Aluminum alloy component quality detection method and system

Abstract

The application discloses a kind of aluminum alloy component quality detection method and system, it is related to intelligent detection technical field, including the complex optical field of calculating optical image image plane, the relative phase difference of calculating image plane complex optical field and reference optical field, conversion is the normal microdisplacement of component surface, generate space-time microdisplacement field;The value interval of time coefficient and dynamic reference value is divided into interval grid respectively, calculate maximum normalized empirical mutual information, generate defect sensitive mode set, calculate the residual field of sensitive mode, and carry out binary processing, generate defect binary mask, and calculate defect position.The application is analyzed by combining the multi-source information of optical image and vibration data, and reconstructs and POD mode analysis using space-time microdisplacement field, not only enhances the detection capability of aluminum alloy component surface small deformation, also improves the comprehensive analysis capability to dynamic response.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection technology, and in particular to a method and system for quality inspection of aluminum alloy components. Background Technology

[0002] With the continuous development of industrial manufacturing, aluminum alloys, as a lightweight and high-strength material, are widely used in aerospace, automobiles, construction and other fields. During the production and use of aluminum alloy components, they are easily affected by various mechanical forces, heat and environmental factors, which can lead to minor defects. These defects can not only affect the quality and performance of products, but also cause safety hazards. How to conduct efficient and accurate quality inspection of aluminum alloy components, especially real-time monitoring on the production line or on-site environment, has become a research hotspot in the current manufacturing industry.

[0003] Existing technologies still face some challenges when dealing with complex aluminum alloy components. Existing methods often rely on a single data source for analysis, resulting in limited ability to detect complex defects. Relying solely on optical images cannot accurately capture minute vibrations or deformations on the aluminum alloy surface, while relying solely on vibration data lacks effective analysis of surface morphology changes. Existing vibration analysis methods often ignore the influence of spatiotemporal factors, making it difficult to fully reflect the dynamic response of aluminum alloy components under excitation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the inventors have proposed the present invention.

[0005] Therefore, this invention provides a method and system for quality inspection of aluminum alloy components, which solves the problem that existing methods often rely on a single data source for analysis, resulting in limited ability to detect complex defects. Simply relying on optical images cannot accurately capture the minute vibrations or deformations on the aluminum alloy surface, while relying solely on vibration data lacks effective analysis of surface morphology changes. Existing vibration analysis methods often ignore the influence of spatiotemporal factors, making it difficult to fully reflect the dynamic response of aluminum alloy components under excitation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for quality inspection of aluminum alloy components, comprising the following steps:

[0008] Collect and preprocess multi-source data, define a reference intensity image, execute an iterative phase retrieval algorithm on the reference intensity image to generate a reference light field, calculate the complex light field of the image plane of the optical image, calculate the relative phase difference between the complex light field of the image plane and the reference light field, convert it into the normal micro-displacement of the component surface, and generate a spatiotemporal micro-displacement field.

[0009] The vibration velocity sequence is mapped to the pseudo-state space to form the phase space trajectory matrix, and singular value decomposition is performed. The left singular vector corresponding to the largest singular value in the decomposition result is extracted and defined as the dynamic reference value. The spatiotemporal micro-displacement field is reshaped into a data matrix, and the fluctuation data matrix and spatiotemporal correlation matrix are calculated. The spatiotemporal correlation matrix is ​​decomposed by eigenvalue decomposition to solve the time coefficient vector. By projecting the fluctuation data matrix onto the time coefficient vector, the vector of the POD spatial mode is calculated to obtain the spatial mode.

[0010] The time coefficient and dynamic reference value ranges are divided into interval grids, the maximum normalized empirical mutual information is calculated and defined as the maximum information coefficient, the POD mode index is rearranged from largest to smallest according to the maximum information coefficient, a defect sensitive mode set is generated, the residual field of the sensitive mode is calculated and binarized to generate a defect binary mask, and the defect location is calculated.

[0011] The multi-source data includes optical images and vibration velocities.

[0012] As a preferred embodiment of the aluminum alloy component quality inspection method of the present invention, the step of calculating the complex light field of the image plane of the optical image, calculating the relative phase difference between the complex light field of the image plane and the reference light field, converting it into a normal micro-displacement of the component surface, and generating a spatiotemporal micro-displacement field includes:

[0013] An optical image of the component when it is completely stationary is acquired and defined as a reference intensity image;

[0014] An iterative phase retrieval algorithm is performed on the reference intensity image. The square root of the reference intensity image is set as the initial image plane light field on the object surface. The initial image plane light field is propagated to the camera target surface through the angular spectrum propagation formula, and the complex amplitude of the image plane light field is calculated.

[0015] Replace the amplitude of the image plane light field complex amplitude with the square root of the reference intensity image to generate the updated image plane light field complex amplitude;

[0016] The updated image plane light field is backpropagated to the object plane to obtain the object plane light field estimate for the next iteration. A maximum number of iterations is set. When the maximum number of iterations is reached, the image plane light field calculated by the last forward propagation is defined as the reference light field.

[0017] Calculate the complex light field of the image plane of an optical image;

[0018] The relative phase difference between the complex optical field on the image plane and the reference optical field is calculated by complex conjugate multiplication. The phase angle of the relative phase difference is extracted to obtain the wrapped relative phase difference field. The wrapped relative phase difference field is then subjected to phase unwrapping processing to obtain the continuous absolute phase difference field.

