Foamed aluminum insulation detection method based on eddy current effect

Abstract

The application relates to the technical field of foam aluminum detection, and discloses a foam aluminum insulation detection method based on eddy current effect. The method divides a foam aluminum sample to be detected into multiple detection units, collects eddy current signal data, material structure data and environmental parameter data of each unit to form a multi-source detection data set. Then, insulation defect evaluation indexes and material stability evaluation indexes are constructed according to the multi-source detection data set, and an initial defect value is generated through a preset threshold mechanism. The initial defect value is dynamically corrected based on the material correlation of adjacent detection units to obtain a corrected defect value. Finally, a priority list is generated according to the corrected defect value sorting, and an insulation performance evaluation result is output in combination with a visualization tool. The method realizes comprehensive and accurate detection of the insulation performance of foam aluminum, provides a systematic and effective technical scheme for identification and evaluation of foam aluminum insulation defects, and is suitable for the insulation performance detection scene in the foam aluminum production and application process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of foam aluminum detection, in particular to a foam aluminum insulation detection method based on eddy current effect. BACKGROUND

[0002] As a new type of material with both structural performance and functional characteristics, foam aluminum has been widely used in many fields such as construction, rail transit, aerospace, etc. due to its lightweight, high strength, good heat and sound insulation performance. In these application scenarios, the insulation performance of foam aluminum is directly related to the safety and stability of the overall structure. Once insulation defects occur, it may cause a series of problems such as energy loss, equipment failure and even safety accidents. Therefore, it is crucial to accurately and efficiently detect the insulation performance of foam aluminum.

[0003] There are various methods for detecting the insulation performance of foam aluminum. Among them, eddy current detection technology is commonly used in metal material defect detection due to its non-contact and fast response characteristics. However, the existing detection methods based on eddy current effect have many limitations. Traditional eddy current detection often only focuses on single eddy current signal data, ignoring the material structure characteristics of foam aluminum itself and the influence of environmental factors during the detection process. The porous structure distribution, pore size, porosity and other material structure parameters of foam aluminum will directly affect the propagation characteristics of eddy current, and changes in temperature, humidity and other environmental parameters may also interfere with the collection accuracy of eddy current signals. Relying solely on single signal data for detection can easily lead to false positives or false negatives of defects.

[0004] The existing detection methods lack systematicity in the division of detection units, and often use a relatively rough division method, which fails to make fine division according to the actual structural characteristics of foam aluminum samples, resulting in low spatial resolution of the detection results and difficulty in accurately locating small defects. In the defect evaluation stage, the existing methods usually judge the detection signals based on fixed empirical thresholds, and the generated defect values lack consideration of the overall stability of the material and the correlation of the material characteristics of adjacent regions. Foam aluminum material has a certain overall stability, and the structure and performance of adjacent regions often have mutual influence, and ignoring this correlation will greatly reduce the accuracy of the defect evaluation results.

[0005] The output form of the existing detection results is relatively single, mostly in the form of data tables or simple curves, lacking intuitive visual display, which is not convenient for detection personnel to quickly identify key defect areas and evaluate overall insulation performance. For large foam aluminum components, the amount of detection data is large, and the traditional result output method is difficult to meet the needs of efficient analysis and decision-making. The existence of these problems makes the existing detection methods unable to achieve ideal results in terms of defect detection rate, positioning accuracy and evaluation comprehensiveness of foam aluminum insulation defects, and it is difficult to meet the high standard requirements of foam aluminum insulation performance detection in actual applications. SUMMARY

[0006] The present application aims to provide a method for detecting the insulation of foamed aluminum based on the eddy current effect to solve the problems raised in the background.

[0007] To achieve the above-mentioned purpose, the present application provides a method for detecting the insulation of foamed aluminum based on the eddy current effect, which comprises:

[0008] Divide the foamed aluminum sample to be detected into multiple detection units, and collect the eddy current signal data, material structure data and environmental parameter data of each detection unit to form a multi-source detection data set;

[0009] Construct an insulation defect evaluation index and a material stability evaluation index according to the multi-source detection data set, and generate an initial defect value through a preset threshold mechanism;

[0010] Dynamically correct the initial defect value based on the material correlation of adjacent detection units to generate a corrected defect value;

[0011] Generate a priority list according to the corrected defect value ranking, and output the insulation performance evaluation result in combination with a visualization tool.

[0012] Preferably, the division rule of the detection unit is as follows:

[0013] Based on the geometric characteristics and material distribution boundary of the foamed aluminum sample, the sample is divided into grid units with uniform area, each grid unit corresponds to an independent detection area, and the grid unit is taken as the detection unit.

[0014] Preferably, the construction process of the insulation defect evaluation index is as follows:

[0015] Extract the signal amplitude variation feature from the eddy current signal data;

[0016] Obtain the porosity and density parameters from the material structure data, and calculate the structure defect parameters through a weight distribution model;

[0017] Generate a defect severity index according to the relationship between the structure defect parameters and the preset standard threshold value;

[0018] Extract the temperature influence factor from the environmental parameter data, and calculate the insulation defect evaluation score value in combination with the defect severity index and the temperature influence factor.

[0019] Preferably, the construction process of the material stability evaluation index is as follows:

[0020] Extract the strength coefficient and uniformity from the material structure data, and perform normalization processing;

[0021] Weighted fusion calculation is performed on each parameter after normalization processing to obtain the material stability evaluation score value;

[0022] When the uniformity falls below a set threshold, the material stability assessment score is updated, and the update rules are as follows:

[0023] If the proportion of consecutive monitoring times with uniformity below the set threshold exceeds the limit of the total number of monitoring times, a preset deduction ratio will be triggered, and the material stability assessment score will be deducted according to the deduction ratio.

[0024] Otherwise, the gradient interval is divided according to the difference between the uniformity and the set threshold, and deduction is made in an increasing proportion, with the upper limit of deduction being the preset deduction proportion.

[0025] Preferably, the process for generating the initial defect value is as follows:

[0026] Set early warning thresholds for insulation defect assessment indicators and material stability assessment indicators;

[0027] If the evaluation index of any detection unit is lower than the set warning threshold, the initial defect value of that detection unit will be assigned the preset highest defect level.

[0028] If all evaluation indicators are higher than the set warning threshold, the initial defect value is calculated based on the insulation defect evaluation indicator and the material stability evaluation indicator using the attenuation function.

[0029] Preferably, the execution steps of the dynamic correction include:

[0030] Select a detection unit as the target detection unit and obtain the initial defect values ​​of all its adjacent detection units;

[0031] Based on the material similarity and spatial distance between the target detection unit and its adjacent detection units, spatial correlation index values ​​are generated.

[0032] Functional compatibility index values ​​are generated by matching the material types of the target detection unit and adjacent detection units.

