Fatigue life prediction method for rocket first stage shell based on nondestructive testing

Abstract

The application discloses a rocket one-substage rocket body shell fatigue life prediction method based on nondestructive testing, relates to the technical field of nondestructive testing, and comprises the following steps: using a geometric reference file, performing multi-mode nondestructive testing on the external structure of a rocket one-substage rocket body shell to generate an original detection data set; extracting size parameters, morphology parameters and position parameters of defects from the original detection data set, integrating the parameters and generating an initial defect list; mapping the initial defect list to a digital modeling structure, establishing an initial state of the digital modeling structure, calculating local stress and generating a transfer function; constructing a fatigue load spectrum according to the transfer function, calculating the stress response of the rocket one-substage rocket body shell through finite element simulation, synchronously integrating strain measurement signals, acceleration measurement signals and acoustic emission monitoring signals, and forming time sequence observation data. The application is suitable for comprehensive flaw detection and life evaluation of a reusable one-substage rocket body shell after each recovery.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology, and in particular to a method for predicting the fatigue life of the first-stage rocket body shell based on nondestructive testing. Background Technology

[0002] With the continuous evolution of aerospace engine propulsion technology, reliability assessment of the cryogenic propellant-delivered first-stage rocket casing under complex operating conditions has become a key direction in structural integrity research. Existing engineering practices widely employ multi-mode non-destructive testing methods, such as ultrasonic testing, eddy current testing, X-ray testing, and acoustic emission monitoring, to identify and quantify defects on the external surface and near-surface region of the first-stage rocket casing. Simultaneously, by combining methods such as three-dimensional geometric reconstruction, point cloud registration, and finite element analysis, stress distribution and structural weak areas under different defect morphologies can be inferred.

[0003] In the unique structural scenario of the first-stage rocket casing using cryogenic propellants, existing technologies still have limitations in their ability to correlate "detection data—damage characteristics—fatigue life," particularly in reflecting the dynamic characteristics of defect size and morphology evolving with load, which still relies heavily on single-moment detection results. Because materials in cryogenic environments exhibit strongly coupled temperature sensitivity, load sensitivity, and micro-damage sensitivity, traditional life prediction methods based on static detection or offline stress analysis are insufficient to efficiently characterize the temporal evolution of defects under real service conditions. Summary of the Invention

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

[0005] Therefore, this invention provides a method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing, which solves the problem of insufficient dynamic mapping of defect parameters to life prediction.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for predicting the fatigue life of a rocket first-stage body shell based on non-destructive testing. The method includes: performing a three-dimensional geometric scan of the scannable external structure of the rocket first-stage body shell to generate a geometric reference file; using the geometric reference file, performing multi-mode non-destructive testing on the external structure of the rocket first-stage body shell to generate an original test dataset; extracting and integrating the size, shape, and location parameters of defects from the original test dataset to generate an initial defect list; mapping the initial defect list to a digitally modeled structure to establish the initial state of the digitally modeled structure, calculating local stress, and generating a transfer function; constructing a fatigue load spectrum based on the transfer function, and calculating the stress response of the rocket first-stage body shell through finite element simulation, simultaneously integrating strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals to form time-series observation data; assimilating the parameters and damage state of the time-series observation data to generate a post-correction parameter set, and calculating crack propagation and remaining life distribution.

[0007] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for generating the geometric reference file are as follows: Using a high-precision laser scanner, three-dimensional point cloud data was collected from the surface of the external structure of the first stage rocket body shell to generate raw point cloud data. The original point cloud data is denoised and filtered to generate a clean point cloud dataset, and the surface fitting method is used to reconstruct the three-dimensional geometric surface of the external structure. Features are extracted from the three-dimensional geometric surface of the external structure to generate geometric feature annotation data, which is then fused with the three-dimensional geometric surface of the external structure to generate a geometric reference file.

[0008] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for generating the original detection dataset are as follows: Extract the three-dimensional geometric information and feature positions of the first-stage rocket body shell from the geometric reference file to generate path planning data; Based on the path planning data, configure the scanning position, scanning angle, and scanning accuracy parameters for multi-mode non-destructive testing, and generate a non-destructive testing execution plan; The non-destructive testing implementation scheme is used to perform multi-mode scanning on the external structure of the rocket's first-stage shell, collect signal data in real time, and form the original test signal stream. The original detection signal stream is initially synchronized and labeled in time, and each signal is associated with the corresponding geometric position of the external structure to generate the original detection dataset.

[0009] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for generating the initial defect list are as follows: The original detection dataset is processed by time synchronization, denoising enhancement and spatial registration to generate a multi-mode detection set; The system identifies suspected defect regions from a multi-mode detection set, generates defect region segmentation results, and analyzes and calculates the size, shape, and spatial location parameters of the defects to generate candidate defect data. The candidate defect data is fused using multiple modes and its quality is checked to identify duplicates and conflicts, and then integrated to generate an initial defect list.

[0010] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for establishing the initial state of the digital modeling structure are as follows: Map the size, shape, and location parameters of defects in the initial defect list to a geometric reference file to generate defect annotation data; The size, shape, and location parameters of defects in the defect annotation data are summarized and processed to establish the initial state of the digital modeling structure.

[0011] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for generating the transfer function are as follows: The initial state of the digitally modeled structure is fused with the material property data from the geometric reference file to generate local stress analysis data; Based on the local stress analysis data, the local stress in the defect area under different vibration loads is calculated, and a local stress dataset is generated. Time-domain and frequency-domain analyses were performed on the local stress dataset to establish the mapping relationship between the defect stress response and the external vibration load. The results were then verified and optimized to generate the transfer function.

[0012] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for forming time-series observation data are as follows: The transfer function is coupled with the local stress dataset to generate a service load reconstruction scheme; According to the service load reconstruction scheme, the flight phase load, separation load, landing and recovery impact load and ground hoisting load are input into the digital modeling structure, the load parameters and timestamps are recorded, and the loading execution record is generated. Strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals from the flight measurement records are collected and time-synchronized and labeled in conjunction with the comprehensive flaw detection results after recovery, generating preliminary time-series observation data; Noise filtering, signal enhancement, and spatial location correlation are performed on the preliminary time-series observation data to form time-series observation data.

[0013] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for performing noise filtering, signal enhancement, and spatial location correlation on the preliminary time-series observation data to form time-series observation data are as follows. The initial time-series observation data are subjected to amplitude enhancement, spectrum enhancement, and filtering and denoising to generate enhanced and denoised signal data. The enhanced and denoised signal data is spatially matched with the geometric reference file, and outlier samples are removed to generate verified spatially labeled signal data. The spatially labeled signal data after verification are organized into a unified format to form time-series observation data.

[0014] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for generating the post-correction parameter set are as follows: The time-series observation data is aligned with the loaded execution records in time and space to generate synchronization calibration signal data; The synchronous calibration signal data is mapped and compared with the size parameters, shape parameters and spatial location parameters of the defects in the initial state of the digital modeling structure, and the defect response characteristics are adjusted to generate corrected defect response data. The parameters and damage state of the corrected defect response data are assimilated to generate a set of post-correction parameters.