[0019] Based on the fundamental formulas of laser interferometry, the absolute phase difference is converted into a normal micro-displacement on the surface of the component;

[0020] Using spatiotemporal tensor reconstruction, the normal micro-displacements are arranged according to their spatial coordinates in the image and the order of acquisition time to form a three-dimensional data array, which is defined as the spatiotemporal micro-displacement field.

[0021] As a preferred embodiment of the aluminum alloy component quality inspection method of the present invention, wherein: the step of mapping the vibration velocity sequence to a pseudo-state space to form a phase space trajectory matrix, and performing singular value decomposition, extracting the left singular vector corresponding to the maximum singular value in the decomposition result, and defining it as a dynamic reference value, includes:

[0022] Calculate the mean of the vibration velocity and subtract the mean from the vibration velocity to obtain the detrended vibration velocity. Sort the detrended vibration velocity sequence according to time order to generate the detrended vibration velocity sequence. Apply zero-phase filtering to the detrended vibration velocity sequence to obtain the filtered vibration velocity sequence.

[0023] By setting a time delay and an embedding dimension m, the filtered vibration velocity sequence is mapped to an m-dimensional pseudo-state space to obtain a phase space vector.

[0024] Arrange the phase space vectors in time order to form a phase space trajectory matrix. Perform singular value decomposition on the phase space trajectory matrix to obtain an orthogonal matrix and a diagonal matrix. The diagonal elements are called singular values.

[0025] Extract the left singular vector corresponding to the largest singular value in the decomposition result and define it as the dynamic reference value;

[0026] The spatiotemporal micro-displacement field is aligned with the time frame of the dynamic reference value by linear interpolation to obtain the final dynamic reference value.

[0027] As a preferred embodiment of the aluminum alloy component quality inspection method of the present invention, the method involves: reshaping the spatiotemporal micro-displacement field into a data matrix, calculating the fluctuation data matrix and the spatiotemporal correlation matrix, performing eigenvalue decomposition on the spatiotemporal correlation matrix to solve for the time coefficient vector, and calculating the vector of the POD spatial mode by projecting the fluctuation data matrix onto the time coefficient vector to obtain the spatial mode, including:

[0028] The spatiotemporal micro-displacement field is reshaped into a data matrix. The mean of each row of the data matrix is ​​calculated to obtain the average displacement field. The average displacement field is then subtracted from each row of the data matrix to obtain the fluctuation data matrix. The spatiotemporal correlation matrix of the fluctuation field is then calculated.

[0029] Eigenvalue decomposition is performed on the spatiotemporal correlation matrix to obtain the eigenvalues ​​of the POD mode and its corresponding time coefficient vector. The time coefficient vector is defined as the time coefficient of the POD mode.

[0030] By projecting the fluctuation data matrix onto the time coefficient vector, the vector of the POD spatial mode is calculated, and the vector of the POD spatial mode is reshaped to obtain the spatial mode.

[0031] As a preferred embodiment of the aluminum alloy component quality inspection method of the present invention, the method involves dividing the value ranges of the time coefficient and the dynamic reference value into interval grids, calculating the maximum normalized empirical mutual information (defined as the maximum information coefficient), rearranging the POD modal indexes from largest to smallest according to the maximum information coefficient, and generating a defect-sensitive mode set, including:

[0032] The ranges of time coefficient and final dynamic reference value are divided into interval grids, and the empirical mutual information under the grid division is calculated.

[0033] Calculate the maximum normalized empirical mutual information, which is defined as the maximum information coefficient;

[0034] The POD modal indexes are rearranged from largest to smallest according to the maximum information coefficient to obtain a new index sequence. The number of sensitive modes is set using the fixed number method, and the corresponding mode pairs are extracted from the POD modal set to form a defect sensitive mode set.

[0035] As a preferred embodiment of the aluminum alloy component quality inspection method of the present invention, the step of calculating the residual field of the sensitive mode, performing binarization processing to generate a defect binary mask, and calculating the defect location includes:

[0036] Based on each sensitive mode in the defect sensitive mode set, the residual field of the sensitive mode is calculated. For each pixel position in the residual field, the maximum value of all residual field values ​​at that position is taken as the comprehensive residual field.

[0037] The optimal segmentation threshold is set using the Otsu method, the comprehensive residual field is binarized to generate an initial defect binary mask, and morphological opening operation is performed on the initial defect binary mask.

[0038] Connected component labeling is performed on the morphologically processed binary image to identify all connected foreground regions. The centroid coordinates of each connected region are calculated and defined as the spatial location region of the defect.

[0039] As a preferred embodiment of the aluminum alloy component quality inspection method of the present invention, the step of collecting multi-source data and performing preprocessing includes:

[0040] Multi-source data of aluminum alloy components are collected through smart sensors and then denoised and standardized.

[0041] The intelligent sensor includes a high-speed camera and a laser vibrometer.

[0042] Secondly, the present invention provides a quality inspection system for aluminum alloy components, comprising:

[0043] The data collection and processing module is used to collect multi-source data of aluminum alloy components and perform noise reduction and standardization processing.

