temperature, high pressure and high stress. The thickness uniformity of the surface coating (such as thermal barrier coating, wear-resistant coating, etc.) of the single-crystal turbine blade directly affects the service life and operational safety of the blade. The traditional coating thickness detection methods (such as metallographic sectioning method and eddy current thickness measurement method) have limited detection range, destructive detection or insufficient precision, and cannot meet the full-domain uniformity evaluation requirements of the single-crystal turbine blade complex curved surface coating.

[0003] With the development of three-dimensional reconstruction and artificial intelligence technology, non-contact detection methods based on three-dimensional models have gradually become a research hotspot. However, in the prior art, the combination of three-dimensional models and thickness analysis is not close enough, the definition of key areas lacks mechanical load basis, and the uniformity evaluation index is single, resulting in insufficient reliability and engineering practicability of the detection results. Therefore, there is an urgent need for a coating thickness uniformity detection method that can realize full-domain accurate detection, key area analysis and rigorous evaluation logic. SUMMARY

[0004] The present application aims to overcome the limitations of the prior art in detecting the coating thickness uniformity of the aero-engine single-crystal turbine blade, and provides a detection method based on AI and three-dimensional reconstruction. Through three-dimensional model reconstruction, AI-assisted analysis and multi-dimensional evaluation, the coating thickness uniformity is accurately detected and qualified.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] An aero-engine single-crystal turbine blade coating thickness uniformity detection method based on AI and three-dimensional reconstruction, comprising the following steps:

[0007] S1, pre-coating data acquisition: acquiring the surface and internal structure data of the uncoated turbine blade to form pre-coating volume data;

[0008] S2, post-coating data acquisition: acquiring the surface and internal structure data of the coated turbine blade to form post-coating volume data;

[0009] S3, double model reconstruction: three-dimensional reconstruction is performed on the pre-coating volume data and the post-coating volume data to construct the pre-coating model M1 and the post-coating model M2 with consistent coordinate systems;

[0010] S4, registration and difference: precisely align M1 and M2 through an AI registration model, and then extract the three-dimensional coating model M3 through difference processing;

[0011] S5, thickness analysis and evaluation: based on M3, obtain the global thickness data set, the key area thickness data set and the uniformity coefficient through the AI thickness analysis model, and generate the evaluation result set;

[0012] S6, result determination: determine whether the coating is qualified according to the comparison result of the evaluation result set and the preset threshold, and output a complete report.

[0013] The detection method constructs a full-process closed loop of "data acquisition - three-dimensional reconstruction - AI registration - difference modeling - thickness analysis - result determination", and realizes the detection upgrade from "two-dimensional observation" to "three-dimensional global" by collecting the surface and internal structure data of the pre-coating and post-coating blades in stages, aiming at the core needs of the coating thickness uniformity detection of the single crystal turbine blade of the aero-engine. The method first reconstructs the pre-coating / three-dimensional model with consistent coordinate system based on double volume data, then breaks through the traditional registration precision bottleneck with the help of AI registration technology, accurately strips the three-dimensional coating model through difference processing, and finally quantifies the global, key area thickness data and uniformity coefficient through the AI thickness analysis model, and completes the qualification determination and outputs a complete report in combination with the preset threshold, which takes into account the accuracy, comprehensiveness and efficiency of the detection throughout the process, and provides a systematic solution for the quality control of the core components of the aero-engine.

[0014] Preferably, the step S5 comprises:

[0015] a) voxelization processing: discretizing the three-dimensional coating model M3 into a plurality of cubic voxels according to a preset resolution, and assigning a three-dimensional coordinate to each voxel;

[0016] b) global thickness data acquisition: applying an alternating electromagnetic field to each voxel, and the AI thickness analysis model inverts the thickness based on the skin effect to generate a global thickness data set of "coordinate-thickness" mapping;

[0017] c) key area thickness data acquisition: based on the global thickness data, extracting the voxel thickness data corresponding to the key area, and calculating the statistical parameters of the voxel thickness data of the key area and the proportion of the coating volume of the key area to the total coating volume, the statistical parameters of each voxel thickness data of the key area at least including the average thickness, the minimum thickness of the key area voxel thickness data, thereby forming a key area thickness data set;

[0018] d) uniformity coefficient acquisition: based on the global thickness data, calculating the uniformity coefficient according to a preset algorithm;

[0019] e) Result integration: integrate the global thickness dataset, the key area thickness dataset, and the uniformity coefficient to generate an evaluation result set.

