Thermal barrier coating multi-dimensional terahertz nondestructive testing method and system based on adaptive optimization mechanism

Abstract

The invention discloses a thermal barrier coating multi-dimensional terahertz nondestructive testing method and system based on an adaptive optimization mechanism, and the method comprises the steps: constructing a matrix sample library, and preparing thermal barrier coating samples with different thicknesses, porosities and uniformity; obtaining terahertz detection data of three dimensions of geometric dimension stability, microstructure integrity and material structure uniformity of the thermal barrier coating; establishing a multi-parameter correction model based on surface roughness, and performing feature extraction, fusion analysis and error compensation on the detection data by adopting a multi-modal integration algorithm; an adaptive optimization model is established through a deep learning algorithm, and continuous iterative optimization of the detection system is realized in combination with a feedback correction mechanism; the method is applied to an engine blade thermal barrier coating, an actual blade coating is detected, detection data are collected for system optimization, system parameters and models are updated, and system applicability and robustness are improved. The method can be widely applied to the field of nondestructive testing of coating structures such as aero-engine blade thermal barrier coatings.

Description

Technical Field

[0001] The present invention belongs to the technical field of nondestructive testing and evaluation, and specifically relates to a multi-dimensional terahertz nondestructive testing method and system for thermal barrier coatings based on an adaptive optimization mechanism. Background Art

[0002] As a key protective material for aircraft engines, thermal barrier coatings are widely used in aviation hot end components, such as turbine blades and combustion chambers, to reduce the operating temperature of the base material and extend its service life. With the advancement of aircraft engine technology, the thrust ratio and operating temperature continue to increase, which puts higher requirements on the performance of thermal barrier coatings. Effective thermal barrier coatings not only insulate, but also need to have excellent oxidation resistance and corrosion resistance to ensure long-term stable operation in harsh environments. The service life of thermal barrier coatings is affected by multiple factors, of which the three main factors are thickness, porosity and uniformity. Thickness is an important parameter affecting the thermal insulation performance of the coating. Too thin may cause overheating of the base material, and too thick may cause the coating to peel off. Porosity directly affects the strength and thermal conductivity of the coating. Too much porosity will weaken the overall structural stability of the coating. Uniformity reflects the distribution of the coating material. A uniform coating can effectively avoid local overheating and stress concentration, thereby improving its overall performance.

[0003] At present, the detection technology for thermal barrier coatings mainly focuses on the evaluation of a single factor, such as ultrasonic testing, eddy current testing and scanning electron microscopy. However, these methods often ignore the interaction between thickness, porosity and uniformity, and cannot fully reflect the comprehensive performance and service status of the coating. Therefore, the existing detection methods face great limitations in practical applications and are difficult to meet the comprehensive evaluation needs of high-performance thermal barrier coatings. Based on this, it is particularly necessary to develop a multi-dimensional terahertz non-destructive testing method and system for thermal barrier coatings based on an adaptive optimization mechanism. The system will integrate multiple parameters, and through a comprehensive evaluation of thickness, porosity and uniformity, it will use the high sensitivity and non-contact characteristics of terahertz non-destructive technology to provide more accurate and reliable detection data. At the same time, the adaptive optimization algorithm is used, combined with the feedback mechanism, to continuously improve the detection accuracy and system applicability, and provide solid technical support for the prediction of the service life of thermal barrier coatings for aircraft engines. This innovative method will open up a new direction for the development of non-destructive testing technology in the field of aircraft engines. Summary of the invention

[0004] The present invention aims to solve the deficiencies of the prior art and proposes a multi-dimensional terahertz nondestructive testing method and system for thermal barrier coatings based on an adaptive optimization mechanism. On the basis of the traditional single-factor detection method, the multi-dimensional detection technology, roughness error compensation and adaptive optimization mechanism are comprehensively utilized to comprehensively evaluate the structural characteristics and performance of thermal barrier coatings, and provide a reliable detection method for the safe operation of aircraft engines. This method significantly improves the prediction accuracy of the service life of thermal barrier coatings and provides a scientific basis for subsequent maintenance and management. In order to improve and solve the problems existing in the prior art.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism comprises the following steps:

[0007] S1. Build a matrix coating sample library to prepare thermal barrier coating samples with different thickness, porosity and uniformity. Each characteristic parameter corresponds to multiple different surface roughnesses.