[0044] The displacement recovery module is used to define a reference intensity image, perform an iterative phase recovery algorithm on the reference intensity image, generate a reference light field, calculate the complex light field of the image plane of the optical image, calculate the relative phase difference between the complex light field of the image plane and the reference light field, convert it into the normal micro-displacement of the component surface, and generate a spatiotemporal micro-displacement field.

[0045] The decomposition mode module is used to map the vibration velocity sequence to the pseudo-state space, construct the phase space trajectory matrix, and perform singular value decomposition. It extracts the left singular vector corresponding to the maximum singular value in the decomposition result and defines it as the dynamic reference value. It reshapes the spatiotemporal micro-displacement field into a data matrix, calculates the fluctuation data matrix and the spatiotemporal correlation matrix, performs eigenvalue decomposition on the spatiotemporal correlation matrix, solves the time coefficient vector, and calculates the vector of the POD spatial mode by projecting the fluctuation data matrix onto the time coefficient vector to obtain the spatial mode.

[0046] The defect detection module is used to divide the value ranges of the time coefficient and the dynamic reference value into interval grids, calculate the maximum normalized empirical mutual information, which is defined as the maximum information coefficient, rearrange the POD mode indexes in descending order of the maximum information coefficient, generate a defect-sensitive mode set, calculate the residual field of the sensitive mode, perform binarization processing, generate a defect binary mask, and calculate the defect location.

[0047] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the aluminum alloy component quality inspection method as described in the first aspect of the present invention.

[0048] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the aluminum alloy component quality inspection method as described in the first aspect of the present invention.

[0049] The beneficial effects of this invention are as follows: By combining multi-source information analysis of optical images and vibration data, and utilizing the reconstruction of spatiotemporal micro-displacement fields and POD modal analysis, this invention not only enhances the detection capability of small deformations on the surface of aluminum alloy components, but also improves the comprehensive analysis capability of dynamic response, thereby significantly improving detection accuracy and defect location capability. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of the quality inspection method for aluminum alloy components in Example 1.

[0052] Figure 2 This is a schematic diagram of the aluminum alloy component quality inspection system in Example 1. Detailed Implementation

[0053] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0056] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for quality inspection of aluminum alloy components, including the following steps:

[0057] S1. Collect multi-source data and preprocess it, define a reference intensity image, execute an iterative phase retrieval algorithm on the reference intensity image to generate a reference light field, calculate the complex light field of the image plane of the optical image, calculate the relative phase difference between the complex light field of the image plane and the reference light field, convert it into the normal micro-displacement of the component surface, and generate a spatiotemporal micro-displacement field.

[0058] Specifically, this involves collecting and preprocessing multi-source data, including:

[0059] Multi-source data of aluminum alloy components are collected through smart sensors and then denoised and standardized.

[0060] The intelligent sensor includes a high-speed camera and a laser vibrometer;

[0061] The multi-source data includes optical images and vibration velocities;

[0062] The aluminum alloy component under test is fixed to an optical vibration isolation platform using a non-rigid clamp (such as a low-stiffness elastic rope) to isolate it from environmental vibration interference. An electromagnetic exciter is used, with its excitation head making perpendicular contact with a pre-marked excitation point on the component surface via a force sensor. The exciter receives a standard half-sine pulse voltage signal from a signal generator and converts it into a standard half-sine pulse excitation force applied to the component. The formula is as follows:

[0063] ,

[0064] in For the excitation force at time t, The amplitude of the pulse force is set based on the component mass and the desired vibration amplitude. Where t is the pulse width and t is the time variable;

[0065] At the rising edge of the excitation pulse, the central control unit generates a synchronous TTL trigger signal, which is simultaneously sent to the high-speed camera and laser vibration meter to acquire optical images and vibration velocities on the component surface.

[0066] By combining optical images and vibration velocity data, richer information can be provided, covering the morphological information of the component surface and its dynamic response. Denoising and standardization can effectively improve the reliability of subsequent analysis, remove interference factors, and reduce the impact of measurement noise, thereby ensuring the accuracy and reliability of the analysis results. Using non-rigid fixtures and optical vibration isolation platforms can effectively isolate the interference of environmental vibration on the measurement results, ensuring that the detection accuracy can still be maintained under complex working conditions.

[0067] Furthermore, the complex light field of the image plane is calculated, the relative phase difference between the complex light field and the optical field is calculated, and converted into the normal micro-displacement of the component surface, generating a spatiotemporal micro-displacement field, including:

[0068] Before the excitation is applied, an optical image of the component when it is completely stationary is acquired and defined as the reference intensity image;

[0069] An iterative phase retrieval algorithm (Gerchberg-Saxton algorithm) is applied to the reference intensity image. The square root of the reference intensity image is used as the initial image plane light field on the object surface (component surface), with the initial phase set to zero. The initial image plane light field is propagated to the camera target surface using the angular spectrum propagation formula. The complex amplitude of the image plane light field is calculated using the following formula:

[0070] ,

[0071] ,

[0072] in Let U be the complex amplitude of the image plane light field in the l-th iteration, and let U represent the complex light field, which is a complex number containing amplitude and phase information. The two-dimensional spatial coordinates of the camera image plane. and These are the two-dimensional inverse Fourier transform operator and the two-dimensional Fourier transform operator, respectively. This is the current estimate of the complex amplitude of the optical field on the object plane (i.e., the surface of the aluminum alloy component) at the l-th iteration. The coordinates of the component's surface are two-dimensional spatial coordinates. and Spatial frequency coordinates Let be the angular spectral transfer function, describing the effect of light propagating a distance z in free space, where z is the optical propagation distance from the object plane (component surface) to the image plane (camera target surface), measured by a laser rangefinder, and j is the imaginary unit. The wavelength of light waves in the propagation medium;

[0073] The square root of the reference intensity image is used to replace the amplitude of the complex amplitude of the image plane light field, while preserving the phase, to generate the updated complex amplitude of the image plane light field. The formula is as follows:

[0074] ,

[0075] ,

[0076] in The image plane light field complex amplitude is updated after the l-th iteration. The image intensity (grayscale value) of the reference frame. To perform phase angle calculations for complex numbers, the result is in interval, This refers to the phase component of the image plane light field calculated during the l-th iteration;

[0077] The updated image plane light field is backpropagated to the object plane to obtain the object plane light field estimate for the next iteration, as shown in the formula:

[0078] ,

[0079] in This is the new estimate of the complex amplitude of the object plane light field that will be used in the (l+1)th iteration (i.e., the next iteration). For the complex conjugate of the angular spectrum transfer function;

[0080] Based on empirical engineering practices, a maximum number of iterations is set. Once the maximum number of iterations is reached, the image plane light field obtained from the last forward propagation calculation is defined as the reference light field, with the following formula:

[0081] ,

[0082] in As a reference light field, The amplitude of the reference light field for reconstruction. The phase of the reconstructed reference optical field;

[0083] The formula for calculating the complex light field of the image plane of an optical image is:

[0084] ,

[0085] in For at any time Complex light field of the image plane, For the reconstructed transient optical field amplitude, For the reconstructed transient optical field phase, For the k-th time;

[0086] The iterative phase retrieval algorithm (Gerchberg-Saxton algorithm) that propagates the angular spectrum of an optical image follows the same procedure as the reference light field;

[0087] The relative phase difference between the complex optical field on the image plane and the reference optical field is calculated using complex conjugate multiplication. The phase angle of the relative phase difference is extracted to obtain the wrapped relative phase difference field. The wrapped relative phase difference field is then subjected to phase unwrapping processing (based on the method of solving the Poisson equation using FFT) to eliminate the phase map. A jump is achieved, resulting in a continuous absolute phase difference field;

[0088] Based on the fundamental formulas of laser interferometry, the absolute phase difference is converted into a normal micro-displacement of the component surface, as given by the following formula:

[0089] ,

[0090] in For at any time The component surface normal micro-displacement, The wavelength of the illumination laser in a vacuum, For at any time A continuous absolute phase difference field;

[0091] Using spatiotemporal tensor reconstruction, the normal micro-displacements are arranged according to their spatial coordinates in the image and the order of acquisition time to form a three-dimensional data array, which is defined as the spatiotemporal micro-displacement field.

[0092] The Gerchberg-Saxton algorithm effectively recovers phase information from complex optical images, greatly improving the ability to perceive minute surface changes. Particularly in detecting minute vibrations and deformations on aluminum alloy surfaces, it provides more accurate displacement information. This algorithm reconstructs the dynamic light field through a continuous iterative process, reflecting the optical changes of aluminum alloy components under excitation in real time. By using angular spectrum propagation and complex amplitude calculation, this method overcomes the limitations of traditional optical image analysis, accurately capturing minute optical differences in different measurement environments and enhancing sensitivity to subtle defects on aluminum alloy surfaces. It calculates the relative phase difference using complex conjugate multiplication, enabling... It can effectively extract information about surface micro-displacements, making the quality defects of aluminum alloy components more clearly apparent. By using the formula of laser interferometry, the phase difference is converted into micro-displacement, which can accurately quantify the minute displacements on the surface of aluminum alloy components. In particular, it has extremely high detection accuracy for displacement changes caused by minute defects. Through the reconstruction of the spatiotemporal micro-displacement field, the micro-displacement information of the component surface can be generated into a dynamic response overview map in terms of time dimension, which greatly enhances the spatiotemporal analysis capability of component surface defects. The spatiotemporal micro-displacement field combines spatial coordinates and time factors, providing multi-dimensional data for further modal analysis and defect localization, thus enhancing the comprehensiveness and accuracy of detection.

[0093] S2. Map the vibration velocity sequence to the pseudo-state space to form the phase space trajectory matrix, and perform singular value decomposition. Extract the left singular vector corresponding to the largest singular value in the decomposition result and define it as the dynamic reference value. Reshape the spatiotemporal micro-displacement field into a data matrix, calculate the fluctuation data matrix and the spatiotemporal correlation matrix, perform eigenvalue decomposition on the spatiotemporal correlation matrix, solve for the time coefficient vector, and calculate the vector of the POD spatial mode by projecting the fluctuation data matrix onto the time coefficient vector to obtain the spatial mode.