[0020] The creativity of this step S5 lies in the systematic and innovative design of the coating thickness analysis process, breaking the limitations of traditional thickness detection "data fragmentation and single analysis": first, it discretizes the three-dimensional coating model M3 into cubic voxels with three-dimensional coordinates through voxelization processing, laying a structured data foundation for accurate tracing of global thickness, solving the pain point of difficulty in accurately correlating complex coating structure thickness data with spatial positions; it innovatively uses the combination of "alternating electromagnetic field + skin effect + AI inversion" to obtain global thickness data, realizing a leap from "local sampling detection" to "global dead-angle-free measurement", and the intelligent inversion of the AI model on electromagnetic response signals greatly improves the extraction accuracy and efficiency of thickness data; in the key area analysis, not only the voxel thickness data of the core stress area are extracted, but also the dual quantitative indicators of "statistical parameters + volume ratio" are innovatively introduced, making the thickness characteristic evaluation of the key area more in line with the actual use requirements under extreme working conditions of the blade; at the same time, the uniformity coefficient is constructed by a preset algorithm (such as the coefficient of variation), realizing the quantitative representation of the uniformity of the coating thickness distribution, and then forming a complete evaluation result set through the integration of multiple types of data, constructing a "global coverage - key focus - quantitative closed loop" thickness analysis logic, which not only guarantees the comprehensiveness and accuracy of the analysis, but also provides multi-dimensional and high-credibility data support for subsequent qualification judgment, significantly better than the traditional single-dimensional thickness analysis method.

[0021] Preferably, the step S6 specifically comprises:

[0022] comparing the average thickness in the key area thickness dataset with a preset average thickness allowable range to determine whether the average thickness of the key area is within the allowable range;

[0023] comparing the minimum thickness in the key area thickness dataset with a preset minimum thickness threshold to determine whether the minimum thickness of the key area is greater than or equal to the minimum thickness threshold;

[0024] comparing the proportion of the key area coating volume to the total coating volume with a preset minimum volume proportion to determine whether the volume proportion is greater than or equal to the minimum volume proportion;

[0025] comparing each voxel thickness in the global thickness dataset with a preset global voxel thickness threshold to determine whether all voxel thicknesses are within the global voxel thickness threshold interval;

[0026] The uniformity coefficient is compared with a preset maximum allowable variation coefficient to determine whether the uniformity coefficient is less than or equal to the maximum allowable variation coefficient;

[0027] When the determination results of the above 5 items are all "yes", it is determined that the coating uniformity is qualified; if any one of the determination results is "no", it is determined that the coating uniformity is unqualified;

[0028] The final output includes a comprehensive test report including the qualification conclusion, the thickness distribution cloud map, the key area statistical report and the volume proportion analysis diagram.

[0029] Preferably, the step b) of extracting the global thickness data specifically includes:

[0030] In the voxelized model M3, an alternating electromagnetic field of a specific frequency is applied to the coating area corresponding to each cubic voxel; the AI thickness analysis model receives the electromagnetic response signal generated by the skin effect of each coating area; based on the attenuation amount or phase change of the electromagnetic response signal, the AI thickness analysis model maps the coating thickness value at the normal direction from the voxel center point to the pre-coating model M1 through the inversion function trained in the AI thickness analysis model; after traversing all voxels, the three-dimensional coordinates and the inversion thickness values of all voxels are integrated to generate a "coordinate-thickness" mapping data set covering the entire coating three-dimensional model M3, i.e. the coating global thickness data set.

[0031] Preferably, the key area in the step c) is a core function and stress area predefined based on the thermodynamic load and structural mechanics load analysis of the turbine blade of the aero-engine, which at least includes:

[0032] The blade region is the main part of the gas passage and bears the highest thermal load, directly affecting the high-temperature protection effect of the coating;

[0033] The crown region participates in blade vibration suppression and air gap sealing and bears periodic vibration load, and the coating integrity affects the stability of the blade operation;

[0034] The tenon region is the key connection between the blade and the disc, bearing the maximum mechanical stress and assembly stress, and the coating uniformity is directly related to the connection reliability.

[0035] Preferably, in the step d) of calculating the uniformity coefficient, the preset algorithm is:

[0036] The ratio of the standard deviation σ and the average value μ of all thickness values in the coating global thickness data set is calculated, i.e. the variation coefficient CV, and the variation coefficient is defined as the uniformity coefficient U, and the calculation formula is:

[0037] U = σ / μ

[0038] The smaller the value of the uniformity coefficient is, the more uniform the coating thickness distribution is.