[0033] Adjustment weights are assigned based on spatial correlation index values ​​and functional compatibility index values;

[0034] The initial defect values ​​of the target detection unit are weighted and adjusted based on the corrected weights to generate corrected defect values;

[0035] The correction is completed by traversing all detection units, and the corrected defect value of each detection unit is output.

[0036] Preferably, the process of generating the spatial correlation index value includes:

[0037] The compositional difference degree is calculated based on the material composition differences between the target detection unit and adjacent detection units;

[0038] The proximity is calculated based on the geometric distance between the target detection unit and its adjacent detection units;

[0039] The basic spatial correlation index value is obtained by weighted fusion of the comprehensive component difference and proximity, and a compensation factor is set according to the connection type of adjacent detection units;

[0040] The basic spatial correlation index values ​​are adjusted by a compensation factor to generate the final spatial correlation index values.

[0041] The rules for setting the compensation factor are as follows:

[0042] If the connection type is direct contact, a first compensation factor value is assigned; if the connection type is indirect interval, a second compensation factor value is assigned, and the first compensation factor value is greater than the second compensation factor value.

[0043] Preferably, the process of generating the functional compatibility index value includes:

[0044] Match the dominant material type of adjacent detection units with the material type of the target detection unit;

[0045] If the matching result belongs to a preset compatible type combination, the function compatibility index value is calculated based on each material type and preset weight;

[0046] If the combination does not belong to a compatible type, the correlation degree is calculated according to the material compatibility specification, and the correlation degree is converted into a functional compatibility index value through a mapping function.

[0047] Preferably, the calculation process for the correlation degree is as follows:

[0048] Obtain a historical dataset of aluminum foam samples and statistically analyze the co-occurrence frequency of the target detection unit and the corresponding material types of adjacent detection units.

[0049] The material query function depends on the weight table to obtain the initial weight values ​​of the material types of the target detection unit and adjacent detection units.

[0050] Calculate the compatibility coefficient according to the compatibility rules for aluminum foam materials;

[0051] Adjust the initial weight value of the function dependency based on the compatibility coefficient, and generate the corrected weight value of the function dependency.

[0052] The comprehensive correction function relies on weight values ​​and co-occurrence frequencies to calculate the correlation.

[0053] Preferably, the allocation process of the corrected weights is as follows:

[0054] Obtain the detection target category of the aluminum foam sample, and set spatial correlation weight and functional compatibility weight according to the detection target category;

[0055] The corrected weights are calculated based on the set weights, the final spatial correlation index value, and the functional compatibility index value.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] The eddy current effect-based method for testing aluminum foam insulation exhibits numerous advantages through multi-dimensional technical optimization. In the data acquisition phase, this method divides the aluminum foam sample into multiple detection units and simultaneously collects eddy current signal data, material structure data, and environmental parameter data from each unit, forming a multi-source detection dataset. This multi-source data fusion approach overcomes the limitations of traditional detection methods that rely solely on a single eddy current signal, comprehensively capturing various factors affecting the insulation performance of aluminum foam. Eddy current signal data directly reflects changes in the material's internal electromagnetic properties; material structure data, such as pore distribution and pore size, reveals the material's physical basis; and environmental parameter data eliminates interference from external conditions. The combination of these three factors provides richer and more reliable information support for subsequent defect assessment.

[0058] In the defect assessment index construction and initial defect value generation stages, the method simultaneously constructs insulation defect assessment indices and material stability assessment indices based on multi-source detection datasets, and generates initial defect values ​​through a preset threshold mechanism. Compared to traditional methods that only focus on the simple judgment of the presence or absence of defects, this approach focuses on both the specific characteristics of insulation defects and the overall stability of the material. This ensures that the generated initial defect values ​​not only reflect the local defect situation but also correlate with the overall performance of the material, laying a more scientific foundation for subsequent evaluations.

[0059] Another significant advantage of this method is its dynamic correction of initial defect values ​​based on the material correlation between adjacent detection units. The structure of aluminum foam is continuous, and the material properties and defect distribution of adjacent units often influence each other. Traditional methods, which ignore this correlation, can easily lead to biased defect assessments. However, by analyzing the correlation between adjacent units in terms of material structure and eddy current signals, and dynamically adjusting the initial defect values, errors caused by local interference factors can be effectively eliminated. This results in corrected defect values ​​that more closely reflect the actual defect conditions, improving the accuracy and reliability of defect assessment.

[0060] The method generates a priority list based on the corrected defect values ​​and outputs insulation performance evaluation results using visualization tools, further enhancing the practicality of the detection method. The priority list sorts detected defect areas according to severity, allowing inspectors to quickly focus on key problem areas and improving detection efficiency. The visualization tools transform complex detection data into intuitive images or charts, clearly displaying the location, distribution, and severity of defects, making the results easier to understand and analyze. This provides a more intuitive and effective reference for both quality control in the production process and maintenance during use. Attached Figure Description

[0061] Figure 1 This is a schematic diagram illustrating the working principle of the foamed aluminum insulation detection method based on eddy current effect described in this invention.

[0062] Figure 2 A flowchart for constructing insulation defect assessment indicators;

[0063] Figure 3 A flowchart for constructing material stability assessment indicators;

[0064] Figure 4 This is a flowchart illustrating the generation process of spatial correlation index values. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Please see Figure 1 This invention provides a method for detecting aluminum foam insulation based on the eddy current effect, the method comprising:

[0067] This method achieves accurate evaluation of the insulation performance of aluminum foam materials through multi-source data fusion and dynamic correction mechanisms. The aluminum foam samples to be tested are divided into several detection units through gridding. Each unit simultaneously collects eddy current signals, material structural parameters, and environmental parameters. Based on the collected data, insulation defect evaluation indicators and material stability evaluation indicators are constructed, and initial defect values ​​are generated through a preset threshold mechanism. A dynamic correction model is established using the material correlation characteristics between adjacent units to spatially weight and adjust the initial defect values, ultimately generating corrected defect values ​​and forming a priority evaluation list. This method uses visualization tools to output the test results, achieving a multi-dimensional evaluation of the insulation performance of aluminum foam.

[0068] Example 1: See Figure 2The detection unit division process is based on the physical properties and structural features of the aluminum foam sample. The meshing process employs an adaptive algorithm, automatically adjusting the mesh density according to the actual size and shape of the sample. For regularly shaped samples, an orthogonal meshing method is used, with mesh lines evenly distributed along the length and width directions. When processing irregularly shaped samples, the system first identifies the sample outline boundary, generating a structured mesh within the boundary, supplemented by an unstructured triangular mesh in the boundary region. The mesh size is determined by comprehensively considering detection accuracy requirements and data processing efficiency; in typical applications, the side length of a single mesh unit is controlled within the range of 3-7 cm. The system has a built-in intelligent recognition module that can automatically detect abrupt material transitions. When a significant material difference exists within a single mesh, a mesh subdivision mechanism is triggered, further dividing the unit into finer sub-mesh. This layered processing method ensures detection resolution while avoiding unnecessary waste of computational resources.

