Low-altitude technology product quality evaluation method and system based on artificial intelligence

Abstract

The present invention discloses an artificial intelligence-based low-altitude technical product quality assessment method and system, comprising the following steps: S1. collecting and constructing an aircraft structure inspection dataset for key structural parts of a low-altitude aircraft; S2. generating a preprocessed aircraft inspection dataset; S3. generating an aircraft microcrack detection feature vector set; S4. obtaining a preliminary candidate aircraft microcrack identification set; S5. outputting a globally optimized final aircraft microcrack identification result; S6. mapping the final aircraft microcrack identification result to a low-altitude aircraft three-dimensional structural model, and constructing a microcrack structural parameter model based on the mapping result. The specific location, size, morphology, direction, and expansion trend of the aircraft microcracks are quantitatively described to form an aircraft microcrack quality assessment report. The present invention improves the optimization effect of the crack identification candidate area in terms of morphological integrity and spatial coherence.

Description

Technical Field

[0001] The present invention relates to the technical field of low-altitude product evaluation, and in particular to a low-altitude technical product quality evaluation method and system based on artificial intelligence. Background Art

[0002] There are many low-altitude technology products, the most common of which are low-altitude aircraft. With the widespread application of low-altitude aircraft in military reconnaissance, urban logistics, and emergency rescue, their structural reliability and safety issues are receiving increasing attention. Low-altitude aircraft usually have lightweight and highly maneuverable structural characteristics. During long-term operation in complex environments, tiny cracks are very likely to appear in key parts of the structure. Due to the small size and unstable shape of microcracks, it is often difficult to detect and accurately evaluate them early through traditional methods. However, their potential structural degradation and failure risks pose a serious threat to the safe operation of aircraft.

[0003] In existing technologies, microcrack detection mostly relies on single data source methods such as ultrasonic testing, infrared thermal imaging, strain response analysis or vibration characteristic identification. Although the crack location or morphology can be preliminarily judged under specific conditions, there are significant limitations: First, the single physical quantity monitoring method lacks a comprehensive reflection of the multi-dimensional characteristics of the crack and cannot fully characterize the morphology, depth and expansion trend of the crack; second, traditional image processing and signal analysis methods have low accuracy in identifying microcracks in complex structures, and often misjudge or miss detection due to background noise interference, data dimension redundancy or diverse crack morphology; in addition, the current mainstream optimization and identification algorithms are mostly based on continuous space modeling, lacking effective modeling and global search capabilities for discrete features of microcrack detection, and are prone to falling into local optimality, resulting in limited accuracy and stability of detection results.

[0004] In recent years, some studies have begun to attempt to apply multi-source data fusion technology to aircraft structural health testing to improve the comprehensiveness and robustness of microcrack identification. However, there are still deficiencies in data processing procedures, feature extraction methods, and crack identification mechanisms. For example, multi-source data cannot be effectively synchronized and standardized, resulting in information redundancy and signal distortion in the fusion results. At the same time, there is a lack of specialized modeling and optimization mechanisms for complex features such as edge gradients and texture directions, making it difficult to fully explore the structural semantic characteristics of microcracks. In addition, existing identification processes based on traditional optimization algorithms often lack discrete feature adaptability and multi-peak distribution optimization mechanisms, and are unable to cope with the complex and changeable distribution patterns of microcrack areas, thereby affecting the reliability and accuracy of the overall detection results.

[0005] In summary, there is an urgent need for a new method for microcrack quality evaluation that integrates multi-source data fusion, high-dimensional feature modeling and intelligent optimization algorithm. Summary of the Invention

[0006] One purpose of the present invention is to propose a low-altitude technical product quality assessment method and system based on artificial intelligence, which improves the optimization effect of crack identification candidate areas in terms of morphological integrity and spatial coherence.

[0007] According to an embodiment of the present invention, a method for evaluating the quality of low-altitude technology products based on artificial intelligence includes the following steps:

[0008] S1. Collect key structural parts of low-altitude aircraft and construct an aircraft structure detection dataset;

[0009] S2. Preprocess the aircraft structure detection data set to generate a preprocessed aircraft detection data set;

[0010] S3. Generate a set of aircraft microcrack detection feature vectors by extracting crack edge gradient, energy distribution, and texture features from the preprocessed aircraft detection dataset;

[0011] S4. A discrete bitterfish optimization algorithm is used to perform a global initial search on the set of aircraft microcrack detection feature vectors. A jump search and local optimization mechanism in the discrete data space are used to obtain a preliminary candidate set of aircraft microcrack identification.

[0012] S5. Apply a multimodal distribution gravity algorithm to perform global fusion optimization on the preliminary candidate aircraft microcrack identification set. By simulating the attraction of multimodal distribution in the gravitational field, multiple candidate microcrack regions are integrated and screened, and the globally optimized final aircraft microcrack identification result is output.

[0013] S6. Map the final identification results of aircraft microcracks to the three-dimensional structural model of the low-altitude aircraft, and construct a microcrack structural parameter model based on the mapping results. Quantify the specific location, size, shape, direction and expansion trend of the aircraft microcracks to form an aircraft microcrack quality assessment report.

[0014] Optionally, the S1 includes the following steps:

[0015] S11. Use strain sensors, ultrasonic detection devices, infrared imagers, and micro-vibration detectors to collect raw structural signal data from key structural parts of low-altitude aircraft. The collected raw structural signal data includes strain response raw data, ultrasonic detection raw image data, infrared thermal imaging raw data, and micro-vibration raw signal data.

[0016] S12. Digitally process the collected raw structural signal data and convert them into strain response signal data, ultrasonic detection image data, infrared thermal imaging data, and micro-vibration signal data;

[0017] S13. The strain response signal data , ultrasonic detection image data , infrared thermal imaging data and micro-vibration signal data According to the predetermined data fusion rules, time and space synchronization and format unification are performed to build the aircraft structure detection data set :

[0018] .

