Multi-mode non-destructive testing method and system for turbine blades

Abstract

This invention discloses a multi-mode non-destructive testing method and system for turbine blades, relating to the field of non-destructive testing technology for aero-engine and gas turbine turbine blades. Addressing the problems of isolated and spatially uncorrelated data from existing fluorescence penetrant testing, X-ray fluorescence spectroscopy, and X-ray diffraction testing, which makes it difficult to quantitatively assess the risk of penetration coupled with surface defects and residual stress, this invention uses the three-dimensional point cloud of the turbine blade as a benchmark to achieve pixel-level spatial registration of the three types of testing data. By calculating the distance from the bottom of the surface defect to the high-stress zone and combining it with a safety threshold dynamically correlated with the material's fracture toughness and residual stress peak value, the invention automatically and quantitatively determines the penetration risk and outputs quality conclusions based on multi-modal features. It can also automatically generate laser repair paths and G-codes. This invention significantly improves testing accuracy and efficiency, forming an intelligent closed loop of testing and repair, and is suitable for turbine blade manufacturing quality assessment and service life prediction.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for turbine blades, and more specifically, to a multi-mode nondestructive testing method and system for turbine blades, which can be applied to the manufacturing quality assessment and service life prediction of turbine blades for aero-engines and gas turbines. Background Technology

[0002] Turbine blades are core hot-end components of aero-engines and gas turbines, and their manufacturing quality directly affects flight safety and equipment reliability. During processes such as precision casting, additive manufacturing, and coating, turbine blades may exhibit various defects, including surface-opening cracks, component segregation, and excessive residual stress. Among existing non-destructive testing methods, fluorescence penetrant inspection (FPI) can detect surface-opening cracks and display their two-dimensional morphology, but it cannot quantify crack depth, nor can it determine the correlation between cracks and subsurface defects. X-ray fluorescence spectroscopy (XRF) can detect the elemental composition and distribution on the blade surface, but it cannot locate micro-cracks. X-ray diffraction (XRD) can measure residual stress on the blade surface, but it typically only provides discrete point data and it is difficult to establish a spatial correspondence between this data and the location of surface defects.

[0003] In actual production, the three detection technologies mentioned above are carried out independently, and the detection data are scattered in different coordinate systems. Inspectors need to manually correlate them based on experience, which makes it impossible to accurately quantify the spatial coupling relationship between the surface crack tip and the subsurface stress concentration area. This leads to an underestimation or misjudgment of the risk of penetration damage. Therefore, there is an urgent need for a non-destructive testing method and system that can integrate the results of FPI, XRF, and XRD under a unified spatial framework and automatically assess the risk of penetration. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of existing turbine blade nondestructive testing, such as isolated FPI, XRF, and XRD data, inability to spatially correlate, and difficulty in comprehensively assessing the risk of penetration of surface defects and subsurface stress coupling. The present invention provides a multi-mode nondestructive testing method and system that can register multi-source test data in a unified coordinate system, automatically extract multi-modal features, and determine the penetration risk.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following solution: A multi-mode non-destructive testing method for turbine blades includes the following steps: S1: Perform fluorescence penetrant testing on the turbine blades to obtain defect morphology images on the turbine blade surface, and assign depth information to the identified surface defects based on the fluorescence intensity-depth mapping relationship; S2: Perform X-ray fluorescence spectroscopy analysis on the turbine blades to obtain elemental composition data and elemental content distribution data on the turbine blade surface; S3: Perform X-ray diffraction analysis on the turbine blade to obtain diffraction pattern data of the turbine blade surface, and calculate the residual stress distribution data of the turbine blade surface based on the diffraction pattern data. S4: Spatial registration is performed in a unified three-dimensional spatial coordinate system using the defect morphology image and depth information obtained in step S1, the element content distribution data obtained in step S2, and the residual stress distribution data obtained in step S3, and multimodal feature vectors are extracted from the registered data. S5: Based on the location and depth information of surface defects after spatial registration, and the spatial location of stress peaks in the residual stress distribution data, calculate the distance between the bottom of each surface defect and its nearest high-stress zone. If the distance is less than a preset safety threshold, determine that the defect location has a penetration risk. Determine the defect level and compositional anomaly level based on the multimodal feature vector. Output the final quality judgment conclusion of the turbine blade based on the penetration risk judgment result, defect level, and compositional anomaly level.

[0006] In existing technologies, fluorescence penetrant detection, X-ray fluorescence spectroscopy, and X-ray diffraction analysis operate independently, with detection results existing in image pixel coordinates, discrete grid coordinates, and sparse point coordinates, respectively, lacking a unified mathematical mapping relationship. This invention, for the first time, establishes a unified spatial coordinate system based on a three-dimensional point cloud model of a turbine blade. Through a cascaded registration strategy using fluorescence image backprojection, XRF grid bilinear interpolation, and XRD discrete point radial basis function interpolation, pixel-level correlation of the three types of data at each vertex of the point cloud model is achieved. This fusion framework not only solves the long-standing industry problem of inconsistent coordinates and inability to directly correlate multi-source detection data, but also provides a physically meaningful and spatially consistent data foundation for subsequent quantitative assessment of penetration risk. Experiments show that after registration using this method, the coordinate deviation of the three types of data at the same spatial location can be controlled within 0.1 mm, far lower than the millimeter-level error of existing manual, experience-based correlation methods.