[0069] The construction of insulation defect assessment indicators is a complex process involving the fusion of multiple parameters. The eddy current signal processing stage employs digital signal analysis technology to reduce noise and extract features from the original acquired signals. Signal preprocessing includes three main steps: baseline correction, power frequency interference cancellation, and random noise filtering. The feature extraction stage focuses on the time and frequency domain characteristics of the signal, with a particular emphasis on analyzing the degree of waveform distortion and harmonic component changes. Material structural parameters are obtained using non-destructive testing methods, reconstructing the internal structure of the sample through 3D imaging technology. Porosity calculation is based on image processing algorithms, performing binarization and connected component analysis on the scanned tomographic images. Density parameter measurement combines sample geometric dimensions and mass data to calculate the average density value of each grid cell. The environmental parameter monitoring system records temperature, humidity, and other data in real time during the testing process, establishing a correlation model between environmental parameters and material properties.

[0070] The calculation of structural defect parameters employs a multi-index weighted fusion method. The system maintains a configurable parameter weight database, selecting appropriate weight combinations based on different application scenarios. In a typical insulation performance testing scenario, the weight of the porosity parameter is set to 0.6, and the weight of the density parameter is 0.4. The weight allocation model supports dynamic adjustment; when special material combinations or abnormal environmental conditions are detected, the system automatically calls upon backup weight schemes. The generation of the defect severity index is based on a graded comparison mechanism, matching the calculated structural defect parameters with preset standard value ranges. The standard value ranges are defined according to material specifications and industry standards, and are divided into five levels: excellent, acceptable, minor defects, moderate defects, and severe defects. The index range corresponding to each level is used to calculate the accurate defect severity index through linear interpolation.

[0071] The calculation of the temperature influence factor takes into account the time-varying characteristics of ambient temperature. The system records the temperature change curve during the detection process and extracts characteristic parameters such as the maximum temperature difference and the rate of temperature change. The temperature correction model establishes a quantitative relationship between material performance parameters and temperature changes, quantifying the impact of temperature fluctuations on the detection results. For aluminum foam samples of different materials, the system uses differentiated temperature correction coefficients, which are determined through extensive basic experiments and stored in the material property database. The final calculation of the insulation defect assessment score uses a multi-layer neural network model. This model has been trained with a large amount of sample data and can accurately reflect the nonlinear relationship between various parameters. The neural network input layer contains three types of data: signal feature parameters, structural defect parameters, and temperature influence factor. The hidden layer adopts a double hidden layer structure, and the output layer generates an assessment score of 0-100. The model parameters are updated regularly to maintain the accuracy and reliability of the assessment results.

[0072] A special algorithm is used for boundary handling during the detection unit generation process. When the mesh boundary coincides with the material physical boundary, the system automatically marks it as a boundary unit and activates the boundary detection mode. The boundary detection mode employs a higher precision sampling frequency and a more rigorous analysis algorithm to ensure the detection accuracy of the boundary region. For material transition regions with gradient changes, the system adopts a progressive mesh generation strategy, gradually refining the mesh in the transition region to achieve a smooth transition in detection resolution. This approach effectively avoids measurement errors caused by boundary effects.

[0073] The eddy current signal acquisition system employs a multi-band excitation method, automatically selecting the optimal detection frequency based on material thickness and expected defect type. The signal acquisition probes utilize an array design, with each grid cell corresponding to an independent detection channel, enabling parallel data acquisition. The probe spacing and arrangement are optimized to ensure comprehensive detection coverage while minimizing signal interference between adjacent channels. The acquired raw signals undergo analog front-end processing, including signal amplification, filtering, and analog-to-digital conversion, ultimately converting them into digital signals for subsequent analysis.

[0074] Multimodal imaging technology was used to acquire material structure data. An X-ray imaging system provided high-resolution 3D images of the sample's internal structure, while an optical surface scanning system recorded the sample's surface morphology. The two data sources were precisely aligned using a coordinate registration algorithm to construct a complete material structure model. Image processing algorithms automatically identified structural anomalies, including typical defects such as uneven pore distribution, cracks, and inclusions. Voxel analysis was used to calculate structural parameters, performing voxel-by-voxel processing on the 3D image data and statistically analyzing the material property parameters within each mesh cell.

[0075] The environmental monitoring system employs a distributed sensor network, deploying multiple monitoring points across the detection area to collect environmental parameters such as temperature and humidity in real time. Sensor data is wirelessly transmitted to a central processing unit for time synchronization and data fusion. The spatial distribution model of environmental parameters is constructed using the Kriging interpolation method, calculating the distribution of environmental parameters across the entire detection area based on measurement data from discrete monitoring points. This processing approach fully considers the spatial variability of environmental parameters, providing accurate foundational data for subsequent temperature-related corrections.

[0076] The calculation process for insulation defect assessment indicators incorporates a quality control mechanism. The system monitors the quality status of each input parameter in real time, and automatically triggers a data verification process when data anomalies or deviations from reasonable ranges are detected. The verification process includes three stages: sensor calibration check, data acquisition process review, and manual confirmation. After the assessment results are generated, the system performs a consistency check, comparing the current results with historical data to identify potential anomalies. For questionable results, the system automatically schedules a re-inspection to ensure the final output assessment results are accurate and reliable. The entire processing flow adopts a modular design, with each functional module relatively independent yet collaborative, exchanging data through standard interfaces, facilitating system maintenance and functional expansion.

[0077] Example 2: See Figure 3 The construction of material stability assessment indices is based on the mechanical properties and microstructure characteristics of aluminum foam. The strength coefficient was obtained using standardized mechanical testing methods, with test samples precisely cut from the aluminum foam sample according to the grid cell positions. Complete load-displacement curves were recorded during testing, from which characteristic parameters of the elastic and plastic deformation stages were extracted. The elastic modulus was calculated using the slope value of the initial linear segment of the curve, and the yield strength was determined using the offset method, with the stress value at which a specific plastic strain was generated as the criterion. Test data for each grid cell was recorded independently, establishing a complete mechanical property distribution map.

[0078] The uniformity parameter was measured using a multi-scale analysis method. At the macroscopic scale, optical surface scanning was used to acquire the morphological features of the sample surface; at the mesoscopic scale, X-ray tomography was used to reconstruct the three-dimensional pore structure; and at the microscopic scale, scanning electron microscopy was used to observe the microscopic morphology of the pore walls. Image processing algorithms comprehensively analyzed the acquired multi-scale images to calculate the uniformity index of the pore distribution. This index comprehensively considers multiple dimensions of features such as pore size, shape, orientation, and spatial distribution, and is converted into a normalized value within the range of 0-1 using statistical algorithms. A sliding window technique was used in the data processing to calculate local uniformity at multiple scales, and then a weighted average was used to obtain the overall uniformity score.