[0015] As a preferred embodiment of the rocket first-stage body shell fatigue life prediction method based on non-destructive testing described in this invention, the specific steps for calculating crack propagation and remaining life distribution are as follows: Based on the post-correction parameter set, calculate the crack growth curve of the defect under different loads and generate crack propagation trajectory data. Based on crack propagation trajectory data and structural failure criteria, the remaining life of defects is assessed, and crack propagation and remaining life distributions are generated.

[0016] The beneficial effects of this invention are as follows: By mapping the initial defect list to the digital modeling structure, the actual defects are accurately restored in terms of size, shape, and spatial location, so that the digital modeling structure has the same initial defect state as the actual object, avoiding the deviation caused by the reliance on idealized defects in traditional life assessment; and by integrating material property data to establish the transfer function between defect stress response and external vibration load, complex loads can be accurately converted into the local stress field of the defect, providing a quantitative basis for fatigue loading and damage calculation, thereby ultimately achieving high-precision fatigue life prediction based on the characteristics of the actual structure and actual defects, and improving the credibility of life prediction results. Attached Figure Description

[0017] 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.

[0018] Figure 1 The flowchart shows a method for predicting the fatigue life of the first-stage rocket body shell based on non-destructive testing. Figure 2 Flowchart generated for geometric reference file; Figure 3 A flowchart for the generation of time-series observation data; Figure 4 A flowchart for calculating crack propagation and remaining lifetime distribution. Detailed Implementation

[0019] 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.

[0020] 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.

[0021] 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.

[0022] Reference Figures 1-4As one embodiment of the present invention, this embodiment provides a method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing, comprising the following steps: S1. Perform a three-dimensional geometric scan of the external structure of the rocket's first-stage body shell to generate a geometric reference file.

[0023] S1.1. Using a high-precision laser scanner, three-dimensional point cloud data is collected on the surface of the external structure of the first stage rocket body to generate raw point cloud data.

[0024] Specifically, a high-precision laser scanner is positioned in the visible area outside the first-stage rocket body shell. By adjusting the scanner's mounting attitude, scanning spacing, and focal length, the laser beam is designed to completely cover the external structural surface of the first-stage rocket body shell. The scanner then scans the external structural surface segment by segment according to a pre-defined scanning trajectory. During the scanning process, the distance information and corresponding angle parameters of each laser pulse are recorded, forming a set of points stored in coordinate form. After the high-precision laser scanner completes a full-coverage scan of the external structural surface of the first-stage rocket body shell, the point sets obtained from all scanned locations are arranged and stitched together according to time sequence and spatial location to generate raw point cloud data corresponding to the external structural surface of the first-stage rocket body shell.

[0025] It should also be noted that the specific steps for setting the scanning trajectory are as follows: determine the scanning area boundary based on the spatial shape of the first-stage rocket body shell; divide the local scanning zone according to the curvature and bending sections of the first-stage rocket body shell, and determine the start and end coordinates of each scanning zone; use a spatial uniform sampling method to calculate the scanning line spacing based on the resolution of the high-precision laser scanner. For example, when the angular resolution of the high-precision laser scanner is 0.02 degrees, convert the angular resolution to a line spacing of two millimeters, and arrange them in order from the start point to the end point to generate continuous scanning lines; combine all scanning lines into a complete scanning trajectory to achieve full coverage scanning of the external structural surface of the first-stage rocket body shell. For example, a high-precision laser scanner has a ranging accuracy of one millimeter, an angular resolution of 0.02 degrees, a scanning spacing of two millimeters, and a focal length adjustment range of 0.5 meters to 3 meters.

[0026] It should also be noted that the accuracy of the sampling values ​​is mainly based on the design requirements of the rocket's first-stage shell, sensor resolution, and actual environmental factors. Strict accuracy requirements are typically set based on the specifications of high-precision laser scanners, strain gauges, and accelerometers. For the rocket's first-stage shell, especially under extreme temperature and high vibration conditions, accuracy requirements may be no less than 1 mm. Whether the accuracy is appropriate needs to be confirmed in conjunction with the sensor's capabilities and the measurement environment. Excessively high accuracy requirements may exceed existing measurement capabilities, leading to error accumulation. Accuracy standards are usually based on design standards and safety factors to ensure the reliability and safety of the structure.

[0027] S1.2. The original point cloud data is denoised and filtered to generate a clean point cloud dataset, and the surface fitting method is used to reconstruct the three-dimensional geometric surface of the external structure.

[0028] Specifically, a statistical outlier removal method is used to remove outliers from the original point cloud data, generating preliminary cleaned point cloud data; a Gaussian filtering method is used to smooth the preliminary cleaned point cloud data, generating smoothed point cloud data; a surface fitting method is used to fit the smoothed point cloud data into blocks according to local surface regions, generating locally fitted surface data; each locally fitted surface data is then stitched together to generate a clean point cloud dataset and reconstruct the three-dimensional geometric surface of the external structure.

[0029] It should also be noted that the statistical outlier removal method is to analyze the distance relationship between each point in the point cloud and its neighboring points, and remove outliers that deviate from the normal distribution range. The normal range is, for example, the distance between neighboring points is concentrated in the range of one to three millimeters and the distribution is continuous without sudden increases. Gaussian filtering is a smoothing method based on a Gaussian weighting function. It reduces local height variations and achieves noise smoothing by weighting each point in the point cloud and its neighboring points. The surface fitting method extracts a set of points for fitting from point cloud data, divides local surface patches according to spatial distribution, uses mathematical fitting algorithms to obtain the fitting parameters of each surface patch, generates continuous surface patches based on the fitting parameters, and splices all surface patches into a three-dimensional geometric surface according to spatial adjacency.

[0030] S1.3 Extract features from the three-dimensional geometric surface of the external structure, generate geometric feature annotation data, and fuse it with the three-dimensional geometric surface of the external structure to generate a geometric reference file.

[0031] Specifically, a curvature calculation method is used to calculate the principal curvature and Gaussian curvature of each mesh on the three-dimensional geometric surface of the external structure, generating curvature feature data; a boundary detection method is used to identify the connection points, branch points, and endpoint positions on the three-dimensional geometric surface of the external structure, generating boundary feature data; a cross-section analysis method is used to extract several cross-sections along the axial direction on the three-dimensional geometric surface of the external structure, and calculate the cross-section diameter, wall thickness, and center position, generating cross-section feature data; the curvature feature data, boundary feature data, and cross-section feature data are integrated to generate geometric feature annotation data; and the geometric feature annotation data is fused with the three-dimensional geometric surface of the external structure according to the spatial correspondence to generate a geometric reference file.