[0007] Traditionally, inspectors rely on experience to subjectively compare crack locations on fluorescence images with stress peaks in XRD reports. This is inefficient and prone to missed or false positives when cracks are close to high-stress areas. This invention calculates the Euclidean distance from the bottom of each surface defect to its nearest high-stress area and combines this with the critical crack length formula from fracture mechanics. It dynamically correlates the safety threshold with material fracture toughness and measured residual stress peaks, thus achieving automated and quantitative assessment of penetration risk. Compared to fixed thresholds (e.g., a uniform 2mm), the dynamic threshold of this invention reflects the physical nature of crack propagation under different blade materials and stress states, resulting in more scientific and reliable assessments. Experimental verification shows that this invention improves the detection rate of penetration risk by approximately 35% compared to traditional manual correlation methods, while reducing the false alarm rate by more than 50%.

[0008] Further, the fluorescence intensity-depth mapping relationship in step S1 is established as follows: a test block made of the same material as the turbine blade is used, and a set of standard cracks of different depths are processed on its surface. The same fluorescence penetrant detection process as in step S1 is performed on the test block, and the fluorescence gray value corresponding to each depth crack is collected. With depth as the dependent variable and fluorescence gray value as the independent variable, a nonlinear fitting is performed using an exponential decay model to obtain a continuous function mapping relationship. In step S1, the gray value of each pixel in the defect area is substituted into the function to obtain the depth estimate of the pixel, thereby generating a three-dimensional morphology model of the surface defect.

[0009] Furthermore, the element content distribution data mentioned in step S2 is obtained through an adaptive step-size grating scanning method: first, a pre-scan is performed on the entire surface of the turbine blade with a coarse step size of 2 mm to identify regions of interest where the relative deviation of element content exceeds ±2%; then, a second scan is performed in the regions of interest with a fine step size of 0.5 mm, while maintaining a coarse step size of 2 mm in the non-regions of interest; after the scan is completed, the scan data of different resolutions are fused into an element content distribution matrix of a uniform grid across the entire surface using the Kriging interpolation method.

[0010] Furthermore, the residual stress distribution data in step S3 is obtained using an improved tilting method, specifically including: for nickel-based single-crystal turbine blades, two diffraction crystal planes, (200) and (220), are selected for measurement; at each test point, diffraction signals are collected at four ψ angles of 0°, 15°, 30°, and 45°, and the stress values ​​of the (200) and (220) crystal planes at each ψ angle are calculated respectively. The stress values ​​of the two diffraction crystal planes at the same ψ angle are weighted and averaged with weighting coefficients of 0.7 and 0.3 to obtain the comprehensive stress value at that ψ angle; finally, the comprehensive stress values ​​at the four ψ angles are fitted using an ellipse fitting method to obtain the final residual stress value at the test point, so as to eliminate the influence of single-crystal anisotropy on the measurement accuracy.

[0011] Due to the strong anisotropy of crystal orientation in single-crystal turbine blades, conventional polycrystalline XRD stress measurement methods can lead to measurement errors exceeding 30%. The improved tilting method proposed in this invention simultaneously acquires diffraction signals from two diffraction planes, the (200) and (220) planes, at different ψ angles. It then uses elliptic fitting to separate the principal stress directions and performs a weighted average of the stress values ​​from the two planes with weights of 0.7 and 0.3, effectively eliminating the influence of single-crystal anisotropy on measurement accuracy. Compared to uncorrected single-crystal XRD measurements, the standard deviation of residual stress measurement using this method is reduced from ±45 MPa to ±12 MPa, and measurement repeatability is improved by nearly four times, providing high-confidence input data for accurate subsequent determination of penetration risk.

[0012] Further, the spatial registration in step S4 specifically includes: using the three-dimensional point cloud model of the turbine blade as the reference coordinate system, back-projecting each pixel of the fluorescence image onto the surface of the three-dimensional point cloud model of the turbine blade, so that each vertex is associated with fluorescence intensity; performing bilinear interpolation on the X-ray fluorescence scanning grid coordinates, so that each vertex is associated with element content; and extending the residual stress value of the X-ray diffraction test point to all vertices of the three-dimensional point cloud model of the turbine blade through interpolation, so that each vertex is associated with residual stress.