[0079] The normalization stage standardizes the original test data. Different normalization benchmarks are used for the strength coefficient and evenness parameter. The strength coefficient is referenced to the test results of standard samples from the same batch, while the evenness parameter is based on an ideal uniform distribution model. The transformation process preserves the distribution characteristics of the original data and eliminates dimensional differences between different parameters. The normalized data is stored in a three-dimensional matrix, with the dimensions corresponding to the spatial coordinates of the sample and the parameter type, respectively.

[0080] The weighted fusion calculation employs a dynamic weight allocation strategy. The system has multiple pre-set weight combination schemes, automatically selecting the most suitable scheme based on the testing objective and material type. In the standard testing mode, the base weights for strength coefficient and uniformity are set to 0.6 and 0.4, respectively. Weight allocation considers the correlation between parameters to avoid redundant calculations of the same features. A nonlinear correction factor is introduced into the fusion calculation process; when a parameter value exceeds the normal range, its weight ratio is automatically adjusted. The calculation results are smoothed to eliminate abnormal scores caused by local fluctuations.

[0081] Uniformity threshold monitoring employs a real-time comparison mechanism. The system continuously tracks uniformity changes in each detection unit, triggering an early warning signal when the value falls below a preset threshold. The threshold setting considers material specifications and the usage environment, typically using the lower quartile of historical data statistical distribution. The monitoring process utilizes moving average technology to reduce false alarms caused by instantaneous fluctuations. After an early warning signal is triggered, the system automatically records the time, location, and relevant parameters of the event, providing complete data for subsequent analysis.

[0082] The number of consecutive monitoring sessions is calculated using a sliding time window method. The system maintains a fixed-length monitoring record queue to record the uniformity status of the most recent few tests. The queue length is dynamically adjusted based on the testing frequency and material characteristics, typically ensuring sufficient coverage for a complete production batch. During statistical calculations, the system scans the entire queue, calculating the frequency and duration of low uniformity events. When the proportion of abnormal events exceeds a set percentage, the system classifies it as a persistent defect and initiates a scoring deduction process.

[0083] The deduction ratio is calculated using a tiered response mechanism. The system divides the deduction ratio into multiple levels based on the severity and duration of the abnormal event. Minor anomalies use a linear deduction model, with the deduction amount proportional to the duration of the anomaly. Severe anomalies use a stepped deduction model; when the anomaly persists beyond a critical time, the deduction ratio jumps to a higher level. Deduction operations are implemented by modifying the scoring matrix. The system retains the original scores and deduction records, supporting result traceability and review.

[0084] The gradient interval deduction mechanism is designed for short-term fluctuations. When the uniformity is slightly below the threshold but does not meet the criteria for sustained abnormality, the system divides the system into multiple gradient intervals based on the magnitude of the deviation. Each interval corresponds to a deduction ratio, with larger deviations resulting in higher deduction ratios. The deduction ratio increases using a non-linear curve, changing gradually near the threshold and drastically away from it. This design makes the scoring system insensitive to minor fluctuations but highly sensitive to significant anomalies.

[0085] The material stability assessment score is updated using a transaction processing model. Each score modification is treated as an independent transaction, comprising four steps: data reading, calculation, modification, and verification. Transaction processing ensures the atomicity and consistency of score updates, avoiding data errors caused by concurrent operations. The updated score takes effect immediately, simultaneously triggering the status flag update of the relevant detection units. The system periodically performs consistency checks on the score data, automatically initiating a data repair procedure when anomalies are detected.

[0086] The dynamic update mechanism relies on real-time data stream processing. Raw data collected by sensors is preprocessed before entering the streaming computing engine. The engine calculates various feature parameters in real time, compares them with thresholds, and triggers corresponding event processing logic. The stream processing employs a distributed architecture, supporting high-concurrency data throughput and low-latency response. Processing results are written to a time-series database, supporting efficient historical data querying and analysis.

[0087] The scoring system employs a multi-layered verification mechanism for fault tolerance. Each scoring calculation undergoes input data validity verification, including range checks, consistency checks, and integrity checks. Intermediate results are validated during the calculation process; any anomalies immediately halt processing and trigger an alarm. The final score output undergoes rationality verification, comparing it with adjacent units and historical data. The system also includes a rollback mechanism to restore the system to a previous stable state in the event of a serious error.

[0088] The visual interface displays the material stability assessment results in real time. The score for each testing unit is displayed in color-coded form on the sample's 3D model, forming an intuitive heatmap. Users can interactively view detailed score data and related parameters at any location. The system supports multiple view modes, including 2D slice view, 3D volume rendering view, and parametric curve view, to meet different analytical needs. The displayed content is automatically refreshed periodically to reflect the latest assessment status.

[0089] Data storage employs a tiered architecture. Raw test data is stored in a cache for real-time processing. Processed feature parameters are stored in a relational database, supporting complex queries. Historical evaluation results are archived to a time-series database, optimizing long-term storage and batch analysis. The storage system implements automatic data migration, moving data to the appropriate storage tier based on access frequency. Data backup uses an incremental strategy, periodically synchronizing changed data to the backup storage system.

[0090] System maintenance functions include automatic calibration and diagnostics. Regular sensor calibration procedures are executed to ensure data acquisition accuracy. The system status monitoring module continuously monitors the operating status of each component, issuing maintenance alerts when performance degradation or functional abnormalities are detected. The logging module records system operations and events in detail, supporting troubleshooting and performance analysis. The maintenance interface provides management functions such as system configuration, user management, and data cleanup to ensure long-term stable system operation.

[0091] The anomaly handling process comprises two levels: automated response and manual intervention. For foreseeable routine anomalies, the system automatically processes and records them according to preset rules. For complex anomalies, the system pauses automated processing and notifies operators for diagnosis and decision-making. Anomalies are categorized and managed, and a knowledge base is established to accumulate handling experience. Root cause analysis is performed after each anomaly handling to optimize the system and prevent similar problems from recurring.

[0092] Integration with other systems utilizes standardized interfaces. Basic data is exchanged with the production management system via a web service interface, communication with testing equipment is achieved through a message queue, and third-party analytical data is imported via a file interface. The interface design supports extensibility, allowing for easy integration of new data sources or functional modules. A data conversion module handles format differences between different systems, ensuring accurate information transmission. A security authentication mechanism controls interface access permissions to prevent unauthorized operations.

[0093] Example 3: The initial defect value generation process employs a multi-level decision-making mechanism, combining graded threshold judgment and continuous value calculation. Early warning thresholds are set based on material performance standards and statistical analysis of historical testing data, establishing a dynamic adjustment model. The early warning thresholds for insulation defect assessment indicators are divided into three levels according to material type: Class I materials have a threshold of 65 points, Class II materials have a threshold of 60 points, and Class III materials have a threshold of 55 points. The early warning threshold for material stability assessment indicators adopts a unified standard, set at 70% of the material's initial design value. The threshold comparison module monitors two indicators of each testing unit in real time; when either indicator falls below the corresponding threshold, the highest defect level marker is immediately triggered.