[0008] S2. Obtain multi-dimensional terahertz data, perform terahertz testing on thermal barrier coatings in the sample library, and obtain test data in three dimensions: geometric dimensional stability, microstructural integrity, and material organization uniformity;

[0009] S3. Establish a roughness compensation model, establish a multi-parameter correction model based on surface roughness, use a multi-modal integration algorithm to extract features and perform fusion analysis on the detection data, and realize error compensation of roughness on the detection signal;

[0010] S4. Implement an adaptive optimization mechanism, use deep learning algorithms to establish an adaptive optimization model, and combine feedback correction mechanisms to achieve continuous iterative optimization of the detection system;

[0011] S5. Applied to thermal barrier coating of engine blades, the actual blade coating is tested, the test data is collected for system optimization, the system parameters and models are updated, and the system applicability and robustness are improved.

[0012] Preferably, the matrix coating sample library comprises preparing thermal barrier coating samples of different thicknesses, porosities and uniformities, and introducing different surface roughnesses for each characteristic parameter to form a matrix thermal barrier coating sample library.

[0013] The prepared sample groups with different thicknesses have a thickness range of 100-800 μm and 10 gradients are set;

[0014] The prepared sample groups with different porosities have a porosity range of 5%-25% and 10 gradients are set;

[0015] The prepared sample groups with different uniformities are controlled by group distribution ratio, and 10 gradients are set;

[0016] The different roughnesses prepared are as follows: each sample prepared has a roughness ranging from 1 to 10 μm, and 10 gradients are set.

[0017] Preferably, in S2, the data of terahertz detection of the matrix sample includes:

[0018] The multi-dimensional terahertz data includes first-dimensional data, second-dimensional data and third-dimensional data;

[0019] The first-dimensional data are terahertz detection data of samples with different thicknesses and roughness;

[0020] The second dimension data is the terahertz detection data of samples with different porosity and roughness;

[0021] The third dimension data is the terahertz detection data of samples with different uniformity and roughness.

[0022] Preferably, in S3, performing multimodal integration on the multi-dimensional terahertz data includes:

[0023] The multimodal integration processes the collected terahertz signals and extracts multi-dimensional characteristic parameters. The first, second and third dimensional data are subjected to signal filtering, and then the time domain, frequency domain and time-frequency characteristic parameters are extracted. The time domain characteristic parameter extraction uses Hilbert transform to extract the signal envelope, calculate the signal peak amplitude, peak position and waveform width, extract the signal root mean square value and kurtosis coefficient, and analyze the signal energy attenuation characteristics;

[0024] The frequency domain feature parameter extraction uses fast Fourier transform to obtain spectrum characteristics, calculates spectrum center of gravity and spectrum width parameters, extracts main frequency components and frequency distribution characteristics, and analyzes spectrum energy distribution characteristics; the time-frequency feature parameter extraction uses wavelet transform to perform multi-scale analysis, extracts wavelet coefficient energy distribution characteristics, calculates time-frequency joint distribution characteristics, and analyzes instantaneous frequency characteristics of the signal.