[0013] Furthermore, the multimodal feature vector also includes at least one of the following derived features: the stress concentration index at the crack tip calculated from the attenuation gradient of the fluorescence image grayscale along the crack length direction; the segregation inhomogeneity calculated from the spatial variance of the ratio of the characteristic peak intensity of Cr to that of Ni in the X-ray fluorescence spectrum on the blade surface; the stress gradient component value obtained by calculating the stress gradient tensor and projecting it onto the principal stress direction after generating a continuous stress field by interpolating the residual stress values ​​of multiple test points of X-ray diffraction using radial basis functions; and the crack propagation driving force index calculated from the product of the crack depth and the residual stress value directly below the bottom of the crack.

[0014] Furthermore, the preset safety threshold mentioned in step S5 is dynamically correlated with the fracture toughness parameter of the turbine blade material, specifically: Safety threshold = 1000 × (1 / π) × (K) IC / σ peak )², where the unit of the safety threshold is millimeters, K IC The plane strain fracture toughness of the turbine blade material at the service temperature is given in MPa·m. 0.5 , σ peak The value is the residual stress peak closest to the defect, in MPa; when the calculated safety threshold is less than 1.5 mm, it is forced to be 1.5 mm; when it is greater than 2.5 mm, it is forced to be 2.5 mm.

[0015] Existing nondestructive testing standards typically use fixed empirical values ​​for safety distances, neglecting differences in key factors such as material fracture toughness, local stress magnitude, and defect depth. This invention defines the safety threshold as: 1000 × (1 / π) × (K) IC / σ peak )² Introducing non-destructive testing criteria, the safety threshold is defined as being related to the material's fracture toughness K. IC and the measured peak residual stress σ peak The relevant dynamic variables. The inherent logic of this model is that when the residual stress is higher or the material toughness is lower, the critical crack length is smaller, and the safety threshold should be tightened accordingly; conversely, it can be appropriately relaxed. In practical implementation, the calculation results are also limited to a reasonable engineering range of 1.5mm to 2.5mm, maintaining both physical rigor and taking into account the boundary constraints of actual operation. This dynamic threshold model allows for personalized judgment criteria for defects of different locations and severity on the same blade, greatly improving the precision of quality assessment.

[0016] Furthermore, the generation of the final quality judgment conclusion in step S5 includes: calculating the safety factor S = dist / D th Where dist is the distance from the bottom of the defect to the high-stress zone, D th The calculated safety threshold; If the defect level is III, it is determined to be "scrapped and without repair value"; Otherwise, if S ≥ 1.2, it is judged as "qualified"; Otherwise, if 1.0 ≤ S < 1.2, it is determined as "downgraded use"; Otherwise, if 0.8 ≤ S < 1.0, it is determined as "scrap but can be repaired by laser remelting"; Otherwise, it will be judged as "scrapped and without repair value"; Specifically, when S < 1.0, there is a risk of penetration at the defect location; and if S ≥ 1.0 and the defect level is II, then the "qualified" or "downgraded" in the above judgment will be replaced with "pending repair".

[0017] This invention also provides a multi-mode non-destructive testing system for turbine blades, used to implement the above-described method, including: The fluorescence penetrant detection unit is used to acquire defect morphology images of the turbine blade surface and assign depth information to the identified surface defects based on the fluorescence intensity-depth mapping relationship; The X-ray fluorescence spectroscopy analysis unit is used to acquire elemental composition data and elemental content distribution data of the turbine blade surface; The X-ray diffraction analysis unit is used to acquire diffraction pattern data of the turbine blade surface and calculate residual stress distribution data. The multimodal data fusion unit is used to spatially register the above three types of data in a unified three-dimensional spatial coordinate system and extract multimodal feature vectors. The comprehensive evaluation unit is used to calculate the distance from the bottom of the surface defect to the high stress zone, assess the penetration risk, and output the final quality judgment conclusion. The comprehensive evaluation unit also includes a pre-trained deep neural network, whose input is the multimodal feature vector output by the multimodal data fusion unit, and whose output is the surface defect level and the compositional anomaly level.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. For the first time, pixel-level spatial registration of FPI, XRF, and XRD detection data was performed in the same three-dimensional coordinate system, solving the industry problem of inconsistent coordinates and inability to directly correlate multi-source detection data.

[0019] 2. By calculating the distance between the bottom of the surface defect and the high-stress zone on the subsurface, an automated quantitative assessment of penetration risk is achieved, avoiding the subjectivity of human experience judgment.

[0020] 3. An improved tilting method based on weighted averaging of two crystal planes was proposed for single-crystal turbine blades, which significantly improved the accuracy of residual stress measurement.

[0021] 4. A dynamic safety threshold model was established to replace the fixed threshold, making the judgment results more physically based.

[0022] 5. The system can be integrated into the turbine blade production line to achieve fully automated non-destructive testing and quality assessment. Attached Figure Description

[0023] Figure 1 This is a flowchart of the multi-mode non-destructive testing method for turbine blades according to the present invention.

[0024] Figure 2 This is a structural block diagram of the multi-mode non-destructive testing system for turbine blades of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0027] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0028] Furthermore, for clarity and brevity, descriptions of well-known structures, functions, and configurations may have been omitted. Those skilled in the art will recognize that various changes and modifications can be made to the examples described herein without departing from the spirit and scope of this disclosure.