[0032] It should also be noted that the curvature calculation method is to calculate the second derivative of the local neighborhood of each grid on the three-dimensional geometric surface of the external structure to obtain the principal curvature and Gaussian curvature, which are used to describe the degree of bending and shape change of the surface locally. Boundary detection methods examine the spatial distribution relationship between endpoints, branch points, and connection points and neighboring points on the three-dimensional geometric surface of the external structure, analyze the number of neighboring points, the geometric continuity of adjacent points, and the magnitude of local surface normal vector changes, in order to determine the structural boundary features of the first-stage rocket body shell. The cross-sectional analysis method involves taking cross-sections at several locations along the axial direction of the first-stage rocket body shell. By calculating geometric parameters such as the cross-sectional diameter, wall thickness, and center position, the local cross-sectional characteristics of the first-stage rocket body shell are described, which is used to accurately characterize the size and morphological changes of the first-stage rocket body shell.

[0033] S2. Using geometric reference files, perform multi-mode non-destructive testing on the external structure of the rocket's first-stage shell to generate the original test dataset.

[0034] S2.1 Extract the three-dimensional geometric information and feature positions of the first-stage rocket body shell from the geometric reference file to generate path planning data.

[0035] Specifically, from the geometric reference file, three-dimensional geometric points are sampled along the axis of the first-stage rocket body shell. The spatial coordinates, diameter, wall thickness, and corresponding feature positions of each sampled point are extracted. Based on the extracted three-dimensional geometric points and feature positions, key path nodes are marked according to scanning requirements, such as inflection points at bends in the first-stage rocket body shell, connection points at branches, and flange or interface positions. The marked key path nodes are used to sort and connect paths, generating a continuous sequence of path points. An interpolation resampling method is used to uniformize the spacing of the path point sequence, for example, setting the spacing between adjacent path points to an example range of 1 mm to 5 mm, generating path point data that meets the scanning accuracy requirements. The path point data is integrated with the corresponding feature positions and the geometric information of the first-stage rocket body shell to generate path planning data.

[0036] It should also be noted that the interpolation resampling method uses linear interpolation or spline interpolation to generate new equally spaced points based on the known numerical relationships between path points, thereby achieving uniformity in the path point sequence.

[0037] S2.2 Based on the path planning data, configure the scanning position, scanning angle, and scanning accuracy parameters for multi-mode non-destructive testing, and generate a non-destructive testing execution plan.

[0038] Specifically, based on the path planning data, the three-dimensional coordinates and characteristic directions of each critical path node are determined. The ray casting method is used to calculate the normal of each detection position relative to the surface of the first-stage rocket body, generating scanning positions. For each scanning position, the scanning angle range and step angle are set according to the requirements of ultrasonic testing, eddy current testing, or magnetic particle testing methods. For example, ultrasonic testing sets a scanning angle every 5°, and eddy current testing sets a scanning position every 2mm. Based on parameters such as the diameter, bending radius, and surface roughness of the first-stage rocket body, the required scanning accuracy for each scanning position is calculated. For example, the ultrasonic sampling interval is 0.1mm, and the eddy current sampling interval is 0.05mm. All scanning positions, scanning angle ranges, and scanning accuracy parameters are sorted according to the detection order to generate a non-destructive testing execution plan.

[0039] It should also be noted that the ultrasonic testing method involves applying an ultrasonic probe to the surface of the first-stage rocket body shell, utilizing the propagation and reflection characteristics of longitudinal waves or shear waves in the material to detect the location, size, and morphology of internal defects and generate defect information. Eddy current detection method uses alternating current to generate induced eddy currents on the surface of a conductive first stage rocket shell. When the eddy currents encounter defects or material discontinuities, they will cause disturbances. The presence and location of defects can be identified by measuring changes in induced voltage or impedance. It is suitable for surface and near-surface defect detection. The magnetic particle testing method involves applying a magnetic field to the surface of the ferromagnetic first-stage rocket body shell, magnetizing the surface and dispersing fine magnetic particles on the surface. Defects or cracks will create magnetic leakage zones, causing the magnetic particles to accumulate at the defects. The accumulation of magnetic particles is observed visually or with optical instruments to determine surface and near-surface defects.

[0040] S2.3. Through the implementation of non-destructive testing, multi-mode scanning is performed on the external structure of the rocket's first-stage shell to collect signal data in real time and form the original test signal stream.

[0041] Specifically, according to the non-destructive testing execution plan, ultrasonic testing probes, eddy current testing probes, and magnetic particle testing probes are sequentially arranged on the external structure of the first-stage rocket body shell. The probes are moved according to the scanning position and scanning angle set by the path planning data. The ultrasonic testing probe is activated to record the echo signal, the eddy current testing probe is activated to record the induced voltage or impedance change signal, and the magnetic particle testing probe is activated to record the magnetic particle aggregation image signal. At the same time, a timestamp and spatial label are added to each acquired signal, and the signal data stream is saved in real time to form the original detection signal stream.

[0042] It should be noted that for large rocket body shell structural components, X-ray inspection, industrial CT inspection or digital radiographic inspection and other radiographic non-destructive testing methods should be given priority to improve the ability to identify internal defects and near-surface defects. For auxiliary verification of local surface defects, other mature detection methods can be combined for verification, but the present invention lies in the correlation between radiation detection results and digital modeling structure and life assessment process.

[0043] S2.4 Perform preliminary time synchronization and labeling on the original detection signal stream, associate each signal with the corresponding external structural geometric position, and generate the original detection dataset.

[0044] Specifically, a timestamp correction method is used to convert the acquisition time of each signal in the original detection signal stream into an absolute timestamp using a unified clock source, and then standardize it by resampling at a fixed time resolution. Based on the scanning path and scanning rate recorded in the non-destructive testing execution plan, the acquisition points corresponding to the signals are mapped to the three-dimensional geometric surface of the external structure, marking the geometric position of each signal. A sequence alignment method is used to initially synchronize the signals of different detection modes in the time dimension, generating a signal record in a unified format, and associating each synchronization signal with the corresponding three-dimensional geometric position of the external structure to generate the original detection dataset.

[0045] It should also be noted that the timestamp correction method refers to the standardization of the time information of the acquired signal data, so as to unify the time reference of different acquisition sources or different acquisition cycles, and ensure that the time point of each signal can accurately correspond to the actual scanning position or the time of event occurrence, thereby eliminating the deviation caused by the difference in time reference. Sequence alignment refers to comparing and adjusting signal sequences from different detection modes or different acquisition channels in the time dimension so that the signals correspond to the same physical state or sampling position at the same time point, thereby realizing synchronous processing and unified analysis of multi-mode signals. A unified clock source typically uses GPS time, atomic clocks, or master clocks as a standard reference, and corrects clock deviations through synchronization signals or network protocols to achieve globally consistent time calibration.

[0046] S3. Extract the size, shape, and location parameters of defects from the original detection dataset, integrate them, and generate an initial defect list.

[0047] S3.1 Perform time synchronization, denoising enhancement, and spatial registration on the original detection dataset to generate a multi-mode detection set.

[0048] Specifically, a timestamp correction method is used to unify the time reference of each mode signal in the original detection dataset to generate a time synchronization signal; a digital filtering method is used to denoise the time synchronization signal, and an amplitude enhancement method is used to enhance the signal to generate an enhanced and denoised signal; a spatial registration method is used to match the enhanced and denoised signal with the three-dimensional geometric surface of the external structure and remove abnormal samples to generate a spatial registration signal; the spatial registration signal is organized in a unified format to generate a multi-mode detection set.