[0025] Preferably, in S3, establishing a roughness compensation model includes:

[0026] The roughness compensation model analyzes the impact of roughness, determines the main characteristic parameters of roughness, and constructs a roughness-feature mapping relationship; it uses a feature-level fusion method to integrate multi-dimensional data, construct a comprehensive feature space, and establish an error compensation function based on roughness. Follow the steps below to build:

[0027] S11. Based on the principal component analysis method, the multi-dimensional characteristic parameters are reduced in dimension, the characteristic principal components are obtained, and the sensitivity coefficient of each characteristic parameter to the roughness is calculated;

[0028] S21, constructing a feature weight matrix based on the sensitivity coefficient, and calculating the contribution of each feature parameter;

[0029] S31, based on the contribution of the characteristic parameters, a kernel function method is used to construct a nonlinear mapping relationship to obtain a mapping function between roughness and characteristic parameters;

[0030] S41. Use evidence theory to calculate the basic probability distribution function of multi-feature fusion and construct feature fusion rules;

[0031] S51, based on the fusion rule and mapping function, establishing a mathematical model for roughness compensation, and calculating a compensation coefficient;

[0032] S61, optimizing compensation model parameters by using support vector regression method and calculating compensation error;

[0033] S71. If the compensation error is less than a preset threshold, output the final compensation model; otherwise, return to step S31 to continue iterative optimization.

[0034] Preferably, in S4, roughness error compensation is combined with deep learning to build a model to implement an adaptive optimization mechanism, including:

[0035] The deep learning constructs a deep learning network model structure to achieve model training and optimization; the adaptive optimization mechanism establishes a feedback correction mechanism to achieve online learning and optimization and build a performance evaluation system. It includes the following steps:

[0036] S12, constructing a multi-layer convolutional neural network model, the input layer receives multi-dimensional feature parameters and roughness compensation data, and the output layer gives the optimized detection results;

[0037] S22. Perform initial model training based on matrix sample library data and use cross-validation method to evaluate the performance of the initial model; calculate the error between the model prediction value and the true value, and establish an error evaluation index system, including root mean square error, relative error and prediction deviation;

[0038] S32. Based on the error evaluation index, an adaptive feedback correction mechanism is constructed to dynamically adjust network parameters and learning rate; an online learning method is used to input the newly added detection data into the model to update feature weights and compensation coefficients;

[0039] S42, calculating the performance index of the updated model, if the performance improvement is greater than a preset threshold, saving the current model parameters; otherwise, returning to step S32 to adjust the optimization strategy;

[0040] S52, based on the performance indicators, evaluate the model stability and generalization ability, if the preset conditions are met, output the optimized model; otherwise, return to step S22 to re-evaluate the error;

[0041] S62, continuously collect new test data during the test process, regularly trigger step S32 to update the model, and form a closed-loop optimization mechanism; when the test environment changes significantly, automatically trigger step S22 to perform a comprehensive error assessment to achieve adaptive adjustment of the model;

[0042] S72. Regularly evaluate the performance of the entire optimization system. If the system performance is significantly degraded, return to step S12 to rebuild the model structure.

[0043] Preferably, in S5, the thermal barrier coating of engine blades is applied to collect detection data and update system parameters, including:

[0044] The invention is applied to the thermal barrier coating of engine blades, performs actual blade coating testing, performs multi-batch data collection, and performs test result analysis;

[0045] The updating of system parameters updates feature extraction algorithm parameters, optimizes the deep learning network structure, corrects data fusion weights, realizes system iterative updates, and forms a continuous optimization closed loop.

[0046] The present invention also provides a multi-dimensional terahertz nondestructive testing system for thermal barrier coatings based on an adaptive optimization mechanism, including a signal acquisition module, a data processing module, a feature extraction module, an intelligent identification module, a feedback correction module, and a practical application optimization module:

[0047] The signal acquisition module is used to scan and detect the thermal barrier coating using terahertz non-destructive testing technology to obtain multi-dimensional raw data including geometric dimensional stability, microstructure integrity and material organization uniformity;

[0048] The data processing module is used to preprocess and reduce noise on the multi-dimensional raw data by using signal filtering and data standardization methods;

[0049] The feature extraction module is used to extract features in the time domain, frequency domain and time-frequency domain of the preprocessed data using a multimodal signal processing method to obtain multi-dimensional feature parameters;

[0050] The intelligent identification module is used to build a parameter identification model based on a deep learning algorithm, combined with a surface roughness compensation mechanism, to achieve accurate identification of characteristic parameters of thermal barrier coatings;

[0051] The feedback correction module is used to achieve adaptive optimization and continuous iterative update of the system through error compensation analysis and parameter optimization;

[0052] The practical application optimization module is used to analyze and evaluate actual blade coating detection data, and update system parameters and models based on the evaluation results to improve the applicability and robustness of the system.