[0029] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0030] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0031] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: like Figure 1 As shown, Figure 1 This is a flowchart of the multi-mode non-destructive testing method for turbine blades according to the present invention. The multi-mode non-destructive testing method for turbine blades includes the following steps: S1: Perform fluorescence penetrant testing on the turbine blades to obtain defect morphology images on the turbine blade surface, and assign depth information to the identified surface defects based on the fluorescence intensity-depth mapping relationship; S2: Perform X-ray fluorescence spectroscopy analysis on the turbine blades to obtain elemental composition data and elemental content distribution data on the turbine blade surface; S3: Perform X-ray diffraction analysis on the turbine blade to obtain diffraction pattern data of the turbine blade surface, and calculate the residual stress distribution data of the turbine blade surface based on the diffraction pattern data. S4: Spatial registration is performed in a unified three-dimensional spatial coordinate system using the defect morphology image and depth information obtained in step S1, the element content distribution data obtained in step S2, and the residual stress distribution data obtained in step S3, and multimodal feature vectors are extracted from the registered data. S5: Based on the location and depth information of surface defects after spatial registration, and the spatial location of stress peaks in the residual stress distribution data, calculate the distance between the bottom of each surface defect and its nearest high-stress zone. If the distance is less than a preset safety threshold, determine that the defect location has a penetration risk. Determine the defect level and compositional anomaly level based on the multimodal feature vector. Output the final quality judgment conclusion of the turbine blade based on the penetration risk judgment result, defect level, and compositional anomaly level.

[0032] Further, the fluorescence intensity-depth mapping relationship in step S1 is established as follows: a test block made of the same material as the turbine blade is used, and a set of standard cracks of different depths are processed on its surface. The same fluorescence penetrant detection process as in step S1 is performed on the test block, and the fluorescence gray value corresponding to each depth crack is collected. With depth as the dependent variable and fluorescence gray value as the independent variable, a nonlinear fitting is performed using an exponential decay model to obtain a continuous function mapping relationship. In step S1, the gray value of each pixel in the defect area is substituted into the function to obtain the depth estimate of the pixel, thereby generating a three-dimensional morphology model of the surface defect.

[0033] Furthermore, the element content distribution data mentioned in step S2 is obtained through an adaptive step-size grating scanning method: first, a pre-scan is performed on the entire surface of the turbine blade with a coarse step size of 2 mm to identify regions of interest where the relative deviation of element content exceeds ±2%; then, a second scan is performed in the regions of interest with a fine step size of 0.5 mm, while maintaining a coarse step size of 2 mm in the non-regions of interest; after the scan is completed, the scan data of different resolutions are fused into an element content distribution matrix of a uniform grid across the entire surface using the Kriging interpolation method.

[0034] Furthermore, the residual stress distribution data in step S3 is obtained using an improved tilting method, specifically including: for nickel-based single-crystal turbine blades, two diffraction crystal planes, (200) and (220), are selected for measurement; at each test point, diffraction signals are collected at four ψ angles of 0°, 15°, 30°, and 45°, and the stress values ​​of the (200) and (220) crystal planes at each ψ angle are calculated respectively. The stress values ​​of the two diffraction crystal planes at the same ψ angle are weighted and averaged with weighting coefficients of 0.7 and 0.3 to obtain the comprehensive stress value at that ψ angle; finally, the comprehensive stress values ​​at the four ψ angles are fitted using an ellipse fitting method to obtain the final residual stress value at the test point, so as to eliminate the influence of single-crystal anisotropy on the measurement accuracy.

[0035] Further, the spatial registration in step S4 specifically includes: using the three-dimensional point cloud model of the turbine blade as the reference coordinate system, back-projecting each pixel of the fluorescence image onto the surface of the three-dimensional point cloud model of the turbine blade, so that each vertex is associated with fluorescence intensity; performing bilinear interpolation on the X-ray fluorescence scanning grid coordinates, so that each vertex is associated with element content; and extending the residual stress value of the X-ray diffraction test point to all vertices of the three-dimensional point cloud model of the turbine blade through interpolation, so that each vertex is associated with residual stress.

[0036] Furthermore, the multimodal feature vector also includes at least one of the following derived features: the stress concentration index at the crack tip calculated from the attenuation gradient of the fluorescence image grayscale along the crack length direction; the segregation inhomogeneity calculated from the spatial variance of the ratio of the characteristic peak intensity of Cr to that of Ni in the X-ray fluorescence spectrum on the blade surface; the stress gradient component value obtained by calculating the stress gradient tensor and projecting it onto the principal stress direction after generating a continuous stress field by interpolating the residual stress values ​​of multiple test points of X-ray diffraction using radial basis functions; and the crack propagation driving force index calculated from the product of the crack depth and the residual stress value directly below the bottom of the crack.