[0049] It should also be noted that digital filtering methods use digital filters to process time synchronization signals in the frequency domain or time domain to remove noise components and high-frequency interference while preserving the characteristics of the target signal. Amplitude enhancement methods increase the amplitude of key features in a signal by adjusting the signal amplitude or applying nonlinear amplification methods, thereby improving the signal's distinguishability and analysis accuracy. The spatial registration method matches the positions of multi-mode signals with geometric reference files or surface features of the first-stage rocket body. Through coordinate transformation, point cloud matching, or feature point alignment, the signal data and geometric positions are made to correspond one-to-one, ensuring the spatial consistency of the data of each mode and providing an accurate spatial reference for subsequent defect identification and analysis.

[0050] S3.2. Identify suspected defect regions in the multi-mode detection set, generate defect region segmentation results, and analyze and calculate the size parameters, shape parameters and spatial location parameters of the defects to generate candidate defect data.

[0051] Specifically, a threshold segmentation method is used to analyze the signal intensity or feature indicators of the multi-mode detection set to identify regions that may contain defects and generate preliminary defect region contours. An edge detection method is then used to refine these preliminary defect region contours, generating defect region segmentation results. Based on the defect region segmentation results, the surface features of the defects are calculated, including the surface curvature, expressed as: ; in, Indicates the surface curvature of the defect. Indicates the height value of the defect surface. This represents the local plane coordinates of the defect surface in the X direction. This represents the Y-axis coordinate of the local plane on the defect surface. This represents the first-order partial derivative of the defect surface height in the X direction. This represents the first-order partial derivative of the defect surface height in the Y direction. This represents the second partial derivative of the defect surface height in the X direction. This represents the second partial derivative of the defect surface height in the Y direction. The mixed second-order partial derivative representing the height of the defect surface; The roughness of the defect morphology is calculated using the following expression: ; in, Indicates the roughness of the defect morphology. Indicates the number of sampling points. Indicates the first The height value of each sampling point on the surface of the defect. Indicates the average height of the defect surface. Indicates the index of the sampling point representing the height of the defect surface; The area of ​​the defect's geometric features is calculated using the following expression: ; in, Represents the area of ​​the geometric features of the defect. This indicates the number of local triangular segments in the division. Indicates the first Area of ​​each triangular segment Indicates the index of the triangle segment; By combining the geometric reference file and using the three-dimensional coordinate transformation method, the local coordinates of each defect point in the defect region segmentation result are transformed according to the three-dimensional rotation matrix and the three-dimensional translation vector to obtain the global spatial coordinate position of the defect in the geometric reference file. The global spatial coordinate position is then matched and integrated with the size parameters, shape parameters and geometric features of the defect to generate candidate defect data.

[0052] It should also be noted that the threshold segmentation method sets a threshold based on the signal strength or characteristic index distribution of the original detection signal. For example, areas with a signal amplitude higher than 0.6 or an acoustic emission event count exceeding 5 are classified as areas where defects may exist, while the remaining areas are classified as non-defect areas, thus achieving preliminary defect localization. Edge detection is a processing method that identifies region boundaries by analyzing the gradient or change characteristics of the original detection signal. It is used to refine the preliminary defect contour and generate the defect region boundary. The three-dimensional coordinate transformation method adjusts the directional relationship of three-dimensional coordinates by using a three-dimensional rotation matrix and moves the position of three-dimensional coordinates by using a three-dimensional translation vector, thereby converting local three-dimensional coordinates into global three-dimensional coordinates. For example, spatial position calibration can be completed by performing rotation operations on three-dimensional coordinate points and superimposing translation amounts.

[0053] S3.3 Perform multi-mode fusion and quality verification on the candidate defect data, identify duplicates and conflicts, and perform consistency integration to generate an initial defect list.

[0054] Specifically, candidate defect data are grouped according to their source type, including X-ray inspection, industrial CT inspection, digital radiography inspection, and 3D scan verification results. Spatial location parameters, size parameters, and morphological parameters are extracted from each group of candidate defect data. A distance threshold matching method is used to compare the spatial locations of candidate defect data from different sources, and candidate defect data with spatial location errors and size parameter differences within the allowable range are retained. A parameter conflict detection method is used to compare the differences in size and morphological parameters of candidate defect data, and conflicting parameters are processed using an interval constraint convergence method to achieve consistency. The processed candidate defect data are then integrated, sorted according to spatial location parameters, and an initial defect list is generated.

[0055] It should also be noted that the distance threshold matching method compares the spatial location spacing of candidate defect data recorded from different sources or at different times, and determines defects whose spacing does not exceed the maximum positioning error allowed by the measurement accuracy of the first stage rocket body shell as the same defect, so as to achieve duplicate defect identification and data integration. The maximum positioning error is determined based on the spatial resolution and scanning noise level of the high-precision laser scanner, for example, in the range of 0.5 mm to 1 mm. The parameter conflict detection method is a comparative analysis of quantifiable features such as size and shape parameters of candidate defect data. By comparing the parameter differences of candidate defect data from different sources at the same spatial location, a conflict is determined when the difference of a certain parameter exceeds the normal deviation range determined by the sensor resolution. The normal deviation range is based on the minimum resolvable size and repeatability error of nondestructive testing, for example, the difference in size parameters does not exceed 0.05 mm, and the difference in shape parameters does not exceed 0.5%. The interval constraint convergence method forms a parameter interval by setting upper and lower limits for the size, shape, and spatial location parameters of candidate defect data. Interval intersection operations are then performed on conflicting candidate defect data to converge the parameter values ​​to overlapping intervals that meet the common conditions. The upper and lower limits are set based on historical non-destructive testing accuracy, sensor resolution, and allowable tolerances in processing or measurement. The range of size parameters can be set to ±0.05 mm of the measured value, and the range of shape parameters can be set to ±0.5% of the measured value. For example, the intersection of intervals from different data sources can be "1.0 mm – 1.2 mm", which serves as the final parameter interval.

[0056] S4. Map the initial defect list to the digital modeling structure, establish the initial state of the digital modeling structure, calculate the local stress, and generate the transfer function.

[0057] S4.1 Map the size, shape, and location parameters of the defects in the initial defect list to the geometric reference file to generate defect annotation data.

[0058] Specifically, a three-dimensional coordinate transformation method is used to process the dimensional, morphological, and positional parameters in the initial defect list according to the origin, axial direction, and cross-sectional datum of the three-dimensional geometric surface coordinate system of the external structure, converting the positional parameters of each defect into three-dimensional coordinates in the geometric reference file. A parameter mapping method is used to attach the dimensional and morphological parameters of each defect to the corresponding three-dimensional coordinate positions according to their original values, and generate defect annotation symbols according to the normal direction of the first-stage rocket body shell surface. After spatially aligning all the defect annotation symbols after coordinate transformation and parameter attachment with the corresponding areas of the geometric reference file, the global coordinate positions of the defect annotation symbols are mapped point by point to the three-dimensional surface coordinates of the geometric reference file, and the defect annotation symbols are superimposed on the corresponding surface areas to generate defect annotation data.