[0053] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a multi-dimensional terahertz nondestructive testing method and system for thermal barrier coatings based on an adaptive optimization mechanism, wherein a multi-dimensional terahertz evaluation of thermal barrier coatings is performed by integrating multiple key parameters such as thickness, porosity, and uniformity. A roughness error compensation model is established, combined with a multimodal integration algorithm and a deep learning optimization mechanism, which effectively improves the accuracy of the detection signal and the robustness of the system. This method can not only accurately reflect the microstructural characteristics of the coating, but also adapt to the actual detection needs under different working conditions.

[0054] The multi-dimensional detection method of thermal barrier coating provided by the present invention belongs to non-contact non-destructive testing. Through real-time monitoring and feedback of the thermal barrier coating of the engine blade, the detection system is continuously optimized to have higher applicability and flexibility. The implementation of the present invention will significantly improve the prediction accuracy of the service life of the thermal barrier coating, provide a scientific basis for the safety of aircraft engines, and promote the application and development of non-destructive testing technology in the aviation field. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0056] Figure 1 is a flow chart of a multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to an embodiment of the present invention;

[0057] Figure 2 Schematic diagram of a matrix thermal barrier coating sample library according to an embodiment of the present invention;

[0058] Figure 3 is a flow chart of error compensation based on roughness established in an embodiment of the present invention;

[0059] Figure 4 is a flow chart of establishing an adaptive optimization mechanism according to an embodiment of the present invention;

[0060] Figure 5 It is a schematic diagram of a multi-dimensional terahertz detection system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] Embodiment 1

[0064] like Figure 1 As shown, a multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism includes the following steps:

[0065] S1. Construct a matrix coating sample library to prepare thermal barrier coating samples with different thickness, porosity and uniformity. Each characteristic parameter corresponds to multiple different surface roughnesses.

[0066] Specifically, Figure 2 As shown, a matrix coating sample library is prepared, and thermal barrier coating samples of different thicknesses, porosities and uniformities are prepared. Different surface roughnesses are introduced for each characteristic parameter to form a matrix thermal barrier coating sample library. Different thickness sample groups, the thickness range is 100-800μm, and 10 gradients are set; different porosity sample groups, the porosity range is 5%-25%, and 10 gradients are set; different uniformity sample groups, through group ratio control, 10 gradients are set; different roughness, each prepared sample contains roughness, the range is 1-10μm, and 10 gradients are set.

[0067] S2. Acquire multi-dimensional terahertz data, perform terahertz detection on the thermal barrier coatings in the sample library, obtain detection data in three dimensions: geometric dimensional stability, microstructure integrity, and material organization uniformity, and reflect the geometric, microscopic, and component characteristics of the thermal barrier coatings;

[0068] Specifically, multi-dimensional terahertz data includes first-dimensional data, second-dimensional data, and third-dimensional data;

[0069] The first dimension of data is the terahertz detection data of samples with different thicknesses and roughness;

[0070] The second dimension data is the terahertz detection data of samples with different porosity and roughness;

[0071] The third dimension data is the terahertz detection data of samples with different uniformity and roughness.