[0039] The uniformity coefficient calculation method is creative in that it breaks through the limitations of traditional coating uniformity evaluation "qualitative description" or "single parameter quantification", and builds a scientific quantification system with "relative dispersion degree" as the core: it innovatively introduces the coefficient of variation (CV) into the detection of single crystal turbine blade coating uniformity of an aero-engine, defines the uniformity coefficient U through the ratio of standard deviation to average value, fundamentally eliminates the interference of coating thickness order difference on uniformity evaluation, and solves the industry pain point of "different thickness specifications of coating cannot be directly compared in uniformity" in traditional methods; the coefficient realizes the quantitative characterization of the uniformity of coating thickness distribution through a simple and clear formula (U = s / m), and through the intuitive logic of "the smaller the value is, the more uniform it is", the evaluation result has strong operability and comparability, avoiding the ambiguity of traditional qualitative description (such as "thickness is relatively uniform" and "local unevenness"); at the same time, the algorithm directly adapts to the global thickness data set after voxelization processing, without additional complex data preprocessing, and is highly compatible with the overall detection process, ensuring the scientificity and accuracy of the uniformity evaluation, and also taking into account the efficiency in engineering application, so that the coating uniformity is upgraded from "empirical judgment" to "precise quantification", providing key technical support for the standardization and fine control of coating quality of aero-engine core components.

[0040] The present application has the following beneficial effects:

[0041] 1. Global detection combined with key areas: the acquisition of global coating thickness data is realized through three-dimensional reconstruction and voxelization processing, and the key areas are defined based on mechanical load analysis, taking into account the comprehensiveness and pertinence of the detection, and avoiding quality blind spots.

[0042] 2. AI technology improves detection accuracy: AI registration model is used to realize high-precision alignment of double models, and AI thickness analysis model is used to invert thickness based on electromagnetic response signal, which significantly improves the accuracy and efficiency of detection and meets the micron-level precision requirement.

[0043] 3. Multi-dimensional evaluation logic is rigorous: through the cooperative judgment of key area statistical parameters, volume proportion, global thickness range and uniformity coefficient, a complete quality evaluation system is formed to ensure that the coating uniformity meets the extreme working condition requirements of the aero-engine.

[0044] 4. Non-contact detection: non-contact data acquisition and analysis are used throughout the process to avoid damage to the blade and coating, and it is suitable for batch detection and quality sampling of in-service blades. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 Figure 1 is a flow chart of an AI and three-dimensional reconstruction-based aero-engine single-crystal turbine blade coating thickness uniformity detection method according to the present application. DETAILED DESCRIPTION

[0046] The present application will be further described in conjunction with the embodiments and drawings, but the embodiments of the present application are not limited thereto. EMBODIMENT

[0047] As shown in Figure 1 , Figure 1 Figure 1 is a flow chart of an AI and three-dimensional reconstruction-based aero-engine single-crystal turbine blade coating thickness uniformity detection method according to the present application.

[0048] An AI and three-dimensional reconstruction-based aero-engine single-crystal turbine blade coating thickness uniformity detection method, comprising the following steps:

[0049] S1, pre-coating data acquisition: acquiring surface and internal structure data of uncoated turbine blades to form pre-coating volume data;

[0050] S2, post-coating data acquisition: acquiring surface and internal structure data of coated turbine blades to form post-coating volume data;

[0051] S3, double-model reconstruction: three-dimensional reconstruction is performed on the pre-coating volume data and the post-coating volume data respectively to construct pre-coating model M1 and post-coating model M2 with consistent coordinate systems;

[0052] S4, registration and difference: M1 and M2 are accurately aligned by an AI registration model, and a coating three-dimensional model M3 is extracted through difference processing;

[0053] S5, thickness analysis and evaluation: based on M3, a global thickness data set, a key area thickness data set and a uniformity coefficient are obtained through an AI thickness analysis model to generate an evaluation result set;

[0054] S6, result determination: according to the comparison result of the evaluation result set and the preset threshold value, it is determined whether the coating is qualified, and a complete report is output.