[0059] It should be noted that, through digital modeling, the size, shape, and spatial location parameters of each defect in the initial defect list are mapped to the geometric reference file, and the defects are labeled to ensure the accurate presentation of defect information in the virtual environment.

[0060] S4.2 Summarize and process the size, shape and location parameters of defects in the defect annotation data to establish the initial state of the digital modeling structure.

[0061] Specifically, the size, shape, and location parameters of each defect in the defect annotation data are classified according to defect type (notch, crack, and hole). Spatial regions are divided based on the coordinate position of the defect on the three-dimensional geometric surface of the external structure. The parameters of defects of the same type and in the same spatial region are summarized to generate summarized defect location parameters. The summarized defect location parameters are standardized using a unified three-dimensional coordinate system, and the size and shape parameters are attached to the standardized location coordinates. The attached size, shape, and standardized location parameters are checked for consistency and formatted to generate a parameter set that can be directly used for the initialization of the digital modeling structure, thus establishing the initial state of the digital modeling structure.

[0062] It should also be noted that a unified three-dimensional coordinate system refers to establishing a common three-dimensional coordinate framework for all defects, geometric features, and spatial locations within the entire digital modeling structure, so that spatial data obtained from different sources or different scans can be located, compared, and superimposed under the same coordinates.

[0063] S4.3. The initial state of the digitally modeled structure is fused with the material property data of the geometric reference file to generate local stress analysis data.

[0064] Specifically, the spatial location, size parameters, and morphological parameters of each defect in the initial state of the digitally modeled structure are matched with the material property data in the geometric reference file. Based on a unified three-dimensional coordinate system, the elastic modulus, Poisson's ratio, density, and other properties of the material are assigned to the local region where the digital twin defect is located, forming the input parameter set required for local stress analysis. The finite element analysis method is used to calculate the initial stress response of each defect region under a specified load, generating local stress analysis data, including the stress distribution, principal stress direction, and stress gradient information of the defect region.

[0065] It should also be noted that material property data refers to numerical information describing the mechanical and physical properties of materials, including parameters such as elastic modulus, Poisson's ratio, density, yield strength, hardness, and coefficient of thermal expansion, which are used to characterize the response characteristics of materials under external forces, vibrations, temperature, or other environmental influences.

[0066] S4.4. Based on the local stress analysis data, calculate the local stress in the defect area under different vibration loads and generate a local stress dataset.

[0067] Specifically, based on the local stress analysis data, vibration load amplitude and frequency parameters are set for each defect region, for example, the amplitude is 10 MPa and the frequency is 50 Hz. The vibration load is applied to the geometric surface of the defect region, and the stress response at each measuring point is calculated using the finite element analysis method. The principal stress direction and stress gradient information are extracted to generate local stress data for each defect region under different vibration loads. The local stress data of all defect regions are summarized according to the vibration load amplitude and frequency parameters to form a local stress dataset, including stress amplitude, principal stress direction and stress gradient distribution.

[0068] It should also be noted that the specific steps for setting the vibration load amplitude and frequency parameters are as follows: Based on the design drawings and material strength requirements of the rocket's first-stage shell, determine the range of vibration load action, for example, an amplitude range of 0~20MPa and a frequency range of 0~200Hz, and combine this with the vibration characteristics of the rocket engine launch and ground testing environment; according to the accuracy requirements of local stress response, set the amplitude step size to 2MPa and the frequency step size to 10Hz, and generate an amplitude-frequency combination table; match the amplitude-frequency combination table for each defect area, apply the vibration load conditions corresponding to each amplitude and frequency combination in sequence, record the vibration load amplitude, frequency parameters, and corresponding defect area identification and spatial location, forming a vibration load parameter set for subsequent local stress calculation.

[0069] S4.5 Perform time-domain and frequency-domain analysis on the local stress dataset, establish the mapping relationship between defect stress response and external vibration load, and perform verification and optimization to generate the transfer function.

[0070] Specifically, a time-domain feature extraction method is used to perform time-domain analysis on the local stress dataset. By extracting the stress-time variation curves of each defect region under various vibration loads, the peak stress, mean stress, and standard deviation are calculated, and a stress response time series is established. The stress response time series is converted into a spectrum using the Fast Fourier Transform method, and the main frequency components and amplitude features are extracted. The time-domain and frequency-domain features of the defect region are compared with the vibration response segments corresponding to the amplitude and frequency parameters of the external vibration load according to the time series index. The load condition with the highest correlation coefficient is used as the matching result to generate a mapping relationship between the defect stress response and the external vibration load. An error analysis method is used to verify the mapping relationship. By adjusting the vibration amplitude or frequency combination, the fitting accuracy of the stress response is optimized, and a transfer function is generated, which includes the defect stress transmission characteristics, main response frequencies, and response amplitudes.

[0071] It should also be noted that the time-domain feature extraction method is a technique that describes dynamic characteristics by analyzing the changes of signals on the time axis. It mainly includes calculating the maximum value, minimum value, mean, standard deviation and variance of the signal, which are used to quantify the change law of stress in the defect area over time. The Fast Fourier Transform (FFT) method is a mathematical method that converts time-domain signals into frequency-domain signals. By performing a Fourier transform on time-domain stress data, the amplitude and phase information of each frequency component are obtained, which can be used to analyze the response characteristics of defect areas to stress at different vibration frequencies.

[0072] It should be noted that by mapping the dimensional parameters, morphological parameters and spatial position parameters obtained by nondestructive testing point by point to the digital modeling structure, the geometric shape and spatial distribution of the real defect are reconstructed in a high-fidelity manner in the digital environment. Material properties are integrated and the local stress of the defect is calculated. Furthermore, the transfer function from external vibration load to the local stress response of the defect is constructed, so that a quantitative physical mapping relationship is formed between the load input and the stress output of the defect.

[0073] S5. Construct a fatigue load spectrum based on the transfer function, and calculate the stress response of the rocket's first-stage shell through finite element simulation. Simultaneously integrate strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals to form time-series observation data.

[0074] S5.1 Couple the transfer function with the local stress dataset to generate a service load reconstruction scheme.

[0075] Specifically, based on the transfer function, the stress response time series of the defect area is transformed into the frequency domain. It is then compared and adjusted one-to-one with the amplitude and frequency parameters of the corresponding vibration load according to the frequency correspondence principle, ensuring that each frequency component is associated with the external load conditions. This yields the stress response amplitude curve of the defect area. Using a cyclic accumulation method, the amplitude of the stress response amplitude curve of the defect area is statistically analyzed and accumulated. The cumulative stress amplitude of each defect area at different vibration frequencies is calculated. By comparing the difference between the allowable stress amplitude corresponding to the target fatigue life and the current cumulative stress amplitude, the vibration load amplitude is reduced proportionally or the application time is increased to ensure that the adjusted stress level meets the target fatigue life requirements. This generates the accelerated fatigue load time series for each defect area. The accelerated fatigue load time series of all defect areas are then summarized to generate a service load reconstruction scheme. For example, a maximum load amplitude limit of 50kN, an application time interval of 0.1s, and a frequency step size of 1Hz can be used.