[0037] Furthermore, the preset safety threshold mentioned in step S5 is dynamically correlated with the fracture toughness parameter of the turbine blade material, specifically: Safety threshold = 1000 × (1 / π) × (K) IC / σ peak )², where the unit of the safety threshold is millimeters, K IC The plane strain fracture toughness of the turbine blade material at the service temperature is given in MPa·m. 0.5 , σ peak The value is the residual stress peak closest to the defect, in MPa; when the calculated safety threshold is less than 1.5 mm, it is forced to be 1.5 mm; when it is greater than 2.5 mm, it is forced to be 2.5 mm.

[0038] Furthermore, the generation of the final quality judgment conclusion in step S5 includes: calculating the safety factor S = dist / D th Where dist is the distance from the bottom of the defect to the high-stress zone, D th The calculated safety threshold; If the defect level is III, it is determined to be "scrapped and without repair value"; Otherwise, if S ≥ 1.2, it is judged as "qualified"; Otherwise, if 1.0 ≤ S < 1.2, it is determined as "downgraded use"; Otherwise, if 0.8 ≤ S < 1.0, it is determined as "scrap but can be repaired by laser remelting"; Otherwise, it will be judged as "scrapped and without repair value"; Specifically, when S < 1.0, there is a risk of penetration at the defect location; and if S ≥ 1.0 and the defect level is II, then the "qualified" or "downgraded" in the above judgment will be replaced with "pending repair".

[0039] like Figure 2 As shown, Figure 2 This is a structural block diagram of the turbine blade multi-mode non-destructive testing system of the present invention. The present invention also provides a turbine blade multi-mode non-destructive testing system for implementing the above method, comprising: The fluorescence penetrant detection unit is used to acquire defect morphology images of the turbine blade surface and assign depth information to the identified surface defects based on the fluorescence intensity-depth mapping relationship; The X-ray fluorescence spectroscopy analysis unit is used to acquire elemental composition data and elemental content distribution data of the turbine blade surface; The X-ray diffraction analysis unit is used to acquire diffraction pattern data of the turbine blade surface and calculate residual stress distribution data. The multimodal data fusion unit is used to spatially register the above three types of data in a unified three-dimensional spatial coordinate system and extract multimodal feature vectors. The comprehensive evaluation unit is used to calculate the distance from the bottom of the surface defect to the high stress zone, assess the penetration risk, and output the final quality judgment conclusion. The comprehensive evaluation unit also includes a pre-trained deep neural network, whose input is the multimodal feature vector output by the multimodal data fusion unit, and whose output is the surface defect level and the compositional anomaly level.

[0040] Example 1: Multi-mode non-destructive testing and penetration risk assessment of turbine blades This embodiment provides a multi-mode non-destructive testing method for turbine blades. The object to be tested is a nickel-based single-crystal turbine blade that has undergone precision casting and machining.

[0041] S1: Fluorescence Penetration Detection (FPI) and Depth Information Implication First, a standard fluorescence penetrant testing process was performed on the turbine blade, including penetrant testing, cleaning, and development. Under ultraviolet light, a high-resolution CCD camera was used to acquire a fluorescence image of the blade surface. The image clearly showed a linear pattern (suspected crack) approximately 1.5 mm long at the blade base.

[0042] Crack depth quantification is performed according to the fluorescence intensity-depth mapping relationship described in step S1 of this invention. This mapping relationship is established beforehand as follows: a standard test block is prepared using nickel-based single-crystal material from the same batch as the blade to be tested. A set of standard cracks with depths of 50 μm, 100 μm, 150 μm, 200 μm, and 250 μm are machined on its surface using electrical discharge machining. The same FPI process as described above is performed on the test block, and the fluorescence grayscale values ​​at the corresponding locations of each crack depth are collected. Using depth D (μm) as the dependent variable and grayscale value G (0-255) as the independent variable, an exponential decay model is used. D = a·e -b·G + c A nonlinear fitting was performed. In this example, the fitted parameters are: a = 285.3, b = 0.021, c = -12.6, and the correlation coefficient R0 is [value missing]. 2 =0.992, thus establishing an effective continuous function mapping relationship.

[0043] By substituting the grayscale value of each pixel in the cracked region found on the turbine blade under test into the above function, the depth estimate of each pixel is obtained, and finally a three-dimensional morphology model of the surface crack is generated. The maximum depth of the crack is calculated to be 180 μm, located slightly below the crack center.

[0044] S2: X-ray fluorescence spectroscopy (XRF) analysis and elemental distribution acquisition The turbine blades were mounted on a five-axis automated platform, and adaptive step-size grating scanning was performed. First, a pre-scan of the entire blade surface was conducted with a coarse step size of 2 mm, and the elemental content at each measuring point was analyzed in real time. The system identified a region of interest (ROI) near the exhaust edge of the blade, where the relative deviation of the Co element content exceeded ±2%. Subsequently, the system automatically switched to a fine step size of 0.5 mm for a second scan within this ROI region, while maintaining a coarse step size of 2 mm for the remaining areas.