[0072] S3. Establish a roughness compensation model, establish a multi-parameter correction model based on surface roughness, use a multi-modal integration algorithm to extract features and perform fusion analysis on the detection data, and realize error compensation of roughness on the detection signal;

[0073] Specifically, multi-modal integration is used to process the collected terahertz signals and extract multi-dimensional feature parameters. The first, second and third dimensional data are subjected to signal filtering, and then the time domain, frequency domain and time-frequency feature parameters are extracted. Time domain feature parameter extraction: Hilbert transform is used to extract the signal envelope, calculate the signal peak amplitude, peak position and waveform width, extract the signal root mean square value and kurtosis coefficient, and analyze the signal energy attenuation characteristics; frequency domain feature parameter extraction: Fast Fourier transform is used to obtain spectrum characteristics, calculate the spectrum center of gravity and spectrum width parameters, extract the main frequency component and frequency distribution characteristics, and analyze the spectrum energy distribution characteristics; time-frequency feature parameter extraction: Wavelet transform is used for multi-scale analysis, wavelet coefficient energy distribution characteristics are extracted, time-frequency joint distribution characteristics are calculated, and the instantaneous frequency characteristics of the signal are analyzed. These feature parameters together constitute the feature space of terahertz detection, providing a data basis for subsequent intelligent recognition.

[0074] Specifically, Figure 3 As shown, a roughness compensation model is established to analyze the influence of roughness, determine the main characteristic parameters of roughness influence, and construct the roughness-feature mapping relationship; the feature-level fusion method is used to integrate multi-dimensional data, construct a comprehensive feature space, and establish an error compensation function based on roughness. Follow the steps below to build:

[0075] S11. Based on the principal component analysis method, the multi-dimensional characteristic parameters are reduced in dimension, the characteristic principal components are obtained, and the sensitivity coefficient of each characteristic parameter to the roughness is calculated;

[0076] Specifically, standardize and preprocess the feature data:

[0077]

[0078] Among them, X ij is the jth characteristic parameter of the i-th sample, μ j is the mean of the jth feature, σ j is the standard deviation of the jth feature.

[0079] Compute the covariance matrix of the standardized data:

[0080]

[0081] Among them, C is the covariance matrix, n is the number of samples, and X is the standardized feature matrix.

[0082]

[0083] Among them, S j is the sensitivity coefficient of the jth feature, Y is the roughness value, X j is the jth characteristic parameter. S21, constructing a characteristic weight matrix based on the sensitivity coefficient, and calculating the contribution of each characteristic parameter;

[0084] Weight and contribution calculation:

[0085]

[0086] Among them, W j is the weight of the jth feature, m is the total number of features, C j is the contribution of the jth feature, λ j is the corresponding eigenvalue.

[0087] S31, based on the contribution of the characteristic parameters, a kernel function method is used to construct a nonlinear mapping relationship to obtain a mapping function between roughness and characteristic parameters;

[0088] Gaussian kernel function and mapping function:

[0089]

[0090] Among them, x i ,x j is the feature vector, σ is the kernel function parameter, α i is the mapping coefficient, b is the bias term, and X is the input feature vector.

[0091] S41. Use evidence theory to calculate the basic probability distribution function of multi-feature fusion and construct feature fusion rules;

[0092] S51, based on the fusion rule and mapping function, establishing a mathematical model for roughness compensation, and calculating a compensation coefficient;

[0093]

[0094] Among them, β i is the compensation coefficient, m(A i ) is the fusion support, f(X i ) is the mapping function value.

[0095] S61, optimizing compensation model parameters by using support vector regression method and calculating compensation error;

[0096] S71. If the compensation error is less than a preset threshold, output the final compensation model; otherwise, return to step S31 to continue iterative optimization.

[0097] The above steps realize feature dimension reduction and sensitivity analysis, feature weight calculation and contribution evaluation, nonlinear mapping relationship construction, multi-feature fusion and compensation model optimization. The roughness error compensation is combined with deep learning to build a model to realize an adaptive optimization mechanism.