[0076] It should also be noted that the target fatigue life, under the predetermined service conditions of the rocket's first-stage body shell (including typical vibration intensity, temperature range, and pressure fluctuation combinations), refers to the number of cyclic loads the structure must meet without fatigue crack propagation or failure. For example, it can be set as follows: The next cycle is used as a criterion for fatigue assessment and load adjustment. The cyclic accumulation method is a calculation method used for fatigue analysis. It estimates the degree of fatigue damage by accumulating the stress amplitude of a material or structure under different cyclic loads.

[0077] S5.2. Based on the service load reconstruction scheme, input the flight phase load, separation load, landing and recovery impact load, and ground hoisting load into the digital modeling structure, record the load parameters and timestamps, and generate the load execution record.

[0078] Specifically, the loading sequence, application location, application direction, duration, and load amplitude of flight phase loads, separation loads, landing and recovery impact loads, and ground hoisting loads are extracted from the service load reconfiguration scheme. This loading sequence, application location, application direction, duration, and load amplitude are then written into the digital modeling structure to form the load input configuration. According to the load input configuration, the flight phase loads, separation loads, landing and recovery impact loads, and ground hoisting loads are sequentially input into the digital modeling structure. During the input process, the load type, load amplitude, application location, application direction, and duration of each load are simultaneously recorded to form a load parameter set. Based on the start time, end time, and order of input for the flight phase loads, separation loads, landing and recovery impact loads, and ground hoisting loads, each load parameter set is timestamped. The timestamped load parameter sets are then linked and organized according to the loading sequence to generate a loading execution record.

[0079] S5.3 Collect strain measurement signals, acceleration measurement signals and acoustic emission monitoring signals from the flight measurement records, and combine them with the comprehensive flaw detection results after recovery to perform time synchronization and annotation, and generate preliminary time-series observation data.

[0080] Specifically, after collecting strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals from the flight measurement records, these signals are categorized and organized according to the sampling time, channel number, and flight phase sequence in the flight measurement records, forming signal correspondences corresponding to each sampling time and monitoring location. Based on the unified time reference in the flight measurement records, the strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals are time-aligned to determine the correspondences between them at the same sampling time, forming a unified time signal set. Combined with the defect location, defect type, defect size, and flaw detection confirmation location from the recovered comprehensive flaw detection results, the signals at each time point in the unified time signal set are matched with the corresponding monitoring locations. Relevant signal segments in the unified time signal set are then labeled according to their defect location, defect type, defect size, and flaw detection confirmation location, forming an observation signal set with defect markings and time correspondences. Finally, the time-synchronized and labeled observation signal set is linked and organized according to the sampling time sequence to generate preliminary time-series observation data.

[0081] S5.4. Perform noise filtering, signal enhancement, and spatial location correlation on the preliminary time-series observation data to form time-series observation data.

[0082] S5.4.1 Perform amplitude enhancement, spectrum enhancement, and filtering and denoising processing on the preliminary time series observation data to generate enhanced and denoised signal data.

[0083] Specifically, an amplitude enhancement method is used to linearly amplify each signal in the preliminary time-series observation data according to a preset amplitude gain coefficient to generate an amplitude-enhanced signal. A fast Fourier transform method is used to process the amplitude-enhanced signal, extract the main frequency components, and enhance the spectral characteristics of the signal according to the spectral amplification factor to generate a spectral-enhanced signal. A digital filtering method is used to filter and denoise the spectral-enhanced signal, such as low-pass filtering or band-pass filtering, to filter out high-frequency and low-frequency interference components and generate enhanced and denoised signal data.

[0084] It should also be noted that the specific steps for obtaining the preset amplitude gain coefficient include: determining the upper and lower limits of the original amplitude of the signal based on the sampling range of the enhanced and denoised signal and the sensitivity of the measuring instrument; setting the target amplitude range in combination with actual observation needs, for example, hoping to enhance the signal amplitude to between 1.2 and 2 times the original amplitude; determining the value of the amplitude gain coefficient according to the set target amplitude range; applying the amplitude gain coefficient to each sampling point of each signal in the preliminary time series observation data to achieve linear amplitude enhancement and generate an amplitude-enhanced signal.

[0085] S5.4.2. Spatial matching is performed between the enhanced and denoised signal data and the geometric reference file, and outlier samples are removed to generate verified spatially labeled signal data.

[0086] Specifically, the process involves reading the geometric reference file and sensor placement dataset, finding the corresponding scan path points based on the timestamps in the enhanced and denoised signal data, and determining the corresponding sensor positions. A three-dimensional coordinate transformation method is used to convert the local coordinates of the sensor positions into global coordinates in the geometric reference file. Nearest neighbor matching is performed on each signal sample, calculating the Euclidean distance between the sample coordinates and the surface points of the geometric reference file, and using the minimum distance as the spatial matching error. Samples with excessive spatial matching errors are removed according to a preset spatial error threshold. For the remaining samples, the median and median absolute deviation of the amplitude are calculated for each channel, and samples with abnormal amplitudes are marked and removed using the median absolute deviation discrimination method. Each signal sample after spatial matching and anomaly removal is appended with a timestamp, global coordinates, and sensor identifier, and output in a unified format to generate verified spatially labeled signal data.

[0087] It should also be noted that the steps for setting the spatial error threshold are as follows: read the point cloud resolution information from the geometric reference file, and read the sensor positioning accuracy and installation error from the sensor layout dataset; synthesize the sensor positioning accuracy and installation error to obtain the value of the synthesized positioning error, for example, about one millimeter; obtain the point cloud pitch and take half of the point cloud pitch as the point cloud resolution reference; select the larger value between the synthesized positioning error and the point cloud resolution reference as the basic threshold; and scale up the basic threshold proportionally according to the safety factor, for example, from one to one and a half, to form the spatial error threshold.

[0088] S5.4.3. The spatially labeled signal data after verification is organized into a unified format to form time-series observation data.

[0089] Specifically, the timestamps, 3D spatial coordinates, signal amplitudes, and sensor type information are read from the verified spatially labeled signal data and reordered according to the timestamps in ascending order. The 3D spatial coordinates and signal amplitudes are grouped according to sensor type, with strain gauge data, accelerometer data, and acoustic emission sensor data organized into independent sequences. The timestamps of each sequence are corrected for time step size to ensure consistent time intervals between adjacent records, for example, adjusting the time step to one millisecond. The 3D spatial coordinate format in each record is standardized to a fixed coordinate order, such as arranged in a fixed order of "X, Y, Z". The signal amplitude field is standardized to the same decimal precision, for example, retaining three decimal places. The sorted, grouped, time-stepped, coordinate-formatted, and signal amplitude-precision-standardized records are then integrated according to the timestamp order to form time-series observation data.