[0055] Preferably, the step S5 comprises:

[0056] a) voxelization processing: the coating three-dimensional model M3 is discretized into a plurality of cubic voxels according to a preset resolution, and each voxel is assigned a three-dimensional coordinate;

[0057] b) global thickness data acquisition: an alternating electromagnetic field is applied to each voxel, and the AI thickness analysis model inversely calculates the thickness based on the skin effect to generate a global thickness data set of "coordinate-thickness" mapping;

[0058] c) Key region thickness data acquisition: based on the global thickness data, the voxel thickness data corresponding to the key region is extracted, and the statistical parameters of the voxel thickness data of the key region and the proportion of the coating volume of the key region in the total coating volume are calculated, the statistical parameters of each voxel thickness data of the key region at least including the average thickness, the minimum thickness of the key region voxel thickness data, thereby forming a key region thickness data set;

[0059] d) Uniformity coefficient acquisition: based on the global thickness data, the uniformity coefficient is calculated according to a preset algorithm;

[0060] e) Result integration: integrating the global thickness data set, the key region thickness data set and the uniformity coefficient, an evaluation result set is generated.

[0061] Preferably, the step S6 specifically includes:

[0062] The average thickness in the key region thickness data set is compared with a preset average thickness allowable range to determine whether the average thickness of the key region is within the allowable range;

[0063] The minimum thickness in the key region thickness data set is compared with a preset minimum thickness threshold to determine whether the minimum thickness of the key region is greater than or equal to the minimum thickness threshold;

[0064] The proportion of the coating volume of the key region in the total coating volume is compared with a preset minimum volume proportion to determine whether the volume proportion is greater than or equal to the minimum volume proportion;

[0065] Each voxel thickness in the global thickness data set is compared with a preset global voxel thickness threshold, respectively, to determine whether all voxel thicknesses are within the interval of the global voxel thickness threshold;

[0066] The uniformity coefficient is compared with a preset maximum allowable variation coefficient to determine whether the uniformity coefficient is less than or equal to the maximum allowable variation coefficient;

[0067] When the determination results of the above 5 items are all "yes", it is determined that the coating uniformity is qualified; if any one of the determination results is "no", it is determined as unqualified;

[0068] The final output includes a comprehensive test report including the qualification conclusion, the thickness distribution cloud map, the key region statistical report and the volume proportion analysis diagram.

[0069] Preferably, the step b) global thickness data extraction specifically includes:

[0070] In the voxelized model M3, an alternating electromagnetic field of a specific frequency is applied to the coating area corresponding to each cubic voxel; the AI thickness analysis model receives the electromagnetic response signal generated by the skin effect of each coating area; based on the attenuation or phase change of the electromagnetic response signal, the AI thickness analysis model maps the coating thickness value at the normal direction from the voxel center point to the pre-coating model M1 through the inversion function trained inside; after traversing all voxels, the three-dimensional coordinates and the inversion thickness values of all voxels are integrated to generate a "coordinate-thickness" mapping dataset covering the entire coating three-dimensional model M3, that is, the global coating thickness dataset.

[0071] Preferably, the key area in step c) is a core function and stress area predefined based on the thermodynamic load and structural mechanical load analysis of the turbine blade of the aero-engine, which at least includes:

[0072] The blade region is the main part of the gas passage and bears the highest thermal load, directly affecting the high-temperature protection effect of the coating;

[0073] The crown region participates in blade vibration suppression and air gap sealing and bears periodic vibration load, and the coating integrity affects the stability of the blade operation;

[0074] The tenon region is the key connection between the blade and the disc, bearing the maximum mechanical stress and assembly stress, and the uniformity of the coating is directly related to the connection reliability.

[0075] Preferably, in the uniformity coefficient calculation of step d), the preset algorithm is:

[0076] The ratio of the standard deviation σ and the average value μ of all thickness values in the global coating thickness dataset is calculated, that is, the coefficient of variation CV, and the coefficient of variation is defined as the uniformity coefficient U, and the calculation formula is:

[0077] U = σ / μ

[0078] Wherein, the smaller the value of the uniformity coefficient, the more uniform the coating thickness distribution.

[0079] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. According to the technical essence of the present application, any simple modification, equivalent replacement and improvement of the above embodiment within the spirit and principles of the present application are still within the protection scope of the technical scheme of the present application.