[0094] The attenuation function calculation is performed on detection units where all indicators meet the standards. A nonlinear mapping method is used to convert the evaluation indicators into defect values. This calculation process incorporates environmental correction factors and material property coefficients to reflect performance changes under actual working conditions. The attenuation function expression is as follows:

[0095] ;

[0096] in: This represents the initial defect value of the i-th detection unit. The score is used to assess insulation defects. The score is used to assess the stability of the material. This is the insulation defect weighting coefficient. For decay rate parameters, This represents the material stability weighting coefficient. and These represent the theoretical maximum and minimum values ​​of the insulation defect assessment score, respectively. and This indicates the extreme range of the material stability assessment score. Parameter values ​​are obtained by querying a material database, and the optimal parameter set is automatically matched for different material combinations.

[0097] The dynamic correction process is based on spatial correlation analysis and employs an iterative optimization algorithm to gradually adjust defect values. The neighborhood search window uses a variable-size design, with a basic 3×3 grid that automatically expands to a 5×5 grid when a material boundary or property abrupt change is detected. Correction calculations for each target detection unit are performed independently to avoid cross-interference. Correction weight allocation considers two factors: spatial correlation characterizes the continuity of material properties, and functional compatibility reflects the interaction mechanisms between different materials. Weight calculations employ normalization to ensure that the sum of the weights of all neighborhood units is 1.

[0098] Material similarity assessment is achieved through multi-feature comparison. Compositional similarity is calculated using spectral analysis data, comparing differences in the content of major elements; structural similarity is calculated based on pore distribution characteristic parameters, including average pore size and pore size variation coefficient; mechanical similarity compares the relative differences in elastic modulus and yield strength. Spatial distance calculation incorporates material anisotropy correction, employing differentiated distance metric coefficients along different directions. For aluminum foam samples with obvious orientation characteristics, the distance weight along the pore extension direction is reduced by 30%, while the weight in the perpendicular direction is increased by 20%.

[0099] Functional compatibility assessment employs a knowledge-based reasoning method. The system has a built-in database of interactive properties of aluminum foam materials, containing functional impact data for common material combinations. The assessment process first identifies the material type combinations of the target unit and neighboring units, querying the database to obtain a basic compatibility score. When encountering novel material combinations, a similarity reasoning engine is activated to infer potential compatibility based on the similarity of chemical composition and structural characteristics. The assessment results are divided into five levels, from fully compatible to severely incompatible, with a corresponding correction coefficient for each level.

[0100] The weighted adjustment process employs an incremental update strategy. The initial defect value serves as the base reference; the changes in defect values ​​of neighboring elements are multiplied by their respective weights and then summed to obtain the correction amount. The magnitude of the correction amount is affected by the material state; when the target element is in a critical state, the correction sensitivity increases by 30%. Boundary checks are performed after each correction to ensure the defect value remains within a reasonable range. The entire correction process is divided into two stages: coarse adjustment and fine adjustment. The coarse adjustment stage uses a larger step size to quickly approach the target value, while the fine adjustment stage reduces the step size to improve accuracy.

[0101] The iteration termination condition is set based on the trend of defect value changes. The system is considered to have reached a stable state when the maximum defect value change is less than a set threshold over three consecutive iterations. Convergence is monitored in real time during iteration, and the step size and weight allocation strategy are automatically adjusted when oscillations or divergence trends are detected. The system records complete iteration history data, supporting later analysis and algorithm optimization.

[0102] The priority list generation employs a multi-keyword sorting algorithm. The primary sorting criterion is the corrected defect value, while secondary sorting keywords include defect area, location importance, and environmental impact coefficient. The sorting process preserves the spatial relationship information of the original detection units, facilitating subsequent cluster analysis. The list output format supports multiple data interfaces, including text reports, XML structures, and JSON formats, meeting the integration needs of different systems.

[0103] The visualization output module employs layered rendering technology. The base layer displays the geometric model and mesh division of the sample, the intermediate layer overlays a defect value distribution cloud map, and the top layer annotates key defect areas and priority markers. Display parameters can be dynamically adjusted, including color mapping scheme, transparency settings, and annotation density. User interaction functions support operations such as region selection, profile viewing, and numerical querying, enabling multi-angle analysis of the detection results.

[0104] An anomaly handling mechanism is implemented throughout the entire process. During data acquisition, a reasonableness check is implemented to filter out obviously abnormal data; intermediate results are verified during the calculation process to prevent error accumulation; and consistency checks are performed during the output phase to ensure the reliability of the final result. For detected anomalies, the system automatically handles them according to preset strategies or submits them for manual intervention. All anomalies are logged in detail, including the time, location, type, and handling measures, forming a complete quality traceability chain.

[0105] The system integration architecture adopts a modular design, with each functional module communicating through standard interfaces. The data acquisition module is responsible for acquiring raw signals, the feature extraction module processes sensor data, the analysis and calculation module executes core algorithms, and the visualization module manages the display of results. Communication between modules uses a publish-subscribe pattern, achieving loose coupling and high cohesion. The system supports hot-swapping, allowing specific modules to be replaced or upgraded while the system is running.

[0106] Performance optimization measures include parallelizing computational tasks, caching data access, and preloading algorithm parameters. The parallel computing framework distributes the processing tasks of grid cells across multiple computing nodes, leveraging multi-core processors to improve throughput. The data caching mechanism stores frequently accessed parameters in memory, reducing disk I / O latency. Commonly used parameters and model data are preloaded during startup, shortening the initial response time. System resource usage is monitored in real time, and task scheduling strategies are dynamically adjusted.

[0107] Maintenance and management functions include system self-checks, data backups, and configuration management. Regular self-checks verify the operational status of each component and detect potential problems. Data backup strategies combine full backups and incremental backups to balance storage space and recovery efficiency. The configuration management interface supports parameter adjustments and algorithm selection to adapt to different testing needs. Operation logs record detailed system activities and user actions, supporting auditing and fault diagnosis.

[0108] The update and upgrade mechanism supports incremental improvements. Algorithm updates utilize version control, allowing rollback to previous stable versions. Database expansions maintain compatibility with the original structure, ensuring access to historical data. User interface improvements maintain consistency in operating habits, reducing relearning costs. The upgrade process is automated, with verification mechanisms ensuring the integrity and correctness of the upgrade. System maintenance status is clearly indicated to avoid accidental operations.

[0109] Example 4: See Figure 4The calculation process of the spatial correlation index value is illustrated using a batch of aerospace-grade aluminum foam samples as an example. The sample dimensions are 30cm × 30cm × 5cm, divided into 6×6 grid cells, each cell with a side length of 5cm. During the inspection, an abnormal signal was found in the cell in the third row and second column (marked as C3), requiring the calculation of its spatial correlation index value with adjacent cells. Adjacent cells include eight directly contacting cells (B2, B3, B4, C2, C4, D2, D3, D4) and six indirectly separated cells (A2, A3, A4, E2, E3, E4).