[0019] Optionally, the S2 includes the following steps:

[0020] S21. Aircraft structure detection dataset The noise of various detection data in the dataset is filtered out respectively. A sliding window filter is used to perform time domain noise reduction on the strain response signal data and micro-vibration signal data. A spatial domain high-pass filter is performed on the ultrasonic detection image data and infrared thermal imaging data to form a preliminary noise-reduced aircraft detection data set.

[0021] S22. Remove outliers from the aircraft detection dataset after preliminary noise reduction, remove local outliers in various detection data based on the triple standard deviation method, and construct an outlier-free aircraft detection dataset;

[0022] S23. Perform data normalization on the abnormal aircraft detection data set and scale all detection data to a uniform range. , get the normalized aircraft detection dataset;

[0023] S24. Perform feature enhancement processing on the normalized aircraft detection dataset, extract multi-scale frequency components and short-time energy features from the strain response signal data and micro-vibration signal data, perform edge enhancement and texture gradient enhancement on the ultrasonic detection image data and infrared thermal imaging data, and construct the final preprocessed aircraft detection dataset. .

[0024] Optionally, S3 includes the following steps:

[0025] S31. Based on the final pre-processed aircraft detection dataset Perform multi-dimensional feature extraction operations on various types of data, and extract signal energy distribution characteristics from strain response signal data and micro-vibration signal data , frequency component characteristics And envelope change characteristics ;

[0026] S32. Extracting microcrack edge gradient features from ultrasonic testing image data and infrared thermal imaging data and texture direction characteristics ;

[0027] S33. Set signal features and image feature set After unified standardization, the vectors are spliced into the feature vectors for aircraft microcrack detection. , among which The feature vector corresponding to the detection area is expressed as:

[0028] ;

[0029] in, 、 、 、 、 Respectively represent The characteristic values of signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics and texture direction characteristics in each region;

[0030] S34. Construct a set of feature vectors extracted from all detection areas to form a set of feature vectors for aircraft microcrack detection:

[0031] ;

[0032] in, is the total number of detection areas.

[0033] Optionally, the signal energy distribution feature is used to measure the response intensity of the structure in the detection area under stimulation, and the higher the energy, the deeper the crack penetration or the more significant the stress concentration.

[0034] The frequency component characteristics reflect the dynamic vibration behavior of the crack area, and the frequency drift or abnormal frequency point reveals the stiffness change around the crack;

[0035] The envelope variation feature represents the non-stationarity of the signal's time domain envelope and is suitable for identifying high mutation areas at the microcrack tip;

[0036] The microcrack edge gradient characteristics reflect the clarity of the crack boundary and the degree of edge mutation;

[0037] The texture direction feature describes the crack propagation path regularity on the surface of the structure and is used to determine whether the crack propagates along the principal stress direction of the structure.

[0038] Optionally, the S4 includes the following steps:

[0039] S41. Set the aircraft microcrack detection feature vector Mapping to discrete search space For each feature vector in the aircraft microcrack detection feature vector set Using discrete mapping function based on fuzzy coding Generate the corresponding binary representation , the discrete mapping function Retain key information of microcracks in low-altitude aircraft in terms of morphology, size and texture to form a binary representation set

[0040] ;

[0041] S42. Initialize discrete bitterfish population , where each discrete bitterfish individual is a binary representation sequence of several candidate microcrack regions, and each component in the discrete bitter fish individual is selected from the binary representation set, reflecting the detection path or regional distribution of the current discrete bitter fish individual in the feature space, and setting the maximum number of iterations and the initial adaptive jump step size , define a domain-specific fitness function for the microcrack characteristics of low-altitude aircraft Used to evaluate the feature matching degree and microcrack region coherence of candidate microcrack regions:

[0042] ;

[0043] in, Represents discrete bitterfish individuals The set of candidate microcrack region indices contained in , To reflect the candidate microcrack area Scoring functions for microcrack morphology, size, microcrack edge gradient characteristics, and texture orientation, To describe the penalty function of inconsistency and redundancy of candidate microcrack regions within discrete bitterfish individuals, and are positive weighting coefficients respectively;

[0044] S43. Adopting the adaptive discrete jump update mechanism to optimize each discrete bitter fish individual in the discrete bitter fish population, based on the optimal discrete bitter fish individual of the current generation. With any discrete bitterfish individual The difference between Update discrete bitterfish individuals:

[0045] ;

[0046] in, represents the candidate microcrack region combination corresponding to the mth discrete bitterfish individual in the tth iteration, represents the updated combination of candidate microcrack regions corresponding to discrete bitterfish individuals, Represents a bitwise exclusive OR operation, For the The jump step size is adaptively adjusted according to the feedback of micro-crack characteristic data of low-altitude aircraft. At the same time, the binary difference between the current optimal solution and the discrete bitterfish individuals to be updated is considered;

[0047] S44. Perform local search on the updated discrete bitter fish individuals to further improve the feature consistency of the candidate microcrack region, wherein the local search performs fine-grained perturbation on the binary representation of the discrete bitter fish individuals, uses a neighborhood search operator to fine-tune some bits, and calculates the feature consistency of the candidate microcrack region according to the fitness function. The feedback is corrected in real time, so that the binary representation of the candidate microcrack region accurately reflects the actual distribution and structural characteristics of the microcracks in the low-altitude aircraft;

[0048] S45. Repeat S43 and S44 until the maximum number of iterations is reached After that, the fitness function is selected from the final population The discrete bitterfish individual set with the highest value constitutes the preliminary candidate set for aircraft microcrack identification , each candidate discrete bitterfish individual in the preliminary candidate aircraft microcrack identification set It corresponds to the binary representation of the potential microcrack area in the low-altitude aircraft structure, and comprehensively reflects the distribution, coherence and morphological characteristics of the microcracks in the microcrack area.