Claims

1. An AI and three-dimensional reconstruction-based aero-engine single-crystal turbine blade coating thickness uniformity detection method, characterized in that, The method comprises the following steps: S1, pre-coating data acquisition: collecting surface and internal structure data of a turbine blade without coating to form pre-coating volume data; S2, post-coating data acquisition: collecting surface and internal structure data of a turbine blade with coating to form post-coating volume data; S3, double-model reconstruction: performing three-dimensional reconstruction on the pre-coating volume data and the post-coating volume data respectively to construct a pre-coating model M1 and a post-coating model M2 with consistent coordinate systems; S4, registration and difference: accurately aligning M1 and M2 through an AI registration model, and then extracting a coating three-dimensional model M3 through difference processing; S5, thickness analysis and evaluation: based on M3, obtaining a global thickness data set, a key area thickness data set and a uniformity coefficient through an AI thickness analysis model, and generating an evaluation result set; a) voxelization processing: discretizing the coating three-dimensional model M3 into multiple cubic voxels according to a preset resolution, and assigning a three-dimensional coordinate to each voxel; b) global thickness data acquisition: applying an alternating electromagnetic field to each voxel, and the AI thickness analysis model inverts the thickness based on the skin effect to generate a global thickness data set of the "coordinate-thickness" mapping; c) key area thickness data acquisition: based on the global thickness data, extracting the voxel thickness data corresponding to the key area, and calculating the statistical parameters of the voxel thickness data of the key area and the proportion of the coating volume of the key area to the total coating volume, the statistical parameters of each voxel thickness data of the key area at least including the average thickness, the minimum thickness of the key area voxel thickness data, thereby forming a key area thickness data set; d) uniformity coefficient acquisition: based on the global thickness data, calculating the uniformity coefficient according to a preset algorithm; e) result integration: integrating the global thickness data set, the key area thickness data set and the uniformity coefficient to generate an evaluation result set; The global thickness data extraction specifically comprises: in the voxelized model M3, applying an alternating electromagnetic field of a specific frequency to the coating area corresponding to each cubic voxel; the AI thickness analysis model receives an electromagnetic response signal generated by the skin effect of each coating area; based on the attenuation amount or phase change of the electromagnetic response signal, the AI thickness analysis model maps the coating thickness value at the normal direction from the center point of the voxel to the pre-coating model M1 through the inversion function trained in the AI thickness analysis model; after traversing all voxels, integrating the three-dimensional coordinates of all voxels and the thickness values inverted thereby to generate a "coordinate-thickness" mapping data set covering the entire coating three-dimensional model M3, that is, the coating global thickness data set; ​ ​ S6, result determination: according to the comparison result of the evaluation result set and the preset threshold value, it is determined whether the coating is qualified, and a complete report is output; the average thickness in the key area thickness data set is compared with the preset average thickness allowable range to determine whether the key area average thickness is within the allowable range; the minimum thickness in the key area thickness data set is compared with the preset minimum thickness threshold to determine whether the key area minimum thickness is greater than or equal to the minimum thickness threshold; the proportion of the key area coating volume to the total coating volume is compared with the preset minimum volume proportion to determine whether the volume proportion is greater than or equal to the minimum volume proportion; the thickness of each voxel in the global thickness data set is compared with the preset global voxel thickness threshold, respectively, to determine whether all voxel thicknesses are within the global voxel thickness threshold interval; the uniformity coefficient is compared with the preset maximum allowable variation coefficient to determine whether the uniformity coefficient is less than or equal to the maximum allowable variation coefficient; when the determination results of the above 5 items are all "yes", it is determined that the coating uniformity is qualified; if any one of the determination results is "no", it is determined as unqualified; finally, a comprehensive detection report including the qualification conclusion, thickness distribution cloud map, key area statistical report and volume proportion analysis diagram is output.

2. The AI and three-dimensional reconstruction-based aero-engine single-crystal turbine blade coating thickness uniformity detection method of claim 1, wherein The key area in the step c) is a core function and stress area defined in advance based on the thermodynamic load and structural mechanics load analysis of the turbine blade of the aero-engine, which at least includes: a blade area, which is the main part of the gas passage, bears the highest thermal load, and directly affects the high-temperature protection effect of the coating; a crown area, which participates in blade vibration suppression and air gap sealing, bears periodic vibration load, and the coating integrity affects the stability of the blade operation; a tenon area, which is the key connection between the blade and the disc, bears the maximum mechanical stress and assembly stress, and the coating uniformity is directly related to the connection reliability.

3. The AI and three-dimensional reconstruction-based aero-engine single-crystal turbine blade coating thickness uniformity detection method according to claim 1, characterized in that, In the uniformity coefficient calculation of the step d), the preset algorithm is: calculating the ratio of the standard deviation σ and the average value μ of all thickness values in the global thickness data set of the coating, that is, the variation coefficient CV, and defining the variation coefficient as the uniformity coefficient U, and the calculation formula is: U = σ / μ, wherein the smaller the value of the uniformity coefficient, the more uniform the thickness distribution of the coating.

Images