[0110] The compositional difference was calculated using data collected by X-ray fluorescence spectrometry, primarily analyzing the differences in the content of aluminum, silicon, and magnesium. Taking unit C3 and its directly contacting unit D3 as an example, the elemental content detection data for both units are shown in Table 1.

[0111] Table 1: The elemental content detection data for units C3 and D3 are as follows.

[0112]

[0113] When calculating the compositional difference, the weighted summation was performed according to the following weighting ratios: aluminum 0.6, silicon 0.3, and magnesium 0.1. The result was a compositional difference of 0.44 between the two units (range 0-1, with smaller values ​​indicating more similar compositions). The same calculation method was applied to the compositional comparison of unit C3 with other adjacent units to establish a complete compositional difference matrix.

[0114] The proximity calculation takes into account the actual spatial relationship. For directly contacting units, the physical distance is uniformly recorded as 1 unit; for indirectly separated units, the distance is calculated according to the actual interval distance. For example, if A3 and C3 are separated by two units, the distance is recorded as 3 units. The distance data is converted into a proximity value in the range of 0-1 using a Gaussian kernel function, where the proximity of directly contacting units is set to 0.9, one unit apart is 0.7, and two units apart is 0.5.

[0115] The basic spatial association index value was synthesized using a weighted average method. The weight for component difference was set to 0.6, and the weight for proximity was set to 0.4. Continuing with the C3-D3 unit pair as an example, the calculated basic index value was 0.6 × 0.44 + 0.4 × 0.9 = 0.624. The compensation factor was determined based on the connection type: a compensation factor of 0.95 was assigned to direct contact connections, and a compensation factor of 0.75 was assigned to indirect spacing connections. After compensation adjustment, the final spatial association index value for unit C3-D3 was 0.624 × 0.95 = 0.593.

[0116] The generation process of the functional compatibility index value focuses on the interaction characteristics between materials. Taking the matching of unit C3 (mainly an Al-Si-Mg alloy) and unit B3 (mainly an Al-Cu-Mn alloy) as an example, the system queries the material compatibility knowledge base and finds that these two alloy combinations belong to the "moderate compatibility" category, with a basic compatibility score of 70 points (out of 100). Further analysis shows a significant difference in the coefficients of thermal expansion between the two units, and the correlation correction coefficient calculated according to the material compatibility specification is 0.85. The final functional compatibility index value is 0.595, obtained by multiplying the basic score by the correction coefficient and normalizing.

[0117] The material compatibility knowledge base is built upon extensive experimental data and engineering experience. For aluminum foam, the system includes compatibility data for over 50 common alloy combinations, each scored across three dimensions: thermal properties, mechanical properties, and chemical stability. When encountering novel combinations, the system uses analogical reasoning based on the chemical properties and crystal structure similarities of the constituent elements. For example, for an unlisted Al-Si-Fe combination, the system interpolates and estimates based on known Al-Si-Cu and Al-Fe-Mn combination data.

[0118] The weighting process is adjusted in conjunction with the inspection targets. In this aerospace component inspection, the weight for structural integrity is set at 0.7, and the weight for insulation performance is set at 0.3. In the final weight calculation formula for spatial correlation index values, the basic weight accounts for 60%, and the inspection target adjustment coefficient accounts for 40%. After calculation, the spatial correlation weight for unit C3 is determined to be 0.66, and the functional compatibility weight is 0.34. These weight values ​​will be used in subsequent defect value correction calculations.

[0119] The dynamic correction process employs an iterative approximation method. Taking the initial defect value of element C3 as 0.65 as an example, the defect values ​​of its neighboring elements B3, C2, C4, and D3 are 0.72, 0.68, 0.61, and 0.70, respectively. Based on the calculated weights of each neighboring element, a weighted average is taken to obtain a correction amount of 0.12. The correction sensitivity coefficient is determined to be 1.2 based on the material state of element C3, ultimately generating a corrected defect value of 0.65 + 0.12 × 1.2 = 0.794. After this value is confirmed to be within a reasonable range through boundary checks, it is updated in the defect value matrix.

[0120] Connection type identification employs multi-sensor fusion technology. Direct contact connections are determined by combining contact resistance measurements and ultrasonic flaw detection data; a contact resistance below 10 μΩ and ultrasonic echo attenuation less than 3 dB are considered direct contact. Indirect gap connections use X-ray imaging to identify the characteristics of the intermediate transition layer, measuring the gap distance with an accuracy of 0.1 mm. In special connection cases, such as those with adhesive layers or transition alloy layers, the system initiates a dedicated analysis program to comprehensively evaluate the connection characteristics.

[0121] The spatial correlation analysis is visualized using a 3D heatmap. The sample model is based on grid cells, with each cell displaying a different color according to its spatial correlation index value, gradation from blue (low correlation) to red (high correlation). Direct connections are indicated by solid lines, while indirect connections are indicated by dashed lines. Users can rotate and zoom to observe the correlation between cells from any angle, and clicking on a specific connection line will display detailed index calculation data.

[0122] Anomaly handling employs a tiered response strategy. When a spatial correlation index value below 0.3 is detected, the system marks it as a weak connection; below 0.2, it is marked as a dangerous connection. For weak connection areas, the detection frequency and number of sampling points are increased; for dangerous connection areas, an immediate alarm is triggered and a shutdown for inspection is recommended. All anomaly connections are recorded with detailed information, including location coordinates, connection type, detection time, and handling measures, forming a complete quality tracking record.

[0123] System calibration procedures are performed regularly to ensure detection accuracy. The spatial positioning measurement system undergoes laser tracker calibration weekly, the component analysis system is calibrated daily using standard samples, and the connectivity characteristic detection system automatically runs diagnostic tests upon each power-on. Calibration data is recorded in a dedicated database; if the allowable error is exceeded, the relevant functional modules are automatically locked until addressed by technical personnel, at which point they can be reactivated.

[0124] Data management employs a tiered storage architecture. Raw detection data is stored in a high-speed storage array for real-time analysis; processed feature parameters are stored in a relational database to support complex queries; and long-term historical data is archived to a tape library to save storage space. Data migration strategies are executed automatically based on access frequency: data from the most recent week is retained in online storage, data from the last month is moved to near-line storage, and older data is archived offline.

[0125] Example 5: The correlation calculation process is based on a combination of historical data analysis and materials science principles. The historical aluminum foam sample dataset comes from ten years of accumulated testing records, containing over 1200 rigorously validated sample data sets. Each data set records complete material parameters, process conditions, and performance indicators, forming a multi-dimensional material characteristic space. The data preprocessing stage employs an automatic cleaning algorithm to remove obviously abnormal and redundant records, retaining representative sample data. Data storage uses a time-series database, supporting queries based on multiple conditions such as material type, production process, and testing time.