[0049] Optionally, the S5 includes the following steps:

[0050] S51. For each candidate discrete bitterfish individual in the preliminary candidate aircraft microcrack identification set The gravitational mass model is constructed based on the structural characteristics of the microcrack area of the corresponding low-altitude aircraft, and the gravitational mass of each candidate discrete bitterfish individual is defined. :

[0051] ;

[0052] in, Candidate discrete bitterfish individuals The fitness function of is the total number of individuals in the candidate set;

[0053] S52. Construct a multi-peak gravitational field between each candidate discrete bitter fish individual in the candidate set, and calculate the candidate discrete bitter fish individual through multi-peak gravitational calculation. and candidate discrete bitterfish individuals The gravitational force between:

[0054] ;

[0055] in, is the gravitational constant, Represents a candidate discrete bitterfish individual and candidate discrete bitterfish individuals The discrete distance measure between To prevent small positive numbers from dividing by zero;

[0056] S53. For each candidate discrete bitterfish individual The comprehensive gravitational vector is calculated by weighted aggregation based on the gravitational forces exerted by all other candidate discrete bitterling individuals:

[0057] ;

[0058] And based on the comprehensive gravitational vector The candidate discrete bitterfish individuals are clustered and grouped according to the direction and amplitude of the gravitational force, and the candidate discrete bitterfish individuals with similar gravitational force and reflecting the coherence of the microcrack regional distribution are classified into the same cluster;

[0059] S54. Screen the representative candidate discrete bitterfish individuals with the best gravity vector aggregation effect in each cluster to form the final identification set of aircraft microcracks after global fusion optimization ,Each candidate discrete bitterfish individual in the final identification set of aircraft microcracks can comprehensively reflect the true distribution, coherence and structural characteristics of the microcrack area in the low-altitude aircraft structure, and the output result is used as the final identification result of the low-altitude aircraft microcracks.

[0060] Optionally, the S54 includes the following steps:

[0061] S541. Based on the final identification set of aircraft microcracks after fusion optimization, each candidate discrete bitterfish individual in the final identification set of aircraft microcracks is The corresponding binary representation is mapped to the original structure detection dataset The spatial position of the microcrack area in the three-dimensional structural model , and associated with its feature vector set for aircraft microcrack detection The corresponding eigenvector in ;

[0062] S542. Constructing a comprehensive scoring function for structural attributes The signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics and texture direction characteristics are jointly weighted evaluated with the individual optimization results:

[0063] in, is the weighting coefficient;

[0064] S543. Comprehensive scoring function based on structural attributes The microcrack risk level determination rule is constructed based on the combined interval of signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics, and texture direction characteristic parameters, and the classification judgment is performed in combination with the following characteristic indicators and physical meanings:

[0065] If any two of the following are met:

[0066] : The crack edge has a significant mutation;

[0067] : Crack response energy is high;

[0068] : The envelope discontinuity is strong;

[0069] : High overall score;

[0070] It is judged as a high-risk microcrack area, indicating that the structure has deep cracks, high stress areas or high growth rate risks;

[0071] If only one of the above is met, and at the same time: 、 , it is determined to be a medium-risk microcrack area, indicating that the crack is in the early development stage or the expansion direction is unclear;

[0072] If satisfied: 、 、 、 , it is judged as a low-risk microcrack area with a basically intact structure and only slight surface disturbances or background noise artifacts;

[0073] in, is the high risk judgment threshold, is the low risk judgment threshold, is the boundary of the middle judgment interval;

[0074] S544. Construct the final identification result set of aircraft microcracks:

[0075] ;

[0076] in, is the final risk level label.

[0077] An artificial intelligence-based low-altitude technology product quality evaluation system is used to implement an artificial intelligence-based low-altitude technology product quality evaluation method, including the following modules:

[0078] Structural data acquisition module, used to collect data on key structural parts of low-altitude aircraft and build aircraft structure detection data sets;

[0079] A data preprocessing module is used to preprocess the aircraft structure detection data set to generate a preprocessed aircraft detection data set;

[0080] A microcrack feature extraction module is used to generate a set of aircraft microcrack detection feature vectors by extracting crack edge gradients, energy distribution, and texture features from a preprocessed aircraft detection dataset;

[0081] Discrete bitter fish optimization module, used to perform global initial search on the aircraft microcrack detection feature vector set using the discrete bitter fish optimization algorithm to obtain a preliminary candidate aircraft microcrack identification set;

[0082] The multimodal distribution gravity optimization module is used to apply the multimodal distribution gravity algorithm to perform global fusion optimization on the preliminary candidate aircraft microcrack identification set and output the final aircraft microcrack identification result after global optimization;

[0083] The structural mapping and visualization module is used to map the final identification results of aircraft microcracks into the three-dimensional structural model of the low-altitude aircraft, and construct a microcrack structural parameter model based on the mapping results to quantitatively describe the specific location, size, shape, direction and expansion trend of the aircraft microcracks.

[0084] The beneficial effects of the present invention are:

[0085] (1) Aiming at the discrete characteristics of microcracks, the present invention introduces a discrete mapping mechanism based on fuzzy coding, maps the set of structural detection feature vectors into a set of binary feature representations, and constructs an adaptive jump search mechanism and a fine-grained perturbation local search mechanism that adapt to the characteristics of microcracks. This can effectively avoid local optimal traps and improve the optimization effect of crack identification candidate areas in terms of morphological integrity and spatial coherence.

[0086] (2) The present invention draws on the multi-peak gravitational mechanism in the gravitational field. By establishing a gravitational mass function based on fitness score for each candidate microcrack area and simulating the gravitational attraction process between multiple targets, a stable and globally optimized microcrack recognition model is constructed. Through weighted aggregation of gravitational vectors and regional clustering analysis, dynamic fusion and redundancy removal of microcrack areas can be achieved, thereby ensuring the structural consistency and practical significance of the final recognition results. It is particularly suitable for scenarios with complex multiple cracks coexisting, and can significantly reduce the misrecognition rate and repeated judgment rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0088] Figure 1This is a flow chart of a low-altitude technical product quality evaluation method based on artificial intelligence proposed by the present invention. DETAILED DESCRIPTION

[0089] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0090] refer to Figure 1 , a low-altitude technology product quality evaluation method based on artificial intelligence, including the following steps:

[0091] S1. Collect key structural parts of low-altitude aircraft and construct an aircraft structure detection dataset;

[0092] S2. Preprocess the aircraft structure detection data set to generate a preprocessed aircraft detection data set;

[0093] S3. Generate a set of aircraft microcrack detection feature vectors by extracting crack edge gradient, energy distribution, and texture features from the preprocessed aircraft detection dataset;

[0094] S4. A discrete bitterfish optimization algorithm is used to perform a global initial search on the set of aircraft microcrack detection feature vectors. A jump search and local optimization mechanism in the discrete data space are used to obtain a preliminary candidate set of aircraft microcrack identification.