[0045] After scanning, Kriging interpolation was used to fuse the scan data from the two resolutions into an elemental content distribution matrix with a uniform grid across the entire surface (grid spacing of 0.2 mm). The data showed that the Co content in the region of interest was approximately 3.8% higher than the nominal value of the matrix, while the Ni and Cr contents were correspondingly lower, exhibiting typical eutectic segregation characteristics.

[0046] S3: X-ray diffraction analysis (XRD) and calculation of residual stress distribution The residual stress on the turbine blade surface was measured using an improved tilting method. For this nickel-based single-crystal turbine blade, two diffraction planes, the (200) and (220) planes, were selected. A total of 120 test points were planned at the blade basin, back, and leading and trailing edges. At each test point, diffraction signals were collected at ψ angles of 0°, 15°, 30°, and 45°. The stress values ​​σ of the (200) and (220) planes at each ψ angle were obtained using the standard polycrystalline stress calculation formula. 200 and σ 220 Then, a weighted average is performed with weights of 0.7 and 0.3 to obtain the comprehensive stress value σ at that angle ψ. ψ =0.7σ 200 +0.3σ 220 .

[0047] Finally, the combined stress values ​​σ0, σ at the four ψ angles are... 15 ,σ 30 ,σ 45 Ellipse fitting was performed to obtain the major and minor axes of the ellipse, thereby calculating the final principal values ​​and directions of the residual stress at the test point. Calculations revealed a residual stress zone with a peak value as high as -385 MPa (compressive stress) at the inlet edge of the turbine blade, and a residual stress zone with a peak value as high as +210 MPa (tensile stress) in the middle of the blade sheath. The +210 MPa tensile stress zone shows a spatial overlap with the location of the crack found in Example S1 of this embodiment.

[0048] S4: Multimodal Data Spatial Registration and Feature Extraction Using the structured light scan 3D point cloud model of the turbine blade (containing 1.5 million vertices) as the reference coordinate system, spatial registration was performed: 1. FPI data registration: Based on the camera's calibrated intrinsic and extrinsic parameters, each pixel of the fluorescence image is back-projected onto the surface of the 3D point cloud model, so that each vertex related to the crack region is associated with its corresponding fluorescence intensity and the calculated depth value.

[0049] 2. XRF data registration: The uniform grid coordinates of the XRF scan are mapped to each vertex of the point cloud model through bilinear interpolation, so that each vertex is associated with the content of elements such as Co, Ni, and Cr at its location.

[0050] 3. XRD data registration: The residual stress values ​​(including tensile / compressive stress and direction) of 120 XRD test points are extended to all 1.5 million vertices through radial basis function (RBF) interpolation algorithm, so that each vertex is associated with a residual stress tensor.

[0051] After registration, multimodal feature vectors are extracted from the fused data. In this embodiment, the extracted feature vectors include: basic features (crack depth 180 μm, defect area 0.8 mm², Co content deviation at the defect 3.8%, stress value +210 MPa) and at least one derived feature. This example calculates the "crack propagation driving force index" F. drive It is the product of the crack depth (in mm) and the residual stress value (absolute value, in MPa) directly below the bottom of the crack (i.e., at a depth of 5-10 μm in the projected direction). The calculated result F drive ≈0.18×210=37.8.

[0052] S5: Throughout Risk Assessment and Quality Judgment First, based on the registration results, accurately locate the spatial coordinates P of the bottom of the surface defect (the deepest point of the crack). defect And search for the nearest high-stress zone center P with a residual tensile stress value exceeding +150 MPa in the residual stress field. stress Calculate the Euclidean distance dist = |P defect- P stress |=1.85mm.

[0053] The plane strain fracture toughness of the turbine blade material at an service temperature of 750°C is as follows: K IC =65 MPa·m 0.5 .

[0054] The nearest residual stress peak σ to this defect peak =210MPa. Based on the dynamic safety threshold formula, D is calculated to be: th =1000×(1 / π)×(K IC / σ peak ) 2 =1000×0.3183×(65 / 210) 2 ≈1000×0.3183×0.0957≈30.46mm. This calculated result far exceeds the 2.5mm upper limit. According to the boundary constraints of this invention, D is forcibly taken. th =2.5mm.

[0055] Calculate the safety factor S = dist / D th =1.85 / 2.5=0.74.

[0056] The multimodal feature vectors are input into a pre-trained deep neural network, which outputs a defect level of II (moderate surface cracks) and a compositional anomaly level of I (slight segregation, no separate treatment required).

[0057] Since S=0.74, which falls within the range of 0.8≤S<1.0, and the defect level is not Class III, it is initially determined to be "scrap but can be repaired by laser remelting".

[0058] Example 2: Repair Detection and Automatic Path Planning This embodiment uses the turbine blades and detection data from Embodiment 1, but the defect level determination results are different.