[0126] Co-occurrence frequency statistics employ a sliding window analysis method. The system scans historical databases for material type combinations of the target detection unit and adjacent detection units, counting the frequency of specific combinations. The statistical range is set according to testing requirements, typically including sample data from the last three years or the last 500 similar processes. Frequency calculations consider minor differences in material ratios; materials are considered to be of the same type when the content deviation of major elements is within ±5%. The statistical results are normalized and converted into co-occurrence probability values ​​within the range of 0-1, reflecting the frequency of material combinations.

[0127] The material function dependency weighting table is constructed using the analytic hierarchy process (AHP). The weighting system consists of three levels: the top level represents the material application objectives, including key performance indicators such as structural strength, thermal insulation, and corrosion resistance; the middle level comprises material characteristic parameters such as elastic modulus, thermal conductivity, and electrochemical potential; and the bottom level represents elemental composition and microstructural characteristics. The weights for each level are determined through a combination of expert scoring and regression analysis of actual test data. The weighting table is updated monthly to incorporate the latest research findings and engineering feedback. For novel material systems, the system automatically generates a temporary weighting scheme, which is then formally evaluated after sufficient data has been accumulated.

[0128] The compatibility coefficient calculation incorporates a fuzzy logic inference system. Input variables include key parameters such as the difference in thermal expansion coefficients between materials, electrode potential differences, and crystal structure matching degrees. Each parameter is divided into multiple fuzzy levels; for example, the difference in thermal expansion coefficients is divided into three levels: slight difference, moderate difference, and significant difference. The fuzzy rule base contains 128 empirical rules derived from materials science literature and engineering practice cases. The inference process employs the Mamdani fuzzy inference method, outputting a compatibility coefficient within the range of 0-1. The system supports manual adjustment of fuzzy rules and membership functions to adapt to the evaluation needs of special materials.

[0129] The correction function relies on an iterative optimization algorithm to generate weight values. Initial weight values ​​are retrieved from a weight table based on material type, and then adjusted according to the real-time calculated compatibility coefficient. The adjustment magnitude has a non-linear relationship with the compatibility coefficient; the weight decreases significantly when the compatibility coefficient is below 0.3, and increases slightly when it is above 0.7. The adjusted weight values ​​must meet normalization conditions, ensuring the sum of the weights for all performance indicators remains 1. The system records historical versions of each weight adjustment, supporting result backtracking and comparative analysis.

[0130] The final correlation calculation employs a hybrid model combining probability and rules. Co-occurrence frequency reflects the statistical regularity of material combinations, while functional dependence weights reflect performance requirements; these two are initially integrated through a weighted summation. The calculation results are then corrected using a compatibility coefficient to weaken the correlation score of incompatible combinations. The system sets correlation thresholds: combinations below 0.2 are marked as high-risk, and combinations above 0.8 are marked as preferred. A more refined nine-level scoring system is used for the intermediate value range to facilitate the differentiation of subtle differences between similar combinations.

[0131] The weight allocation process is revised to take into account the priority differences of the detection targets. The system maintains a detection target configuration library, containing parameter settings for common detection scenarios. In structural integrity detection scenarios, the spatial correlation weight is set to 0.7 by default; in insulation performance detection scenarios, the functional compatibility weight is increased to 0.45. Target category identification is automatically determined based on sample identification codes and process parameters. When multiple detection targets exist, a weight allocation strategy is adopted. Users can manually adjust the weight allocation scheme through the configuration interface to meet special detection needs.

[0132] The dynamic adjustment mechanism for spatial correlation weights responds to changes in real-time detection data. The system monitors the material property gradient between adjacent units and automatically increases the spatial correlation weights when abrupt changes are detected. The adjustment magnitude is proportional to the rate of performance change, with the maximum adjustment range controlled within ±0.15. A damping factor is introduced into the dynamic adjustment to prevent drastic fluctuations in weight values ​​from affecting computational stability. After each weight update, the system re-evaluates the correction results of the ten most recent detection units to verify the effectiveness of the adjustment scheme.

[0133] The calculation of functional compatibility weights takes into account material interface effects. For material combinations with obvious transition layers, the system analyzes the microstructural characteristics and element diffusion of the interface region to generate an interface quality score. This score is used as an additional factor in the calculation of functional compatibility weights; a high-quality interface can increase the weight value by up to 0.1. Interface analysis employs a specialized high-resolution detection mode, reducing the scan step size to 1 / 5 of conventional detection to ensure the capture of detailed features in the interface region.

[0134] The weight normalization process employs a soft maximization function. This function preserves the relative magnitudes of the weight values ​​while ensuring the sum is strictly equal to 1. The process retains three decimal places of precision to avoid the accumulation of rounding errors. The normalized weight matrix is ​​stored in a distributed cache for shared access by multiple computing nodes. Cache updates utilize a copy-on-write strategy, ensuring data consistency while improving concurrent access performance.

[0135] The real-time data processing pipeline employs an event-driven architecture. Upon entering the system, detection data triggers a series of processing events, including data verification, feature extraction, correlation calculation, and weight allocation. Each processing stage is implemented as an independent microservice, communicating via a message queue. Event processing results are written to a time-series database, while simultaneously updating a state snapshot in memory. This architecture supports high-throughput real-time data processing, with the processing latency of a single detection unit controlled to within 50 milliseconds.

[0136] Anomaly data handling employs a multi-layered verification mechanism. Raw detection data first undergoes range checks and consistency verification; anomaly data is then marked and transferred to a dedicated processing flow. During calculations, the reasonableness of intermediate results is assessed, and any anomalies immediately halt subsequent processing. The final correlation calculation result is compared with historical data to determine trends; deviations from the normal range trigger a review process. All anomaly events are recorded with detailed contextual information, including input data, processing steps, and system status, facilitating problem diagnosis.

[0137] The system's learning function continuously optimizes parameter settings. After each detection, the system compares the predicted results with the actual validation data and automatically adjusts the calculation parameters to reduce bias. The learning process employs a mini-batch gradient descent method, with parameter update magnitudes strictly controlled to avoid overfitting. Learning results are cross-validated before being applied to formal detection, ensuring system performance stability. Global parameter optimization is performed monthly, recalibrating all calculation models.

[0138] A robust user feedback mechanism improves the evaluation system. Testing personnel can provide feedback on the system-generated relevance scores, collected through standardized forms. An expert review committee regularly evaluates the feedback data, translating reasonable suggestions into system rules or parameter adjustments. A closed-loop management system ensures that every feedback item receives a clear status update and result notification. Valuable feedback cases are incorporated into a knowledge base as reference solutions for similar issues in the future.

[0139] The visualization analysis tool supports multi-dimensional data exploration. Correlation analysis results are displayed as heatmaps on the 3D sample model, along with cross-sectional views and detailed data tables. Users can interactively filter specific material combinations to view the detailed correlation calculation process and influencing factors. The visualization system supports simultaneous multi-view interaction, facilitating comparative analysis of correlation characteristics in different regions. All views support high-resolution export, meeting the needs of report preparation and academic research.