[0095] S5. Apply a multimodal distribution gravity algorithm to perform global fusion optimization on the preliminary candidate aircraft microcrack identification set. By simulating the attraction of multimodal distribution in the gravitational field, multiple candidate microcrack regions are integrated and screened, and the globally optimized final aircraft microcrack identification result is output.

[0096] S6. Map the final identification results of aircraft microcracks to the three-dimensional structural model of the low-altitude aircraft, and construct a microcrack structural parameter model based on the mapping results. Quantify the specific location, size, shape, direction and expansion trend of the aircraft microcracks to form an aircraft microcrack quality assessment report.

[0097] In this embodiment, S1 includes the following steps:

[0098] S11. Use strain sensors, ultrasonic detection devices, infrared imagers, and micro-vibration detectors to collect raw structural signal data from key structural parts of low-altitude aircraft. The collected raw structural signal data includes strain response raw data, ultrasonic detection raw image data, infrared thermal imaging raw data, and micro-vibration raw signal data.

[0099] S12. Digitally process the collected raw structural signal data and convert them into strain response signal data, ultrasonic detection image data, infrared thermal imaging data, and micro-vibration signal data;

[0100] S13. The strain response signal data , ultrasonic detection image data , infrared thermal imaging data and micro-vibration signal data According to the predetermined data fusion rules, time and space synchronization and format unification are performed to build the aircraft structure detection data set :

[0101] .

[0102] In this embodiment, S2 includes the following steps:

[0103] S21. Aircraft structure detection dataset The noise of various detection data in the dataset is filtered out respectively. A sliding window filter is used to perform time domain noise reduction on the strain response signal data and micro-vibration signal data. A spatial domain high-pass filter is performed on the ultrasonic detection image data and infrared thermal imaging data to form a preliminary noise-reduced aircraft detection data set.

[0104] S22. Remove outliers from the aircraft detection dataset after preliminary noise reduction, remove local outliers in various detection data based on the triple standard deviation method, and construct an outlier-free aircraft detection dataset;

[0105] S23. Perform data normalization on the abnormal aircraft detection data set and scale all detection data to a uniform range. , get the normalized aircraft detection dataset;

[0106] S24. Perform feature enhancement processing on the normalized aircraft detection dataset, extract multi-scale frequency components and short-time energy features from the strain response signal data and micro-vibration signal data, perform edge enhancement and texture gradient enhancement on the ultrasonic detection image data and infrared thermal imaging data, and construct the final preprocessed aircraft detection dataset. .

[0107] In this embodiment, S3 includes the following steps:

[0108] S31. Based on the final pre-processed aircraft detection dataset Perform multi-dimensional feature extraction operations on various types of data, and extract signal energy distribution characteristics from strain response signal data and micro-vibration signal data , frequency component characteristics And envelope change characteristics ;

[0109] S32. Extracting microcrack edge gradient features from ultrasonic testing image data and infrared thermal imaging data and texture direction characteristics ;

[0110] S33. Set signal features and image feature set After unified standardization, the vectors are spliced into the feature vectors for aircraft microcrack detection. , among which The feature vector corresponding to the detection area is expressed as:

[0111] ;

[0112] in, 、 、 、 、 Respectively represent The characteristic values of signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics and texture direction characteristics in each region;

[0113] S34. Construct a set of feature vectors extracted from all detection areas to form a set of feature vectors for aircraft microcrack detection:

[0114] ;

[0115] in, is the total number of detection areas.

[0116] In this embodiment, the signal energy distribution characteristics are used to measure the response strength of the structure in the detection area under stimulation. The higher the energy, the deeper the crack penetration or the more significant the stress concentration.

[0117] The frequency component characteristics reflect the dynamic vibration behavior of the crack area, and the frequency drift or abnormal frequency points reveal the stiffness changes around the crack;

[0118] The envelope variation feature represents the non-stationary nature of the signal’s time domain envelope and is suitable for identifying high mutation areas at the microcrack tip.

[0119] The gradient characteristics of microcrack edges reflect the clarity of crack boundaries and the degree of edge mutation;

[0120] The texture direction characteristics describe the crack propagation path on the surface of the structure and are used to determine whether the crack propagates along the principal stress direction of the structure.

[0121] In this embodiment, S4 includes the following steps:

[0122] S41. Set the aircraft microcrack detection feature vector Mapping to discrete search space For each feature vector in the aircraft microcrack detection feature vector set Using discrete mapping function based on fuzzy coding Generate the corresponding binary representation , discrete mapping function Retain key information of microcracks in low-altitude aircraft in terms of morphology, size and texture to form a binary representation set ;

[0123] S42. Initialize discrete bitterfish population , where each discrete bitterfish individual is a binary representation sequence of several candidate microcrack regions, and each component in the discrete bitter fish individual is selected from the binary representation set, reflecting the detection path or regional distribution of the current discrete bitter fish individual in the feature space, and setting the maximum number of iterations and the initial adaptive jump step size , define a domain-specific fitness function for the microcrack characteristics of low-altitude aircraft Used to evaluate the feature matching degree and microcrack region coherence of candidate microcrack regions:

[0124] ;

[0125] in, Represents discrete bitterfish individuals The set of candidate microcrack region indices contained in , To reflect the candidate microcrack area Scoring functions for microcrack morphology, size, microcrack edge gradient characteristics, and texture orientation, To describe the penalty function of inconsistency and redundancy of candidate microcrack regions within discrete bitterfish individuals, and are positive weighting coefficients respectively;

[0126] S43. Adopting the adaptive discrete jump update mechanism to optimize each discrete bitter fish individual in the discrete bitter fish population, based on the optimal discrete bitter fish individual of the current generation. With any discrete bitterfish individual The difference between Update discrete bitterfish individuals:

[0127] ;

[0128] in, represents the candidate microcrack region combination corresponding to the mth discrete bitterfish individual in the tth iteration, represents the updated combination of candidate microcrack regions corresponding to discrete bitterfish individuals, Represents a bitwise exclusive OR operation, For the The jump step size is adaptively adjusted according to the feedback of micro-crack characteristic data of low-altitude aircraft. At the same time, the binary difference between the current optimal solution and the discrete bitterfish individuals to be updated is considered;

[0129] S44. Perform local search on the updated discrete bitter fish individuals to further improve the feature consistency of the candidate microcrack region. The local search performs fine-grained perturbations on the binary representation of the discrete bitter fish individuals, uses the neighborhood search operator to fine-tune some bits, and calculates the feature consistency according to the fitness function. The feedback is corrected in real time, so that the binary representation of the candidate microcrack region accurately reflects the actual distribution and structural characteristics of the microcracks in the low-altitude aircraft;

[0130] S45. Repeat S43 and S44 until the maximum number of iterations is reached After that, the fitness function is selected from the final population The discrete bitterfish individual set with the highest value constitutes the preliminary candidate set for aircraft microcrack identification , each candidate discrete bitterfish individual in the preliminary candidate aircraft microcrack identification set It corresponds to the binary representation of the potential microcrack area in the low-altitude aircraft structure, and comprehensively reflects the distribution, coherence and morphological characteristics of the microcracks in the microcrack area.

[0131] This implementation introduces a discrete mapping mechanism based on fuzzy coding to address the discrete characteristics of microcracks. It maps the set of structural detection feature vectors into a set of binary feature representations, and constructs an adaptive jump search mechanism and a fine-grained perturbation local search mechanism that adapt to the characteristics of microcracks. This can effectively avoid local optimal traps and improve the optimization effect of crack identification candidate areas in terms of morphological integrity and spatial coherence.

[0132] In this embodiment, S5 includes the following steps:

[0133] S51. For each candidate discrete bitterfish individual in the preliminary candidate aircraft microcrack identification set The gravitational mass model is constructed based on the structural characteristics of the microcrack area of the corresponding low-altitude aircraft, and the gravitational mass of each candidate discrete bitterfish individual is defined. :

[0134] ;

[0135] in, Candidate discrete bitterfish individuals The fitness function of is the total number of individuals in the candidate set;

[0136] S52. Construct a multi-peak gravitational field between each candidate discrete bitter fish individual in the candidate set, and calculate the candidate discrete bitter fish individual through multi-peak gravitational calculation. and candidate discrete bitterfish individuals The gravitational force between:

[0137] ;

[0138] in, is the gravitational constant, Represents a candidate discrete bitterfish individual and candidate discrete bitterfish individuals The discrete distance measure between To prevent small positive numbers from dividing by zero;

[0139] S53. For each candidate discrete bitterfish individual The comprehensive gravitational vector is calculated by weighted aggregation based on the gravitational forces exerted by all other candidate discrete bitterling individuals:

[0140] ;

[0141] And based on the comprehensive gravitational vector The candidate discrete bitter fish individuals are clustered and grouped according to the direction and amplitude of the gravitational vector, and the candidate discrete bitter fish individuals with similar gravitational effects and reflecting the coherence of the microcrack regional distribution are classified into the same cluster; S54. The representative candidate discrete bitter fish individuals with the best gravitational vector aggregation effect in each cluster are screened to form the final aircraft microcrack identification set after global fusion optimization. ,Each candidate discrete bitterfish individual in the final identification set of aircraft microcracks can comprehensively reflect the true distribution, coherence and structural characteristics of the microcrack area in the low-altitude aircraft structure, and the output result is used as the final identification result of the low-altitude aircraft microcracks.

[0142] This implementation integrates strain response signals, infrared thermal imaging, ultrasonic images, and micro-vibration data, and performs unified normalization, feature enhancement, and high-dimensional splicing on various signals to construct a multidimensional detection feature vector set. On this basis, a feature-sensitive scoring function and a structural attribute scoring model are introduced to perform weighted evaluations on the key attributes of the crack's energy response, frequency drift, edge gradient, and texture direction, thereby constructing a structural risk level determination mechanism. The output crack identification results not only have spatial coordinates, but also come with quantifiable risk labels and structural interpretation parameters, which are helpful for subsequent structural life prediction and maintenance decision support.

[0143] In this embodiment, S54 includes the following steps:

[0144] S541. Based on the final identification set of aircraft microcracks after fusion optimization, each candidate discrete bitterfish individual in the final identification set of aircraft microcracks is The corresponding binary representation is mapped to the original structure detection dataset The spatial position of the microcrack area in the three-dimensional structural model , and associated with its feature vector set for aircraft microcrack detection The corresponding eigenvector in ;

[0145] S542. Constructing a comprehensive scoring function for structural attributes The signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics and texture direction characteristics are jointly weighted evaluated with the individual optimization results:

[0146] ;

[0147] in, is the weighting coefficient;

[0148] S543. Comprehensive scoring function based on structural attributes The microcrack risk level determination rule is constructed based on the combined interval of signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics, and texture direction characteristic parameters, and the classification judgment is performed in combination with the following characteristic indicators and physical meanings:

[0149] If any two of the following are met:

[0150] : The crack edge has a significant mutation;

[0151] : Crack response energy is high;

[0152] : The envelope discontinuity is strong;

[0153] : High overall score;

[0154] It is judged as a high-risk microcrack area, indicating that the structure has deep cracks, high stress areas or high growth rate risks;

[0155] If only one of the above is met, and at the same time: 、 , it is determined to be a medium-risk microcrack area, indicating that the crack is in the early development stage or the expansion direction is unclear;

[0156] If satisfied: 、 、 、 , it is judged as a low-risk microcrack area with a basically intact structure and only slight surface disturbances or background noise artifacts;

[0157] in, is the high risk judgment threshold, is the low risk judgment threshold, is the boundary of the middle judgment interval;

[0158] S544. Construct the final identification result set of aircraft microcracks:

[0159] ;

[0160] in, is the final risk level label.