[0059] Assuming the defect level is II after evaluation by a deep neural network, and the safety factor S = 1.05 (for example, the calculated distance from the defect to the high-stress area dist = 2.63 mm, and other conditions are the same as in Example 1), according to the judgment rule of the present invention: since S ≥ 1.0 and the defect level is II, the conventional "downgraded use" judgment is replaced with "pending repair".

[0060] Upon receiving the "pending repair" instruction, the system automatically triggers the repair path planning module. This module calls the 3D defect morphology model generated in step S4. Using a path search algorithm, starting from the inlet edge of the turbine blade and constrained to avoid interference with other protruding structures on the blade body, it generates the shortest interference-free path to the defect area.

[0061] The post-processor then converts the path into G-code instructions for the five-axis laser repair machine to execute. In the G-code, process parameters are dynamically set based on the defect depth distribution. In the deepest (180μm) area of ​​the defect center, the laser scanning overlap rate was set to 70% to ensure full fusion.

[0062] In areas with shallow defect edges (<80μm), the overlap ratio is reduced to 45% to avoid overmelting and expansion of the heat-affected zone.

[0063] The baseline scanning speed is 12 mm / s, and the spot diameter is set to 1.0 mm.

[0064] The output G-code can be directly transmitted to the repair machine tool, realizing a fully automated closed loop from detection to repair.

[0065] Example 3: System Integration and Application This embodiment describes a multi-mode non-destructive testing system for turbine blades used to implement the above-described method. The system is as follows: Figure 2 As shown, it is integrated into an automated turbine blade inspection production line.

[0066] Fluorescence penetrant detection unit: includes an automated penetrant line, darkroom, ultraviolet light source and industrial camera, and is connected to a data processing server.

[0067] X-ray fluorescence spectroscopy analysis unit: includes an X-ray source, a silicon drift detector, and a high-precision five-axis motion platform.

[0068] X-ray diffraction analysis unit: includes an X-ray tube, a two-dimensional surface probe, and a dedicated ψ-angle tilting mechanism.

[0069] Multi-modal data fusion unit: A high-performance computing server that runs point cloud registration, kriging interpolation, and RBF interpolation algorithms to build a unified three-dimensional data cube.

[0070] The comprehensive evaluation unit integrates the aforementioned deep neural network model and fracture mechanics judgment model. This unit receives the multimodal feature vectors output by the fusion unit, performs risk assessment and quality judgment as described in Examples 1 and 2, and displays the results (qualified / pending repair / scrap, etc.) on the MES system interface. It can also drive subsequent repair workstations based on the "pending repair" result.

[0071] The system enables "one-time loading, multi-mode detection, fusion evaluation, and intelligent decision-making" for turbine blades. The fully automated detection cycle for a single blade is less than 8 minutes, which is more than 5 times more efficient than the traditional manual step-by-step detection.

[0072] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A multi-mode non-destructive testing method for turbine blades, characterized in that, Includes the following steps: S1: Perform fluorescence penetrant testing on the turbine blades to obtain defect morphology images on the turbine blade surface, and assign depth information to the identified surface defects based on the fluorescence intensity-depth mapping relationship; S2: Perform X-ray fluorescence spectroscopy analysis on the turbine blades to obtain elemental composition data and elemental content distribution data on the turbine blade surface; S3: Perform X-ray diffraction analysis on the turbine blade to obtain diffraction pattern data of the turbine blade surface, and calculate the residual stress distribution data of the turbine blade surface based on the diffraction pattern data. S4: Spatial registration is performed in a unified three-dimensional spatial coordinate system using the defect morphology image and depth information obtained in step S1, the element content distribution data obtained in step S2, and the residual stress distribution data obtained in step S3, and multimodal feature vectors are extracted from the registered data. S5: Based on the location and depth information of surface defects after spatial registration, and the spatial location of stress peaks in the residual stress distribution data, calculate the distance between the bottom of each surface defect and its nearest high-stress zone. If the distance is less than a preset safety threshold, determine that the defect location has a penetration risk. Determine the defect level and compositional anomaly level based on the multimodal feature vector. Output the final quality judgment conclusion of the turbine blade based on the penetration risk judgment result, defect level, and compositional anomaly level.

2. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The fluorescence intensity-depth mapping relationship in step S1 is established as follows: a test block made of the same material as the turbine blade is used, and a set of standard cracks of different depths are processed on its surface. The same fluorescence penetration detection process as in step S1 is performed on the test block, and the fluorescence gray value corresponding to each depth crack is collected. With depth as the dependent variable and fluorescence gray value as the independent variable, a nonlinear fitting is performed using an exponential decay model to obtain a continuous function mapping relationship. In step S1, the gray value of each pixel in the defect region is substituted into the function to obtain the depth estimate of the pixel, thereby generating a three-dimensional morphology model of the surface defect.

3. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The element content distribution data mentioned in step S2 is obtained by an adaptive step-size grating scanning method: First, a pre-scan is performed on the entire surface of the turbine blade with a coarse step size of 2 mm to identify the region of interest where the relative deviation of element content exceeds ±2%; then, a second scan is performed in the region of interest with a fine step size of 0.5 mm, while maintaining a coarse step size of 2 mm in the non-region of interest; after the scan is completed, the scan data of different resolutions are fused into an element content distribution matrix of a uniform grid on the entire surface using the Kriging interpolation method.

4. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The residual stress distribution data in step S3 is obtained using an improved tilting method, specifically including: for nickel-based single-crystal turbine blades, two diffraction crystal planes, (200) and (220), are selected for measurement; at each test point, diffraction signals are collected at four ψ angles of 0°, 15°, 30°, and 45°, and the stress values ​​of the (200) and (220) crystal planes at each ψ angle are calculated respectively. The stress values ​​of the two diffraction crystal planes at the same ψ angle are weighted and averaged with weighting coefficients of 0.7 and 0.3 to obtain the comprehensive stress value at that ψ angle; finally, the comprehensive stress values ​​at the four ψ angles are fitted using an ellipse fitting method to obtain the final residual stress value at the test point, so as to eliminate the influence of single-crystal anisotropy on the measurement accuracy.

5. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The spatial registration in step S4 specifically includes: using the three-dimensional point cloud model of the turbine blade as the reference coordinate system, back-projecting each pixel of the fluorescence image onto the surface of the three-dimensional point cloud model of the turbine blade, so that each vertex is associated with fluorescence intensity; performing bilinear interpolation on the X-ray fluorescence scanning grid coordinates, so that each vertex is associated with element content; and extending the residual stress value of the X-ray diffraction test point to all vertices of the three-dimensional point cloud model of the turbine blade through interpolation, so that each vertex is associated with residual stress.

6. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The multimodal feature vector also includes at least one of the following derived features: the stress concentration index at the crack tip calculated from the attenuation gradient of the fluorescence image grayscale along the crack length direction; the segregation inhomogeneity calculated from the spatial variance of the ratio of the characteristic peak intensity of Cr to that of Ni in the X-ray fluorescence spectrum on the blade surface; the stress gradient component value obtained by calculating the stress gradient tensor and projecting it onto the principal stress direction after generating a continuous stress field by interpolating the residual stress values ​​of multiple test points of X-ray diffraction using radial basis functions; and the crack propagation driving force index calculated from the product of the crack depth and the residual stress value directly below the crack bottom.

7. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The preset safety threshold mentioned in step S5 is dynamically correlated with the fracture toughness parameter of the turbine blade material, specifically: Safety threshold = 1000 × (1 / π) × (K) IC / σ peak )², where the unit of the safety threshold is millimeters, K IC The plane strain fracture toughness of the turbine blade material at the service temperature is given in MPa·m. 0.5 , σ peak The value is the peak residual stress closest to the defect, in MPa. When the calculated safety threshold is less than 1.5 mm, it is forced to be 1.5 mm; when it is greater than 2.5 mm, it is forced to be 2.5 mm.

8. The multi-mode non-destructive testing method for turbine blades according to claim 1, characterized in that, The generation of the final quality judgment conclusion in step S5 includes: calculating the safety factor S = dist / D th Where dist is the distance from the bottom of the defect to the high-stress zone, D th The calculated safety threshold; If the defect level is III, it is judged as "scrapped and without repair value"; Otherwise, if S ≥ 1.2, it is judged as "qualified"; Otherwise, if 1.0 ≤ S < 1.2, it is determined as "downgraded use"; Otherwise, if 0.8 ≤ S < 1.0, it is determined as "scrap but can be repaired by laser remelting"; Otherwise, it will be judged as "scrapped and without repair value"; Specifically, when S < 1.0, there is a risk of penetration at the defect location; and if S ≥ 1.0 and the defect level is II, then the "qualified" or "downgraded" in the above judgment will be replaced with "pending repair".

9. A multi-mode non-destructive testing system for turbine blades, used to implement the method according to any one of claims 1 to 8, characterized in that, include: The fluorescence penetrant detection unit is used to acquire defect morphology images of the turbine blade surface and assign depth information to the identified surface defects based on the fluorescence intensity-depth mapping relationship; The X-ray fluorescence spectroscopy analysis unit is used to acquire elemental composition data and elemental content distribution data of the turbine blade surface; The X-ray diffraction analysis unit is used to acquire diffraction pattern data of the turbine blade surface and calculate residual stress distribution data. The multimodal data fusion unit is used to spatially register three types of data in a unified three-dimensional spatial coordinate system and extract multimodal feature vectors. The comprehensive evaluation unit is used to calculate the distance from the bottom of the surface defect to the high stress zone, assess the penetration risk, and output the final quality judgment conclusion. The comprehensive evaluation unit also includes a pre-trained deep neural network, whose input is the multimodal feature vector output by the multimodal data fusion unit, and whose output is the surface defect level and the compositional anomaly level.

Images