[0090] It should be noted that, based on the digital modeling structure, spatial matching is performed on the preliminary time-series observation data, outlier data is removed and signal enhancement is performed to ensure the accuracy and consistency of the data. Through real-time updates via data processing, accurate time-series observation data is generated.

[0091] S6. Assimilate the parameters and damage state of the time-series observation data to generate a set of post-correction parameters, and calculate the crack propagation and remaining life distribution.

[0092] S6.1 Align the time-series observation data with the loaded execution record in time and space to generate synchronization calibration signal data.

[0093] Specifically, the process involves reading each set of time tags and their corresponding spatial observation positions from the time-series observation data, and reading each set of time tags and their corresponding loading execution positions from the loading execution record. Based on the time tags, the spatial observation positions and loading execution positions are sorted along a unified time axis, and a matching process is performed line by line based on the same time tags. When time tag offsets exist, linear interpolation is performed on the time tags of the loading execution record to obtain interpolated time tags consistent with the time tags of the time-series observation data, and the corresponding interpolated loading execution position is calculated. After the time tags are completely consistent, the loading execution position of the loading execution record is adjusted point-by-point according to the spatial observation positions in the time-series observation data and the loading execution positions in the loading execution record, ensuring consistency with the spatial observation positions in the time-series observation data, thus generating synchronization calibration signal data.

[0094] It should be noted that by using a digital modeling structure, the time-series observation data and the loading execution records are precisely synchronized, and the response characteristics of defects in the model are updated in real time, ensuring the high accuracy and timeliness of the data.

[0095] S6.2. Map and compare the synchronous calibration signal data with the size parameters, shape parameters and spatial position parameters of the defects in the initial state of the digital modeling structure, adjust the defect response characteristics, and generate the corrected defect response data.

[0096] Specifically, the time stamp, spatial observation position, and corresponding response amplitude are read line by line from the synchronous calibration signal data. Simultaneously, the size, shape, and spatial position parameters of the defects in the initial state of the digital modeling structure are read line by line. Based on the one-to-one correspondence of the spatial position parameters, the spatial observation positions in the synchronous calibration signal data are paired with the spatial position parameters of the defects in the initial state of the digital modeling structure. When the spatial positions match, the corresponding size and shape parameters are compared line by line with the response amplitudes in the synchronous calibration signal data. During the comparison process, the response amplitudes in the synchronous calibration signal data are adjusted proportionally, offset, or corrected using existing interpolation methods to ensure consistency between the adjusted response amplitudes and the corresponding size and shape parameters. After adjusting the response amplitudes of all synchronous calibration signal data, the adjusted response amplitudes are combined with the time stamp and spatial observation position of the original synchronous calibration signal data to generate corrected defect response data.

[0097] S6.3 Assimilate the parameters and damage state of the corrected defect response data to generate a set of post-correction parameters.

[0098] Specifically, based on the corrected defect response data, transfer function, and local stress data, a priori parameter vector consisting of the size, shape, and spatial location parameters of defects is extracted from the initial state of the digitally modeled structure and used as initial parameters. According to the transfer function, the dynamic response of each defect's size, shape, and spatial location parameters in the priori parameter vector is calculated. By performing numerical transfer and response mapping on the local stress data, a predicted response sequence under the same loading condition is obtained. The predicted response sequence is compared with the corrected defect response data at each time step, and the residual sequence is calculated as the observation error. Using a particle filtering method, the priori parameter vector is iteratively corrected based on the residual sequence and the preset observation noise covariance to generate a new parameter vector. The new parameter vector is then forward-predicted again through the transfer function, and the residual calculation and parameter update steps are repeated, iterating cyclically according to the time window until the parameters converge. The converged parameter vector is summarized, and consistency verification and formatted output are performed to generate the post-correction parameter set.

[0099] It should also be explained that the specific process for presetting the observation noise covariance is as follows: Observation error samples are selected from the observation data collected under stable loading conditions; the observation error samples are subjected to time-by-time differencing to obtain an error sequence; the error sequences are classified and organized according to channel or sampling location, for example, the error sequences of strain gauges, accelerometers, and acoustic emission sensors are classified separately, and sorted according to the spatial location of the sampling points within the same sensor type; statistical features such as maximum error, minimum error, and average error are extracted from each type of error sequence; the observation noise variance range for each channel is set based on the statistical features, for example, the variance is set to 50% of the square of the maximum error to 150% of the square of the average error, as the noise fluctuation range; the variances of each channel are arranged according to the time sequence of acquisition and combined to form the observation noise covariance. A time window is a fixed or variable length interval defined in continuous time series data, used for data processing, analysis, or calculation within the interval; The specific steps for performing iterations until parameter convergence are as follows: Set initial parameter values ​​and convergence criteria. For example, a convergence threshold of 0.001 indicates that convergence is considered achieved when the parameter change is less than 0.1% between two consecutive iterations. The maximum number of iterations can be set to 50 to prevent infinite loops. Input the initial parameters into the calculation process to obtain the first iteration result. Update the parameters based on the first iteration result to generate the parameter values ​​for the next iteration. Calculate the updated parameters and compare them with the previous iteration result to determine if the change is less than the convergence threshold. If the change is greater than the convergence threshold, continue updating the parameters and repeating the iteration steps. If the change is less than the convergence threshold or the maximum number of iterations is reached, stop the iteration and output the converged parameter values ​​as the final result. The particle filtering method is a recursive Bayesian filtering technique based on random sampling, used to estimate state variables under nonlinear or non-Gaussian conditions.

[0100] S6.4. Based on the later correction parameter set, calculate the crack growth curve of the defect under different loads and generate crack propagation trajectory data.

[0101] Specifically, based on the post-correction parameter set, the initial crack size and crack location of each defect are determined, and load conditions, including load amplitude and frequency, are set. The defect size, morphology, and spatial location parameters in the post-correction parameter set are matched with the load conditions to generate initial crack state data. Using a crack propagation calculation method, the crack growth rate of each defect under a specified load is calculated using the stress intensity factor and material fracture mechanics parameters. The crack length is iteratively updated with the number of load cycles, and the crack propagation amount in each iteration is recorded. The crack propagation iteration results for each defect are summarized and processed to generate a crack propagation trajectory dataset according to spatial location and crack development sequence. The crack growth curves and crack propagation trajectories of each defect are compiled to generate crack propagation trajectory data, recording the crack length increment, crack tip position, and cumulative load cycle count.

[0102] It should also be noted that the crack propagation calculation method refers to the method of calculating the length increase and propagation direction of the crack under each load cycle or load action based on the initial crack size, material properties, load conditions and defect geometry.

[0103] S6.5. Based on crack propagation trajectory data and structural failure criteria, assess the remaining life of the defect and generate crack propagation and remaining life distribution.