[0161] This implementation introduces a multi-dimensional feature joint scoring mechanism, combining five key microcrack structural features: signal energy, frequency components, envelope changes, image edges, and texture directions, to construct a microcrack risk level determination rule for low-altitude aircraft. This significantly improves the accuracy and reliability of microcrack identification, not only achieving precise positioning of microcrack areas in the three-dimensional structural model, but also enabling risk level determination based on the intensity of feature expression, effectively supporting the safe operation and maintenance and early warning decisions of low-altitude aircraft.

[0162] An artificial intelligence-based low-altitude technology product quality evaluation system is used to implement an artificial intelligence-based low-altitude technology product quality evaluation method, including the following modules:

[0163] Structural data acquisition module, used to collect data on key structural parts of low-altitude aircraft and build aircraft structure detection data sets;

[0164] A data preprocessing module is used to preprocess the aircraft structure detection data set to generate a preprocessed aircraft detection data set;

[0165] A microcrack feature extraction module is used to generate a set of aircraft microcrack detection feature vectors by extracting crack edge gradients, energy distribution, and texture features from a preprocessed aircraft detection dataset;

[0166] Discrete bitter fish optimization module, used to perform global initial search on the aircraft microcrack detection feature vector set using the discrete bitter fish optimization algorithm to obtain a preliminary candidate aircraft microcrack identification set;

[0167] The multimodal distribution gravity optimization module is used to apply the multimodal distribution gravity algorithm to perform global fusion optimization on the preliminary candidate aircraft microcrack identification set and output the final aircraft microcrack identification result after global optimization;

[0168] The structural mapping and visualization module is used to map the final identification results of aircraft microcracks into the three-dimensional structural model of the low-altitude aircraft, and construct a microcrack structural parameter model based on the mapping results to quantitatively describe the specific location, size, shape, direction and expansion trend of the aircraft microcracks.

[0169] Example 1

[0170] At a low-altitude aircraft test site, a low-altitude reconnaissance UAV numbered "LQX-17" returned to the ground maintenance center for a routine structural status assessment after completing a night flight mission in the mountains. This aircraft is a lightweight flight platform with a carbon fiber composite structure. It has flown a total of 142 times and accumulated flight time of more than 763 hours.

[0171] During the rapid structural inspection after the flight, the system detected a significant mutation in the strain response signal near the lower chord of the wing. The raw data recorded by the system showed that between 02:31 and 02:33, the strain sensor numbered "SG-12" continuously collected a maximum strain peak of "427 με", an increase of more than 115% from the average value of "198 με" in the previous cycle. This was accompanied by an abnormal frequency drift: the main vibration frequency recorded by the micro-vibration accelerometer "MV-08" suddenly shifted from the conventional 31.4 Hz to 45.7 Hz and continued to fluctuate.

[0172] The inspectors immediately started the method of the present invention to conduct intelligent evaluation of structural microcracks.

[0173] First, the system automatically aggregated multi-source structural data collected during this flight mission, including 2,640 strain response signals, 212 frames of infrared thermal imaging images, 195 sets of micro-vibration time-domain records, and 168 ultrasonic scans. The data preprocessing module successively completed noise filtering, outlier removal, and normalization operations, and applied STFT transformation to the micro-vibration signals to extract short-time energy envelope features, and performed edge detection and texture direction calibration on the infrared images.

[0174] Subsequently, based on the method of the present invention, the system constructed a microcrack feature space model, and generated a total of 258 sets of feature vectors in the feature extraction stage, involving five types of feature dimensions: signal energy (mean 76.2 nJ), frequency offset (maximum deviation up to 14.3 Hz), envelope mutation coefficient (average 0.61), edge gradient change rate (maximum 0.84), and texture direction offset (maximum 46.3°).

[0175] At 03:01, the system started the "Discrete Bitter Fish Optimization Algorithm" to search the feature space, using fuzzy coding to map all feature vectors to the discrete search space, initializing the discrete population size to 100. At the 23rd iteration, it was found that the fitness score of individual "x17" exceeded 0.83 for the first time. Its binary representation "101010011001..." corresponds to the upper edge of the left wing box, which the system determined to be a high potential microcrack area.

[0176] At 03:08, the system automatically entered the "Multi-peak Distribution Gravity Optimization" stage. After gravity model fusion analysis, the system finally output 6 suspected microcrack areas, namely:

[0177] The outer edge of the main wing lower chord beam (ID: R-1); the left wing middle section wing box (ID: R-2); the right wing leading edge seam (ID: R-3); the landing bulkhead connection point (ID: R-4); the battery compartment fixing bracket (ID: R-5); near the main beam in the middle of the main wing (ID: R-6).

[0178] The system automatically calculates a risk score (R score) for each region. For example, the R-2 region score is 0.91, and the risk level is judged to be "high". The corresponding envelope mutation characteristic value is 0.83, the frequency drift is 11.5 Hz, and the energy index reaches 108.3 nJ.

[0179] At 03:11, the system automatically generated a microcrack quality assessment report and pushed an alarm message to the maintenance personnel's terminal through the control center. After confirming the information, maintenance engineer Zhang used a portable crack detection coating agent to apply a coating inspection on the above six areas. The results showed that five of the areas had visible crack absorption stripes. The crack width range was confirmed to be 0.18mm to 0.43mm, and the largest crack was located in the R-2 area.

[0180] In the comparative experiment, the traditional inspection team used the "edge enhancement + manual comparison of thermal imaging images" method. Under the same data conditions, they only detected three areas (R-1, R-3, and R-6), failed to cover the R-2, R-4, and R-5 areas, and misjudged the R-6 area as a thermal artifact, resulting in a false alarm.