[0094] Specifically, the vibration velocity sequence is mapped to a pseudo-state space to form a phase space trajectory matrix, and singular value decomposition is performed. The left singular vector corresponding to the largest singular value in the decomposition result is extracted and defined as the dynamic reference value, including:

[0095] The mean vibration velocity is calculated and then subtracted from the mean to obtain the detrended (zero-mean) vibration velocity. The velocity is then sorted in chronological order to generate a detrended vibration velocity sequence, thereby eliminating the DC component in the signal and ensuring that subsequent analysis fluctuates around zero.

[0096] Zero-phase filtering is applied to the detrended vibration velocity sequence. The purpose of filtering is to retain the effective structural vibration components excited by the excitation, while suppressing low-frequency drift, power frequency interference and high-frequency noise that exceeds the analysis range, thus obtaining the filtered vibration velocity sequence.

[0097] Using the mutual information method and the pseudo nearest neighbor method, respectively, with a time delay and an embedding dimension m, the filtered vibration velocity sequence is mapped to an m-dimensional pseudo-state space to obtain a phase space vector.

[0098] Arrange the phase space vectors in time order to form a phase space trajectory matrix, where each row represents the m-dimensional state at time t. Perform singular value decomposition on the phase space trajectory matrix to obtain an orthogonal matrix (containing the left singular vector and the right singular vector respectively) and a diagonal matrix. The diagonal elements are called singular values, which represent the variance of the data in each principal direction.

[0099] Extract the left singular vector corresponding to the largest singular value in the decomposition result and define it as the dynamic reference value;

[0100] The spatiotemporal micro-displacement field is aligned with the time frame of the dynamic reference value by linear interpolation to obtain the final dynamic reference value.

[0101] Detrended (zero-mean) vibration data removes the DC component, ensuring that subsequent analysis can focus on the dynamic changes of the system, avoiding interference from static errors and improving the accuracy of the analysis. Zero-phase filtering can effectively suppress low-frequency drift and power frequency interference, ensuring that the extracted vibration data is more representative of the true response of the system, rather than the influence of noise and environmental interference, thus improving the quality of the signal. The left singular vector corresponding to the maximum singular value extracted by singular value decomposition can effectively capture the most representative dynamic features of the system, serving as the core basis for subsequent analysis and reducing the influence of redundant data.

[0102] Furthermore, the spatiotemporal micro-displacement field is reshaped into a data matrix, the fluctuation data matrix and the spatiotemporal correlation matrix are calculated, eigenvalue decomposition is performed on the spatiotemporal correlation matrix, and the time coefficient vector is solved. By projecting the fluctuation data matrix onto the time coefficient vector, the vector of the POD spatial modes is calculated to obtain the spatial modes, including:

[0103] The spatiotemporal micro-displacement field is reshaped into a data matrix, the spatial dimension is flattened, each row of the matrix represents a spatial location, the displacement sequence over the entire time history, and each column represents a snapshot of the full field displacement at a certain moment.

[0104] Calculate the mean of each row of the data matrix to obtain the average displacement field, and subtract the average displacement field from each row of the data matrix to obtain the fluctuation data matrix. Calculate the spatiotemporal correlation matrix of the fluctuation field using the following formula:

[0105] ,

[0106] Where R is the spatiotemporal correlation matrix, and M is the total number of spatial points (pixels). Here is the fluctuation displacement data matrix, and T is the transpose operation;

[0107] Eigenvalue decomposition is performed on the spatiotemporal correlation matrix to obtain the eigenvalues ​​of the POD mode and its corresponding time coefficient vector (which has been automatically normalized). The time coefficient vector is defined as the time coefficient of the POD mode.

[0108] POD mode refers to a set of optimal spatiotemporal basis functions extracted from full-field transient micro-displacement data through intrinsic orthogonal decomposition (POD). Each mode consists of a spatial mode shape and a time coefficient sequence. It does not represent the physical intrinsic mode of the structure, but is a data-driven, energy-optimal decomposition unit used to efficiently characterize and screen vibration response features related to defects.

[0109] The vector of POD spatial modes is calculated by projecting the fluctuation data matrix onto the time coefficient vector, using the following formula:

[0110] ,

[0111] in Let be the vector of the nth order POD spatial mode. For the eigenvalues ​​of the nth POD mode, This is the time coefficient vector for the nth POD mode;

[0112] The vectors of the POD spatial modes are reshaped to obtain the spatial modes;

[0113] Its element order corresponds to the sequence obtained by expanding the original two-dimensional image in column-major (e.g., MATLAB, Fortran) or row-major (e.g., C, Python NumPy).

[0114] By reshaping the spatiotemporal micro-displacement field into a data matrix, spatial and temporal information can be processed within a unified framework, simplifying the data processing and analysis process. Calculating the fluctuation data matrix and performing eigenvalue decomposition of the spatiotemporal correlation matrix can effectively capture high-frequency information and important modes in the system's dynamic response, reducing the limitations of traditional analysis methods. POD extracts the optimal spatiotemporal basis function from the full-field transient micro-displacement data, making the characterization of vibration modes more accurate. It can mine the main vibration characteristics of the system based on actual data without directly relying on the physical model, providing more accurate vibration response characteristics for defect detection. After data reshaping, the spatial mode vectors of POD modes can more intuitively represent the key dynamic characteristics in the structural response. The POD spatial mode vectors represent the optimal spatial distribution of vibration characteristics, which can efficiently capture vibration responses related to defects, helping to improve the sensitivity and accuracy of defect identification.