[0098] S4. Implement an adaptive optimization mechanism, use deep learning algorithms to establish an adaptive optimization model, and combine feedback correction mechanisms to achieve continuous iterative optimization of the detection system;

[0099] Specifically, Figure 4 As shown, a deep learning network model structure is constructed to realize model training and optimization; the adaptive optimization mechanism establishes a feedback correction mechanism to realize online learning and optimization and build a performance evaluation system. The following steps are included:

[0100] S12, constructing a multi-layer convolutional neural network model, the input layer receives multi-dimensional feature parameters and roughness compensation data, and the output layer gives the optimized detection results;

[0101] S22. Perform initial model training based on matrix sample library data and use cross-validation method to evaluate the performance of the initial model; calculate the error between the model prediction value and the true value, and establish an error evaluation index system, including root mean square error, relative error and prediction deviation;

[0102] S32. Based on the error evaluation index, an adaptive feedback correction mechanism is constructed to dynamically adjust network parameters and learning rate; an online learning method is used to input the newly added detection data into the model to update feature weights and compensation coefficients;

[0103] S42, calculating the performance index of the updated model, if the performance improvement is greater than a preset threshold, saving the current model parameters; otherwise, returning to step S32 to adjust the optimization strategy;

[0104] S52, based on the performance indicators, evaluate the model stability and generalization ability, if the preset conditions are met, output the optimized model; otherwise, return to step S22 to re-evaluate the error;

[0105] S62, continuously collect new test data during the test process, regularly trigger step S32 to update the model, and form a closed-loop optimization mechanism; when the test environment changes significantly, automatically trigger step S22 to perform a comprehensive error assessment to achieve adaptive adjustment of the model;

[0106] S72. Regularly evaluate the performance of the entire optimization system. If the system performance is significantly degraded, return to step S12 to rebuild the model structure.

[0107] S5. Applied to thermal barrier coating of engine blades, the actual blade coating is tested, the test data is collected for system optimization, the system parameters and models are updated, and the system applicability and robustness are improved.

[0108] Establish a feedback and correction mechanism, realize online learning and optimization, and build a performance evaluation system.

[0109] Specifically, it is applied to the thermal barrier coating of engine blades, conducts actual blade coating inspection, performs multi-batch data collection, and analyzes the inspection results; updates system parameters, updates feature extraction algorithm parameters, optimizes the deep learning network structure, corrects data fusion weights, realizes system iterative updates, and forms a continuous optimization closed loop.

[0110] Embodiment 2

[0111] The present invention also provides a multi-dimensional terahertz nondestructive testing system for thermal barrier coatings based on an adaptive optimization mechanism, which is characterized by comprising a signal acquisition module, a data processing module, a feature extraction module, an intelligent identification module, a feedback correction module and a practical application optimization module:

[0112] Specifically, Figure 5 As shown, the signal acquisition module is used to scan and detect the thermal barrier coating using terahertz non-destructive testing technology to obtain multi-dimensional raw data including geometric dimensional stability, microstructure integrity and material organization uniformity;

[0113] In particular, this module realizes the synchronous collection of multi-dimensional data, improves detection efficiency, uses non-destructive testing technology to ensure sample integrity, and has all-round scanning capabilities to ensure the integrity and accuracy of data collection.

[0114] A data processing module, used to preprocess and reduce noise on the multi-dimensional raw data by using signal filtering and data standardization methods;

[0115] In particular, this module adopts a filtering strategy to effectively suppress environmental interference and system noise, realize data standardization processing, and improve the reliability of subsequent analysis.

[0116] A feature extraction module is used to extract features in the time domain, frequency domain and time-frequency domain of the preprocessed data using a multimodal signal processing method to obtain multi-dimensional feature parameters;

[0117] In particular, this module integrates time domain, frequency domain and time-frequency domain analysis methods to comprehensively extract signal features. It adjusts feature extraction parameters according to signal characteristics to improve feature expression capabilities. It selects the most representative feature parameters through feature importance evaluation.

[0118] Intelligent identification module, which is used to build a parameter identification model based on deep learning algorithm and combine it with surface roughness compensation mechanism to realize accurate identification of characteristic parameters of thermal barrier coating;

[0119] In particular, this module uses deep neural networks to automatically learn high-order features to improve recognition accuracy. It integrates surface roughness compensation algorithms to reduce the impact of surface conditions and achieves accurate simultaneous recognition of multiple characteristic parameters of coatings.