[0181] The following is a detailed comparison of the method of the present invention and the traditional method in this real detection task:

[0182] Detection Dimension Method of the present invention Traditional methods Improvement effect Total detection time 42 minutes 3.6 hours ↓ 80.5% Lower limit of identifiable crack size 0.17 mm 0.32 mm ↑ Resolution increased by 88% Multi-source fusion dimension 5-dimensional (strain + vibration + infrared + ultrasound + spectrum) 2D (image + temperature) ↑ Improved information integration Number of accurately identified areas 6 / 6 3 / 6 ↑ Recognition rate increased by 100% Number of misjudgments 0 2 ↓ Reduce error rate System automatic scoring confidence mean 0.82 No scoring mechanism ↑ Strong interpretability Number of missed inspection risk areas 0 3 ↓ Enhanced risk control capabilities

[0183] The test results were submitted to the maintenance center, and during the structural re-inspection, they fully matched the crack areas and parameter models identified by the method of the present invention, verifying the feasibility and high adaptability of the algorithm in real complex working conditions. In a multi-region, fine-level crack distribution environment, the method of the present invention demonstrated the advantages of high precision, high automation and high real-time performance, providing strong support for subsequent structural fatigue prediction and maintenance scheduling.

[0184] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for evaluating the quality of low-altitude technology products based on artificial intelligence, characterized in that: The steps include: S1. Collect key structural parts of low-altitude aircraft and construct an aircraft structure detection dataset; S2. Preprocess the aircraft structure detection data set to generate a preprocessed aircraft detection data set; S3. Generate a set of aircraft microcrack detection feature vectors by extracting microcrack edge gradients, envelope variation characteristics, energy distribution, frequency component characteristics, and texture characteristics from the preprocessed aircraft detection dataset; S4. A discrete bitterfish optimization algorithm is used to perform a global initial search on the set of aircraft microcrack detection feature vectors. A jump search and local optimization mechanism in the discrete data space are used to obtain a preliminary candidate set of aircraft microcrack identification. The S4 comprises the following steps: S41. Set the aircraft microcrack detection feature vector Mapping to discrete search space For each feature vector in the aircraft microcrack detection feature vector set Using discrete mapping function based on fuzzy coding Generate the corresponding binary representation , the discrete mapping function Retain key information of microcracks in low-altitude aircraft in terms of morphology, size and texture to form a binary representation set ; S42. Initialize discrete bitterfish population , where each discrete bitterfish individual is a binary representation sequence of several candidate microcrack regions, and each component in the discrete bitter fish individual is selected from the binary representation set, reflecting the detection path or regional distribution of the current discrete bitter fish individual in the feature space, and setting the maximum number of iterations and the initial adaptive jump step size , define a domain-specific fitness function for the microcrack characteristics of low-altitude aircraft Used to evaluate the feature matching degree and microcrack region coherence of candidate microcrack regions: ； in, Represents discrete bitterfish individuals The set of candidate microcrack region indices contained in , To reflect the candidate microcrack area Scoring functions for microcrack morphology, size, microcrack edge gradient characteristics, and texture orientation, To describe the penalty function of inconsistency and redundancy of candidate microcrack regions within discrete bitterfish individuals, and are positive weighting coefficients respectively; S43. Adopting the adaptive discrete jump update mechanism to optimize each discrete bitter fish individual in the discrete bitter fish population, based on the optimal discrete bitter fish individual of the current generation. With any discrete bitterfish individual The difference between Update discrete bitterfish individuals: ； in, represents the candidate microcrack region combination corresponding to the mth discrete bitterfish individual in the tth iteration, represents the updated combination of candidate microcrack regions corresponding to discrete bitterfish individuals, Represents a bitwise exclusive OR operation, For the The jump step size is adaptively adjusted according to the feedback of micro-crack characteristic data of low-altitude aircraft. At the same time, the binary difference between the current optimal solution and the discrete bitterfish individuals to be updated is considered; S44. Perform local search on the updated discrete bitter fish individuals to further improve the feature consistency of the candidate microcrack region, wherein the local search performs fine-grained perturbation on the binary representation of the discrete bitter fish individuals, uses a neighborhood search operator to fine-tune some bits, and calculates the feature consistency of the candidate microcrack region according to the fitness function. The feedback is corrected in real time, so that the binary representation of the candidate microcrack region accurately reflects the actual distribution and structural characteristics of the microcracks in the low-altitude aircraft; S45. Repeat S43 and S44 until the maximum number of iterations is reached After that, the fitness function is selected from the final population The discrete bitterfish individual set with the highest value constitutes the preliminary candidate set for aircraft microcrack identification , each candidate discrete bitterfish individual in the preliminary candidate aircraft microcrack identification set It corresponds to the binary representation of the potential microcrack area in the low-altitude aircraft structure, and comprehensively reflects the distribution, coherence and morphological characteristics of microcracks in the microcrack area; S5. Apply a multimodal distribution gravity algorithm to perform global fusion optimization on the preliminary candidate aircraft microcrack identification set. By simulating the attraction of multimodal distribution in the gravitational field, multiple candidate microcrack regions are integrated and screened, and the globally optimized final aircraft microcrack identification result is output. The S5 comprises the following steps: S51. For each candidate discrete bitterfish individual in the preliminary candidate aircraft microcrack identification set The gravitational mass model is constructed based on the structural characteristics of the microcrack area of the corresponding low-altitude aircraft, and the gravitational mass of each candidate discrete bitterfish individual is defined. : ； in, Candidate discrete bitterfish individuals The fitness function of is the total number of individuals in the candidate set; S52. Construct a multi-peak gravitational field between each candidate discrete bitter fish individual in the candidate set, and calculate the candidate discrete bitter fish individual through multi-peak gravitational calculation. and candidate discrete bitterfish individuals The gravitational force between: ； in, is the gravitational constant, Represents a candidate discrete bitterfish individual and candidate discrete bitterfish individuals The discrete distance measure between To prevent small positive numbers from dividing by zero; S53. For each candidate discrete bitterfish individual The comprehensive gravitational vector is calculated by weighted aggregation based on the gravitational forces exerted by all other candidate discrete bitterling individuals: ； And based on the comprehensive gravitational vector The candidate discrete bitterfish individuals are clustered and grouped according to the direction and amplitude of the gravitational force, and the candidate discrete bitterfish individuals with similar gravitational force and reflecting the coherence of the microcrack regional distribution are classified into the same cluster; S54. Screen the representative candidate discrete bitterfish individuals with the best gravity vector aggregation effect in each cluster to form the final identification set of aircraft microcracks after global fusion optimization ,Each candidate discrete bitterfish individual in the final identification set of aircraft microcracks can comprehensively reflect the true distribution, coherence and structural characteristics of the microcrack area in the low-altitude aircraft structure, and the output result is used as the final identification result of low-altitude aircraft microcracks; S6. Map the final identification results of aircraft microcracks to the three-dimensional structural model of the low-altitude aircraft, and construct a microcrack structural parameter model based on the mapping results. Quantify the specific location, size, shape, direction and expansion trend of the aircraft microcracks to form an aircraft microcrack quality assessment report.