[0115] S3. Divide the time coefficient and dynamic reference value ranges into interval grids respectively, calculate the maximum normalized empirical mutual information, define it as the maximum information coefficient, rearrange the POD mode indexes from largest to smallest according to the maximum information coefficient, generate the defect sensitive mode set, calculate the residual field of the sensitive mode, perform binarization processing, generate the defect binary mask, and calculate the defect location.

[0116] Specifically, the value ranges of the time coefficient and the dynamic reference value are divided into interval grids, the maximum normalized empirical mutual information is calculated and defined as the maximum information coefficient, and the POD mode indexes are rearranged in descending order of the maximum information coefficient to generate a defect-sensitive mode set, including:

[0117] The time coefficient and the final dynamic reference value are divided into interval grids, and the empirical mutual information under the grid division is calculated using the following formula:

[0118] ,

[0119] in The empirical mutual information is defined by grid G, where g and h are the number of segments dividing the time coefficient and the final dynamic reference value range, respectively. For data points to fall into the grid The joint empirical probability of the (u,v)th cell in the array. and Let be the marginal empirical probability of a data point falling into the u-th segment in the time coefficient dimension and the marginal empirical probability of falling into the v-th segment in the final dynamic reference value dimension;

[0120] The maximum normalized empirical mutual information, defined as the maximum information coefficient, is calculated using the following formula:

[0121] ,

[0122] in G is the maximum information coefficient between the time coefficient of the nth-order POD mode and the final dynamic reference value. G is the grid, and B is the upper limit of the total number of allowed grids, which is set using empirical rules to control computational complexity.

[0123] The POD modal indexes are rearranged from largest to smallest according to the maximum information coefficient to obtain a new index sequence. The number of sensitive modes is set using the fixed number method, and the corresponding mode pairs are extracted from the POD modal set to form a defect sensitive mode set.

[0124] By calculating empirical mutual information within the range of time coefficients and dynamic reference values, information from multiple dimensions can be effectively integrated, revealing the potential correlations between them and enhancing the understanding of the system's dynamic characteristics. The calculation of empirical mutual information can not only capture linear relationships but also reveal the nonlinear dependence between time coefficients and dynamic reference values. By normalizing the maximum information coefficient and rearranging the POD modes, the accuracy of mode selection during defect detection can be improved, thereby effectively extracting the modes that best reflect the defect characteristics.

[0125] Furthermore, the residual field of the sensitive mode is calculated and binarized to generate a defect binary mask, and the defect location is calculated, including:

[0126] Based on each sensitive mode in the defect-sensitive mode set, the residual field of the sensitive mode is calculated using the following formula:

[0127] ,

[0128] in Let the residual field be the k-th sensitive mode. This represents the k-th sensitive spatial mode shape of the component under test. For the mean mode shape of the k-th sensitive mode, a healthy baseline standard deviation mode shape is selected and generated from historical data. The health baseline standard deviation mode shape for the k-th sensitive mode characterizes the natural fluctuation range of the healthy sample in this mode;

[0129] For each pixel location in the residual field, take the maximum value of all residual field values ​​at that location as the composite residual field;

[0130] The optimal segmentation threshold is set using the Otsu method. The composite residual field is binarized (marking the composite residual field as 1 if it is greater than or equal to the optimal segmentation threshold, and otherwise as 0) to generate an initial defect binary mask. A value of 1 represents a suspected defect pixel.

[0131] A morphological opening operation (erosion followed by dilation) is performed on the initial defect binary mask to eliminate isolated small points caused by noise and smooth the boundaries of the defect region.

[0132] Connected component labeling is performed on the morphologically processed binary image to identify all connected foreground regions (i.e., connected blocks composed of pixels with a value of 1). The centroid coordinates of each connected region are calculated and defined as the spatial location region of the defect.

[0133] The calculation of residual fields can directly characterize the deviations caused by defects, which is of great significance for defect detection. By comparing the residuals with the standard deviation mode shape, it is possible to keenly capture the minute changes caused by defects, thereby improving the sensitivity and accuracy of detection. The combination of Otsu's method and morphological opening operation can effectively remove the interference caused by noise, accurately mark the defect area, reduce misjudgments, and improve the accuracy of detection results. Through binarization and connected component marking, the specific location of defects can be automatically identified and located, eliminating the need for manual intervention and greatly improving detection efficiency. Through centroid coordinate calculation, the location of defects can be accurately located, providing necessary information for the repair or replacement of components and ensuring structural safety. The connected component marking method can effectively identify various types of defects, has high flexibility and adaptability, and can cope with complex defect situations in practical applications.

[0134] Example 2, refer to Figure 2 As a second embodiment of the present invention, an aluminum alloy component quality inspection system includes:

[0135] The data collection and processing module is used to collect multi-source data of aluminum alloy components and perform noise reduction and standardization processing.

[0136] The displacement recovery module is used to define a reference intensity image, perform an iterative phase recovery algorithm on the reference intensity image, generate a reference light field, calculate the complex light field of the image plane of the optical image, calculate the relative phase difference between the complex light field of the image plane and the reference light field, convert it into the normal micro-displacement of the component surface, and generate a spatiotemporal micro-displacement field.