[0120] Feedback correction module, used to achieve adaptive optimization and continuous iterative update of the system through error compensation analysis and parameter optimization;

[0121] In particular, this module establishes a multi-dimensional error evaluation system to monitor system performance in real time. It automatically adjusts system parameters based on the error analysis results to optimize system performance. Through continuous feedback correction, the system performance is continuously improved.

[0122] The practical application optimization module is used to analyze and evaluate the actual blade coating inspection data, and update the system parameters and models according to the evaluation results to improve the applicability and robustness of the system.

[0123] In particular, this module evaluates the test results in real time to ensure the reliability of the test. The model is continuously updated according to the actual application data to improve the adaptability of the system. Through diversified data training, the stability of the system under different conditions is enhanced.

[0124] This system has the following advantages:

[0125] Full process automation: full process automation is achieved from data collection to result output.

[0126] High level of intelligence: Integrates multiple intelligent algorithms to realize intelligent operation of the system.

[0127] Strong adaptability: Through the adaptive optimization mechanism, the system is ensured to have high adaptability in practical applications.

[0128] Good scalability: Modular design facilitates system function expansion and updating.

[0129] High reliability: Multiple optimization mechanisms ensure the reliability of system detection results.

[0130] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism, characterized in that: The following steps are involved: S1. Construct a matrix coating sample library to prepare thermal barrier coating samples with different thickness, porosity and uniformity. Each characteristic parameter corresponds to multiple different surface roughnesses. S2. Obtain multi-dimensional terahertz data, perform terahertz testing on thermal barrier coatings in the sample library, and obtain test data in three dimensions: geometric dimensional stability, microstructural integrity, and material organization uniformity; S3. Establish a roughness compensation model, establish a multi-parameter correction model based on surface roughness, use a multi-modal integration algorithm to extract features and perform fusion analysis on the detection data, and realize error compensation of roughness on the detection signal; S4. Implement an adaptive optimization mechanism, use deep learning algorithms to establish an adaptive optimization model, and combine feedback correction mechanisms to achieve continuous iterative optimization of the detection system; S5. Applied to thermal barrier coating of engine blades, the actual blade coating is tested, the test data is collected for system optimization, the system parameters and models are updated, and the system applicability and robustness are improved.

2. The multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to claim 1, characterized in that: The matrix coating sample library comprises preparing thermal barrier coating samples with different thicknesses, porosities and uniformities, introducing different surface roughnesses for each characteristic parameter, and forming a matrix thermal barrier coating sample library.

3. The multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to claim 2, characterized in that: In S2, the data of terahertz detection of the matrix sample includes: The multi-dimensional terahertz data includes first-dimensional data, second-dimensional data and third-dimensional data; The first-dimensional data are terahertz detection data of samples with different thicknesses and roughness; The second dimension data is the terahertz detection data of samples with different porosity and roughness; The third dimension data is the terahertz detection data of samples with different uniformity and roughness.

4. The multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to claim 3, characterized in that: In S3, multi-modal integration of multi-dimensional terahertz data includes: The multimodal integration processes the collected terahertz signals and extracts multi-dimensional characteristic parameters; performs signal filtering on the first, second and third dimensional data, and then extracts time domain, frequency domain and time-frequency characteristic parameters to achieve comprehensive characteristic analysis and performance evaluation of terahertz signals; The time domain feature parameter extraction is used to analyze the signal energy attenuation characteristics; the frequency domain feature parameter extraction is used to analyze the spectrum energy distribution characteristics; the time-frequency feature parameter extraction is used to analyze the instantaneous frequency characteristics of the signal.