2. The method for evaluating the quality of low-altitude technology products based on artificial intelligence according to claim 1, characterized in that: Said S1 comprises the following steps: S11. Use strain sensors, ultrasonic detection devices, infrared imagers, and micro-vibration detectors to collect raw structural signal data from key structural parts of low-altitude aircraft. The collected raw structural signal data includes strain response raw data, ultrasonic detection raw image data, infrared thermal imaging raw data, and micro-vibration raw signal data. S12. Digitally process the collected raw structural signal data and convert them into strain response signal data, ultrasonic detection image data, infrared thermal imaging data, and micro-vibration signal data; S13. The strain response signal data , ultrasonic detection image data , infrared thermal imaging data and micro-vibration signal data According to the predetermined data fusion rules, time and space synchronization and format unification are performed to build the aircraft structure detection data set : 。 3. The method for evaluating the quality of low-altitude technology products based on artificial intelligence according to claim 1, characterized in that: The S2 comprises the following steps: S21. Aircraft structure detection dataset The noise of various detection data in the dataset is filtered out respectively. A sliding window filter is used to perform time domain noise reduction on the strain response signal data and micro-vibration signal data. A spatial domain high-pass filter is performed on the ultrasonic detection image data and infrared thermal imaging data to form a preliminary noise-reduced aircraft detection data set. S22. Remove outliers from the aircraft detection dataset after preliminary noise reduction, remove local outliers in various detection data based on the triple standard deviation method, and construct an outlier-free aircraft detection dataset; S23. Perform data normalization on the abnormal aircraft detection data set and scale all detection data to a uniform range. , get the normalized aircraft detection dataset; S24. Perform feature enhancement processing on the normalized aircraft detection dataset, extract multi-scale frequency components and short-time energy features from the strain response signal data and micro-vibration signal data, perform edge enhancement and texture gradient enhancement on the ultrasonic detection image data and infrared thermal imaging data, and construct the final preprocessed aircraft detection dataset. .

4. The method for evaluating the quality of low-altitude technology products based on artificial intelligence according to claim 1, characterized in that: The S3 includes the following steps: S31. Based on the final pre-processed aircraft detection dataset Perform multi-dimensional feature extraction operations on various types of data, and extract signal energy distribution characteristics from strain response signal data and micro-vibration signal data , frequency component characteristics And envelope change characteristics ; S32. Extracting microcrack edge gradient features from ultrasonic testing image data and infrared thermal imaging data and texture direction characteristics ; S33. Set signal features and image feature set After unified standardization, the vectors are spliced into the feature vectors for aircraft microcrack detection. , among which The feature vector corresponding to the detection area is expressed as: ； in, 、 、 、 、 Respectively represent The characteristic values of signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics and texture direction characteristics in each region; S34. Construct a set of feature vectors extracted from all detection areas to form a set of feature vectors for aircraft microcrack detection: ； in, is the total number of detection areas.

5. The method for evaluating the quality of low-altitude technology products based on artificial intelligence according to claim 4, characterized in that: The signal energy distribution characteristics are used to measure the response strength of the structure in the detection area under stimulation. The higher the energy, the deeper the crack penetration or the more significant the stress concentration. The frequency component characteristics reflect the dynamic vibration behavior of the crack area, and the frequency drift or abnormal frequency point reveals the stiffness change around the crack; The envelope variation feature represents the non-stationarity of the signal's time domain envelope and is suitable for identifying high mutation areas at the microcrack tip; The microcrack edge gradient characteristics reflect the clarity of the crack boundary and the degree of edge mutation; The texture direction feature describes the crack propagation path regularity on the surface of the structure and is used to determine whether the crack propagates along the principal stress direction of the structure.

6. The method for evaluating the quality of low-altitude technology products based on artificial intelligence according to claim 1, characterized in that: The S54 includes the following steps: S541. Based on the final identification set of aircraft microcracks after fusion optimization, each candidate discrete bitterfish individual in the final identification set of aircraft microcracks is The corresponding binary representation is mapped to the original structure detection dataset The spatial position of the microcrack area in the three-dimensional structural model , and associated with its feature vector set for aircraft microcrack detection The corresponding eigenvector in ; S542. Constructing a comprehensive scoring function for structural attributes The signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics and texture direction characteristics are jointly weighted evaluated with the individual optimization results: ； in, is the weighting coefficient; S543. Comprehensive scoring function based on structural attributes The microcrack risk level determination rule is constructed based on the combined interval of signal energy distribution characteristics, frequency component characteristics, envelope change characteristics, microcrack edge gradient characteristics, and texture direction characteristic parameters, and the classification judgment is performed in combination with the following characteristic indicators and physical meanings: If any two of the following are met: : The crack edge has a significant mutation; : Crack response energy is high; : The envelope discontinuity is strong; : High overall score; It is judged as a high-risk microcrack area, indicating that the structure has deep cracks, high stress areas or high growth rate risks; If only one of the above is met, and at the same time: 、 , it is determined to be a medium-risk microcrack area, indicating that the crack is in the early development stage or the expansion direction is unclear; If satisfied: 、 、 、 , it is judged as a low-risk microcrack area with a basically intact structure and only slight surface disturbances or background noise artifacts; in, is the high risk judgment threshold, is the low risk judgment threshold, is the boundary of the middle judgment interval; S544. Construct the final identification result set of aircraft microcracks: ； in, is the final risk level label.

Images