[0104] Specifically, based on crack propagation trajectory data, a point-by-point comparison method is used to sequentially read the sequence of crack length changes of each defect with load cycles or time, and obtain the critical crack length specified in the structural failure criterion. Each data point in the crack length sequence is compared with the critical crack length point by point, and the first time the crack length reaches or exceeds the critical crack length specified in the structural failure criterion is recorded. For example, 80% of the minimum wall thickness of a sub-stage rocket shell or the critical crack length calculated based on the material fracture toughness is taken as the reference value for determining failure. The cumulative load cycles or time experienced by the defect from the current state to the failure state is taken as the remaining lifetime, and the crack lengths and corresponding remaining lifetimes of all defects are summarized to generate crack propagation and remaining lifetime distribution.

[0105] It should also be noted that structural failure criteria are the standards used to determine when a structure will experience functional failure or damage under external loads; they typically include parameters such as critical stress, critical strain, critical crack length, fatigue limit, or fracture toughness.

[0106] It should be noted that by assimilating multi-source time-series observation data such as strain, acceleration, and acoustic emission with loading records, dynamic updates of material parameters, defect propagation rates, and damage states in the digital modeling structure are achieved, thereby forming a set of post-correction parameters that can reflect the evolution characteristics of the actual service process. In fatigue life prediction, it often relies on fixed material parameters, static stress fields, or preset damage, lacking real-time feedback correction capabilities, which leads to deviations between crack propagation calculations and actual evolution patterns.

[0107] In summary, this invention achieves accurate reproduction of real defects in size, shape, and spatial location by mapping an initial defect list to a digitally modeled structure. This ensures that the digitally modeled structure possesses an initial defect state consistent with the physical object, avoiding the bias caused by traditional life assessments relying on idealized defects. Furthermore, by integrating material property data to establish a transfer function between defect stress response and external vibration loads, complex loads can be accurately converted into local stress fields within the defects, providing a quantitative basis for fatigue loading and damage calculation. Ultimately, this invention achieves high-precision fatigue life prediction based on real structure and real defect characteristics, improving the reliability of life prediction results.

[0108] 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 predicting the fatigue life of a rocket's first-stage casing based on nondestructive testing, characterized in that: include, A three-dimensional geometric scan of the external structure of the first stage rocket body shell was performed to generate a geometric reference file. Using geometric reference files, multi-mode non-destructive testing was performed on the external structure of the first-stage rocket body shell to generate the original test dataset; Extract the size, shape, and location parameters of defects from the original detection dataset, integrate them, and generate an initial defect list; The initial defect list is mapped to the digital modeling structure, the initial state of the digital modeling structure is established, local stress is calculated, and the transfer function is generated. The fatigue load spectrum is constructed based on the transfer function, and the stress response of the rocket's first-stage shell is calculated through finite element simulation. Strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals are integrated simultaneously to form time-series observation data. The parameters and damage state of the time-series observation data are assimilated to generate a set of post-correction parameters, and the crack propagation and remaining life distribution are calculated.

2. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for generating the geometric reference file are as follows: Using a high-precision laser scanner, three-dimensional point cloud data was collected from the surface of the external structure of the first stage rocket body shell to generate raw point cloud data. The original point cloud data is denoised and filtered to generate a clean point cloud dataset, and the surface fitting method is used to reconstruct the three-dimensional geometric surface of the external structure. Features are extracted from the three-dimensional geometric surface of the external structure to generate geometric feature annotation data, which is then fused with the three-dimensional geometric surface of the external structure to generate a geometric reference file.

3. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for generating the original detection dataset are as follows. Extract the three-dimensional geometric information and feature positions of the first-stage rocket body shell from the geometric reference file to generate path planning data; Based on the path planning data, configure the scanning position, scanning angle, and scanning accuracy parameters for multi-mode non-destructive testing, and generate a non-destructive testing execution plan; The non-destructive testing implementation scheme is used to perform multi-mode scanning on the external structure of the rocket's first-stage shell, collect signal data in real time, and form the original test signal stream. The original detection signal stream is initially synchronized and labeled in time, and each signal is associated with the corresponding geometric position of the external structure to generate the original detection dataset.

4. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for generating the initial defect list are as follows: The original detection dataset is processed by time synchronization, denoising enhancement and spatial registration to generate a multi-mode detection set; The system identifies suspected defect regions from a multi-mode detection set, generates defect region segmentation results, and analyzes and calculates the size, shape, and spatial location parameters of the defects to generate candidate defect data. The candidate defect data is fused using multiple modes and its quality is checked to identify duplicates and conflicts, and then integrated to generate an initial defect list.

5. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for establishing the initial state of the digital modeling structure are as follows. Map the size, shape, and location parameters of defects in the initial defect list to a geometric reference file to generate defect annotation data; The size, shape, and location parameters of defects in the defect annotation data are summarized and processed to establish the initial state of the digital modeling structure.

6. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for generating the transfer function are as follows: The initial state of the digitally modeled structure is fused with the material property data from the geometric reference file to generate local stress analysis data; Based on the local stress analysis data, the local stress in the defect area under different vibration loads is calculated, and a local stress dataset is generated. Time-domain and frequency-domain analyses were performed on the local stress dataset to establish the mapping relationship between the defect stress response and the external vibration load. The results were then verified and optimized to generate the transfer function.

7. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for generating time-series observation data are as follows. The transfer function is coupled with the local stress dataset to generate a service load reconstruction scheme; According to the service load reconstruction scheme, the flight phase load, separation load, landing and recovery impact load and ground hoisting load are input into the digital modeling structure, the load parameters and timestamps are recorded, and the loading execution record is generated. Strain measurement signals, acceleration measurement signals, and acoustic emission monitoring signals from the flight measurement records are collected and time-synchronized and labeled in conjunction with the comprehensive flaw detection results after recovery, generating preliminary time-series observation data; Noise filtering, signal enhancement, and spatial location correlation are performed on the preliminary time-series observation data to form time-series observation data.

8. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 7, characterized in that: The process of filtering noise, enhancing signals, and correlating spatial locations in the preliminary time-series observation data to form time-series observation data involves the following specific steps. The initial time-series observation data are subjected to amplitude enhancement, spectrum enhancement, and filtering and denoising to generate enhanced and denoised signal data. The enhanced and denoised signal data is spatially matched with the geometric reference file, and outlier samples are removed to generate verified spatially labeled signal data. The spatially labeled signal data after verification are organized into a unified format to form time-series observation data.

9. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for generating the post-correction parameter set are as follows. The time-series observation data is aligned with the loaded execution records in time and space to generate synchronization calibration signal data; The synchronous calibration signal data is mapped and compared with the size parameters, shape parameters and spatial location parameters of the defects in the initial state of the digital modeling structure, and the defect response characteristics are adjusted to generate corrected defect response data. The parameters and damage state of the corrected defect response data are assimilated to generate a set of post-correction parameters.

10. The method for predicting the fatigue life of a rocket first-stage casing based on non-destructive testing according to claim 1, characterized in that: The specific steps for calculating crack propagation and remaining lifetime distribution are as follows. Based on the post-correction parameter set, calculate the crack growth curve of the defect under different loads and generate crack propagation trajectory data. Based on crack propagation trajectory data and structural failure criteria, the remaining life of defects is assessed, and crack propagation and remaining life distributions are generated.

Images