[0137] The decomposition mode module is used to map the vibration velocity sequence to the pseudo-state space, construct the phase space trajectory matrix, and perform singular value decomposition. It extracts the left singular vector corresponding to the maximum singular value in the decomposition result and defines it as the dynamic reference value. It reshapes the spatiotemporal micro-displacement field into a data matrix, calculates the fluctuation data matrix and the spatiotemporal correlation matrix, performs eigenvalue decomposition on the spatiotemporal correlation matrix, solves the time coefficient vector, and calculates the vector of the POD spatial mode by projecting the fluctuation data matrix onto the time coefficient vector to obtain the spatial mode.

[0138] The defect detection module is used to divide the value ranges of the time coefficient and the dynamic reference value into interval grids, calculate the maximum normalized empirical mutual information, which is defined as the maximum information coefficient, rearrange the POD mode indexes in descending order of the maximum information coefficient, generate a defect-sensitive mode set, calculate the residual field of the sensitive mode, perform binarization processing, generate a defect binary mask, and calculate the defect location.

[0139] This embodiment also provides a computer device applicable to the aluminum alloy component quality inspection method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the aluminum alloy component quality inspection method proposed in the above embodiment.

[0140] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0141] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the aluminum alloy component quality inspection method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0142] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for quality inspection of aluminum alloy components, characterized in that: Includes the following steps: Collect and preprocess multi-source data, define a reference intensity image, execute an iterative phase retrieval algorithm on the reference intensity image to generate a reference light field, calculate the complex light field of the image plane of the optical image, calculate the relative phase difference between the complex light field of the image plane and the reference light field, convert it into the normal micro-displacement of the component surface, and generate a spatiotemporal micro-displacement field. The vibration velocity sequence is mapped to the pseudo-state space to form the phase space trajectory matrix, and singular value decomposition is performed. The left singular vector corresponding to the largest singular value in the decomposition result is extracted and defined as the dynamic reference value. The spatiotemporal micro-displacement field is reshaped into a data matrix, and the fluctuation data matrix and spatiotemporal correlation matrix are calculated. The spatiotemporal correlation matrix is ​​decomposed by eigenvalue decomposition to solve the time coefficient vector. By projecting the fluctuation data matrix onto the time coefficient vector, the vector of the POD spatial mode is calculated to obtain the spatial mode. The time coefficient and dynamic reference value ranges are divided into interval grids, the maximum normalized empirical mutual information is calculated and defined as the maximum information coefficient, the POD mode index is rearranged from largest to smallest according to the maximum information coefficient, a defect sensitive mode set is generated, the residual field of the sensitive mode is calculated and binarized to generate a defect binary mask, and the defect location is calculated. The multi-source data includes optical images and vibration velocities.

2. The method for quality inspection of aluminum alloy components as described in claim 1, characterized in that: The calculation of the complex light field of the image plane of the optical image, the calculation of the relative phase difference between the complex light field of the image plane and the optical field, the conversion into the normal micro-displacement of the component surface, and the generation of the spatiotemporal micro-displacement field include: An optical image of the component when it is completely stationary is acquired and defined as a reference intensity image; An iterative phase retrieval algorithm is performed on the reference intensity image. The square root of the reference intensity image is set as the initial image plane light field on the object surface. The initial image plane light field is propagated to the camera target surface through the angular spectrum propagation formula, and the complex amplitude of the image plane light field is calculated. Replace the amplitude of the image plane light field complex amplitude with the square root of the reference intensity image to generate the updated image plane light field complex amplitude; The updated image plane light field is backpropagated to the object plane to obtain the object plane light field estimate for the next iteration. A maximum number of iterations is set. When the maximum number of iterations is reached, the image plane light field calculated by the last forward propagation is defined as the reference light field. Calculate the complex light field of the image plane of an optical image; The relative phase difference between the complex optical field on the image plane and the reference optical field is calculated by complex conjugate multiplication. The phase angle of the relative phase difference is extracted to obtain the wrapped relative phase difference field. The wrapped relative phase difference field is then subjected to phase unwrapping processing to obtain the continuous absolute phase difference field. Based on the fundamental formulas of laser interferometry, the absolute phase difference is converted into a normal micro-displacement on the surface of the component; Using spatiotemporal tensor reconstruction, the normal micro-displacements are arranged according to their spatial coordinates in the image and the order of acquisition time to form a three-dimensional data array, which is defined as the spatiotemporal micro-displacement field.

3. The method for quality inspection of aluminum alloy components as described in claim 2, characterized in that: The process of mapping the vibration velocity sequence to a pseudo-state space to form a phase space trajectory matrix, performing singular value decomposition, and extracting the left singular vector corresponding to the largest singular value in the decomposition result, which is defined as the dynamic reference value, includes: Calculate the mean of the vibration velocity and subtract the mean from the vibration velocity to obtain the detrended vibration velocity. Sort the detrended vibration velocity sequence according to time order to generate the detrended vibration velocity sequence. Apply zero-phase filtering to the detrended vibration velocity sequence to obtain the filtered vibration velocity sequence. By setting a time delay and an embedding dimension m, the filtered vibration velocity sequence is mapped to an m-dimensional pseudo-state space to obtain a phase space vector. Arrange the phase space vectors in time order to form a phase space trajectory matrix. Perform singular value decomposition on the phase space trajectory matrix to obtain an orthogonal matrix and a diagonal matrix. The diagonal elements are called singular values. Extract the left singular vector corresponding to the largest singular value in the decomposition result and define it as the dynamic reference value; The spatiotemporal micro-displacement field is aligned with the time frame of the dynamic reference value by linear interpolation to obtain the final dynamic reference value.