[0140] System maintenance and management are automated. Daily automatic database backups and log archiving, and weekly system health checks and performance optimizations. Maintenance tasks are scheduled to avoid peak testing periods. Critical components are supported with dual-machine hot standby, with primary and backup nodes synchronizing status data in real time and automatically switching in case of failure. Version upgrades support rolling updates to ensure uninterrupted testing services. All maintenance operations are logged in detail, meeting quality management system requirements.

[0141] Security control mechanisms protect core data and algorithms. User access is managed using role-based access control, with different levels of personnel having differentiated operational permissions. Data is encrypted during transmission and storage, and critical computing nodes are deployed in secure isolation zones. System operations are audited throughout, and modifications to important parameters require dual authentication. Regular security vulnerability scans and penetration tests are conducted to promptly remediate potential risks. Data backup employs a 3-2-1 strategy to ensure data recoverability in extreme circumstances.

[0142] Integration with other systems utilizes standardized interfaces. Basic data is exchanged with the production management system via REST API, real-time monitoring equipment data is received via MQTT protocol, and the material database is accessed through a dedicated data channel. The interface design adheres to the principle of loose coupling, ensuring that changes in internal data structures do not affect external systems. A data conversion layer handles semantic differences between different systems, ensuring accurate information transmission. Interface calls implement flow control and fault isolation to prevent cascading failures.

[0143] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0144] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting aluminum foam insulation based on eddy current effect, characterized in that, The method includes: The aluminum foam sample to be tested is divided into multiple testing units, and eddy current signal data, material structure data and environmental parameter data of each testing unit are collected to form a multi-source testing dataset. Based on the multi-source detection dataset, insulation defect assessment index and material stability assessment index are constructed, and initial defect values ​​are generated through a preset threshold mechanism. The initial defect value is dynamically corrected based on the material correlation between adjacent detection units to generate a corrected defect value. A priority list is generated based on the corrected defect values, and the insulation performance evaluation results are output using visualization tools. The process of constructing the insulation defect assessment index is as follows: Extracting signal amplitude variation characteristics from eddy current signal data; Porosity and density parameters are obtained from material structure data, and structural defect parameters are calculated using a weighted allocation model. A defect severity index is generated based on the relationship between structural defect parameters and preset standard thresholds. Temperature influence factors are extracted from environmental parameter data, and insulation defect assessment scores are calculated by combining the defect severity index and temperature influence factors. The process for constructing the material stability assessment index is as follows: Strength coefficients and uniformity are extracted from material structure data and then normalized. The material stability assessment score is obtained by weighted fusion calculation of the normalized parameters. When the uniformity falls below a set threshold, the material stability assessment score is updated, and the update rules are as follows: If the proportion of consecutive monitoring times with uniformity below the set threshold exceeds the limit of the total number of monitoring times, a preset deduction ratio will be triggered, and the material stability assessment score will be deducted according to the deduction ratio. Otherwise, the gradient interval is divided according to the difference between the uniformity and the set threshold, and deduction is made in an increasing proportion, with the upper limit of deduction being the preset deduction proportion.

2. The foamed aluminum insulation detection method based on eddy current effect according to claim 1, characterized in that: The division rules for the detection units are as follows: Based on the geometric features and material distribution boundaries of the aluminum foam sample, the sample is divided into uniformly sized grid cells, each corresponding to an independent detection area, and the grid cells are used as detection units.

3. The foamed aluminum insulation detection method based on eddy current effect according to claim 1, characterized in that: The process for generating the initial defect value is as follows: Set early warning thresholds for insulation defect assessment indicators and material stability assessment indicators; If the evaluation index of any detection unit is lower than the set warning threshold, the initial defect value of that detection unit will be assigned the preset highest defect level. If all evaluation indicators are higher than the set warning threshold, the initial defect value is calculated based on the insulation defect evaluation indicator and the material stability evaluation indicator using the attenuation function.

4. The foamed aluminum insulation detection method based on eddy current effect according to claim 3, characterized in that: The execution steps of the dynamic correction include: Select a detection unit as the target detection unit and obtain the initial defect values ​​of all its adjacent detection units; Based on the material similarity and spatial distance between the target detection unit and its adjacent detection units, spatial correlation index values ​​are generated. Functional compatibility index values ​​are generated by matching the material types of the target detection unit and adjacent detection units. Adjustment weights are assigned based on spatial correlation index values ​​and functional compatibility index values; The initial defect values ​​of the target detection unit are weighted and adjusted based on the corrected weights to generate corrected defect values; The correction is completed by traversing all detection units, and the corrected defect value of each detection unit is output.

5. The method for detecting aluminum foam insulation based on eddy current effect according to claim 4, characterized in that: The process of generating the spatial correlation index value includes: The compositional difference degree is calculated based on the material composition differences between the target detection unit and adjacent detection units; The proximity is calculated based on the geometric distance between the target detection unit and its adjacent detection units; The basic spatial correlation index value is obtained by weighted fusion of the comprehensive component difference and proximity, and a compensation factor is set according to the connection type of adjacent detection units; The basic spatial correlation index values ​​are adjusted by a compensation factor to generate the final spatial correlation index values. The rules for setting the compensation factor are as follows: If the connection type is direct contact, a first compensation factor value is assigned; if the connection type is indirect interval, a second compensation factor value is assigned, and the first compensation factor value is greater than the second compensation factor value.

6. The foamed aluminum insulation detection method based on eddy current effect according to claim 4, characterized in that: The process for generating the functional compatibility index value includes: Match the dominant material type of adjacent detection units with the material type of the target detection unit; If the matching result belongs to a preset compatible type combination, the function compatibility index value is calculated based on each material type and preset weight; If the combination does not belong to a compatible type, the correlation degree is calculated according to the material compatibility specification, and the correlation degree is converted into a functional compatibility index value through a mapping function.

7. The method for detecting aluminum foam insulation based on eddy current effect according to claim 6, characterized in that: The calculation process for the correlation degree is as follows: Obtain a historical dataset of aluminum foam samples and statistically analyze the co-occurrence frequency of the target detection unit and the corresponding material types of adjacent detection units. The material query function depends on the weight table to obtain the initial weight values ​​of the material types of the target detection unit and adjacent detection units. Calculate the compatibility coefficient according to the compatibility rules for aluminum foam materials; Adjust the initial weight value of the function dependency based on the compatibility coefficient, and generate the corrected weight value of the function dependency. The comprehensive correction function relies on weight values ​​and co-occurrence frequencies to calculate the correlation.

8. The method for detecting aluminum foam insulation based on eddy current effect according to claim 4, characterized in that: The allocation process for the corrected weights is as follows: Obtain the detection target category of the aluminum foam sample, and set spatial correlation weight and functional compatibility weight according to the detection target category; The corrected weights are calculated based on the set weights, the final spatial correlation index value, and the functional compatibility index value.

Images