4. The method for quality inspection of aluminum alloy components as described in claim 3, characterized in that: The process involves reshaping the spatiotemporal micro-displacement field into a data matrix, calculating the fluctuation data matrix and the spatiotemporal correlation matrix, performing eigenvalue decomposition on the spatiotemporal correlation matrix to solve for the time coefficient vector, and calculating the vector of the POD spatial modes by projecting the fluctuation data matrix onto the time coefficient vector to obtain the spatial modes, including: The spatiotemporal micro-displacement field is reshaped into a data matrix. The mean of each row of the data matrix is ​​calculated to obtain the average displacement field. The average displacement field is then subtracted from each row of the data matrix to obtain the fluctuation data matrix. The spatiotemporal correlation matrix of the fluctuation field is then calculated. Eigenvalue decomposition is performed on the spatiotemporal correlation matrix to obtain the eigenvalues ​​of the POD mode and its corresponding time coefficient vector. The time coefficient vector is defined as the time coefficient of the POD mode. By projecting the fluctuation data matrix onto the time coefficient vector, the vector of the POD spatial mode is calculated, and the vector of the POD spatial mode is reshaped to obtain the spatial mode.

5. The method for quality inspection of aluminum alloy components as described in claim 4, characterized in that: The process involves dividing the time coefficient and dynamic reference value ranges into interval grids, calculating the maximum normalized empirical mutual information (defined as the maximum information coefficient), and rearranging the POD mode indexes from largest to smallest according to the maximum information coefficient to generate a defect-sensitive mode set, including: The ranges of time coefficient and final dynamic reference value are divided into interval grids, and the empirical mutual information under the grid division is calculated. Calculate the maximum normalized empirical mutual information, which is defined as the maximum information coefficient; The POD modal indexes are rearranged from largest to smallest according to the maximum information coefficient to obtain a new index sequence. The number of sensitive modes is set using the fixed number method, and the corresponding mode pairs are extracted from the POD modal set to form a defect sensitive mode set.

6. The method for quality inspection of aluminum alloy components as described in claim 5, characterized in that: The calculation of the residual field of the sensitive mode, the binarization process to generate a defect binary mask, and the calculation of the defect location include: Based on each sensitive mode in the defect sensitive mode set, the residual field of the sensitive mode is calculated. For each pixel position in the residual field, the maximum value of all residual field values ​​at that position is taken as the comprehensive residual field. The optimal segmentation threshold is set using the Otsu method, the comprehensive residual field is binarized to generate an initial defect binary mask, and morphological opening operation is performed on the initial defect binary mask. Connected component labeling is performed on the morphologically processed binary image to identify all connected foreground regions. The centroid coordinates of each connected region are calculated and defined as the spatial location region of the defect.

7. The method for quality inspection of aluminum alloy components as described in claim 1, characterized in that: The collection and preprocessing of multi-source data includes: Multi-source data of aluminum alloy components are collected through smart sensors and then denoised and standardized. The intelligent sensor includes a high-speed camera and a laser vibrometer.

8. A quality inspection system for aluminum alloy components, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The data collection and processing module is used to collect multi-source data of aluminum alloy components and perform noise reduction and standardization processing. The displacement recovery module is used to define a reference intensity image, perform an iterative phase recovery algorithm on the reference intensity image, generate a reference light field, calculate the complex light field of the image plane of the optical image, calculate the relative phase difference between the complex light field of the image plane and the reference light field, convert it into the normal micro-displacement of the component surface, and generate a spatiotemporal micro-displacement field. The decomposition mode module is used to map the vibration velocity sequence to the pseudo-state space, construct the phase space trajectory matrix, and perform singular value decomposition. It extracts the left singular vector corresponding to the maximum singular value in the decomposition result and defines it as the dynamic reference value. It reshapes the spatiotemporal micro-displacement field into a data matrix, calculates the fluctuation data matrix and the spatiotemporal correlation matrix, performs eigenvalue decomposition on the spatiotemporal correlation matrix, solves the time coefficient vector, and calculates the vector of the POD spatial mode by projecting the fluctuation data matrix onto the time coefficient vector to obtain the spatial mode. The defect detection module is used to divide the value ranges of the time coefficient and the dynamic reference value into interval grids, calculate the maximum normalized empirical mutual information, which is defined as the maximum information coefficient, rearrange the POD mode indexes in descending order of the maximum information coefficient, generate a defect-sensitive mode set, calculate the residual field of the sensitive mode, perform binarization processing, generate a defect binary mask, and calculate the defect location.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the aluminum alloy component quality inspection method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the aluminum alloy component quality inspection method according to any one of claims 1 to 7.

Images