5. The multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to claim 4, characterized in that: In S3, a roughness compensation model is established, including: The roughness compensation model analyzes the influence of roughness, determines the main characteristic parameters of the roughness influence, and constructs a roughness-feature mapping relationship; uses a feature-level fusion method to integrate multi-dimensional data, constructs a comprehensive feature space, and establishes a roughness-based error compensation function; and is constructed according to the following steps: S11. Based on the principal component analysis method, the multi-dimensional characteristic parameters are reduced in dimension, the characteristic principal components are obtained, and the sensitivity coefficient of each characteristic parameter to the roughness is calculated; S21, constructing a feature weight matrix based on the sensitivity coefficient, and calculating the contribution of each feature parameter; S31, based on the contribution of the characteristic parameters, a kernel function method is used to construct a nonlinear mapping relationship to obtain a mapping function between roughness and characteristic parameters; S41. Use evidence theory to calculate the basic probability distribution function of multi-feature fusion and construct feature fusion rules; S51, based on the fusion rule and mapping function, establishing a mathematical model for roughness compensation, and calculating a compensation coefficient; S61, optimizing compensation model parameters by using support vector regression method and calculating compensation error; S71. If the compensation error is less than a preset threshold, output the final compensation model; otherwise, return to step S31 to continue iterative optimization.

6. The multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to claim 5, characterized in that: In S4, roughness error compensation is combined with deep learning to build a model to implement an adaptive optimization mechanism, including: The deep learning constructs a deep learning network model structure to achieve model training and optimization; the adaptive optimization mechanism establishes a feedback correction mechanism to achieve online learning and optimization and build a performance evaluation system; including the following steps: S12, constructing a multi-layer convolutional neural network model, the input layer receives multi-dimensional feature parameters and roughness compensation data, and the output layer gives the optimized detection results; S22. Perform initial model training based on matrix sample library data and use cross-validation method to evaluate the performance of the initial model; calculate the error between the model prediction value and the true value, and establish an error evaluation index system, including root mean square error, relative error and prediction deviation; S32. Based on the error evaluation index, an adaptive feedback correction mechanism is constructed to dynamically adjust network parameters and learning rate; an online learning method is used to input the newly added detection data into the model to update feature weights and compensation coefficients; S42, calculating the performance index of the updated model, if the performance improvement is greater than a preset threshold, saving the current model parameters; otherwise, returning to step S32 to adjust the optimization strategy; S52, based on the performance indicators, evaluate the model stability and generalization ability, if the preset conditions are met, output the optimized model; otherwise, return to step S22 to re-evaluate the error; S62, continuously collect new test data during the test process, regularly trigger step S32 to update the model, and form a closed-loop optimization mechanism; when the test environment changes significantly, automatically trigger step S22 to perform a comprehensive error assessment to achieve adaptive adjustment of the model; S72. Regularly evaluate the performance of the entire optimization system. If the system performance is significantly degraded, return to step S12 to rebuild the model structure.

7. The multi-dimensional terahertz nondestructive testing method for thermal barrier coatings based on an adaptive optimization mechanism according to claim 5, characterized in that: In the S5, the thermal barrier coating of the engine blade is applied to collect the detection data and update the system parameters, including: The invention is applied to the thermal barrier coating of engine blades, performs actual blade coating testing, performs multi-batch data collection, and performs test result analysis; The updating of system parameters updates feature extraction algorithm parameters, optimizes the deep learning network structure, corrects data fusion weights, realizes system iterative updates, and forms a continuous optimization closed loop.

8. A multi-dimensional terahertz nondestructive testing system for thermal barrier coatings based on an adaptive optimization mechanism to implement the method as described in any one of claims 1 to 7, characterized in that: Including signal acquisition module, data processing module, feature extraction module, intelligent identification module, feedback correction module and practical application optimization module: The signal acquisition module is used to scan and detect the thermal barrier coating using terahertz non-destructive testing technology to obtain multi-dimensional raw data including geometric dimensional stability, microstructure integrity and material organization uniformity; The data processing module is used to preprocess and reduce noise on the multi-dimensional raw data by using signal filtering and data standardization methods; The feature extraction module is used to extract features in the time domain, frequency domain and time-frequency domain of the preprocessed data using a multimodal signal processing method to obtain multi-dimensional feature parameters; The intelligent identification module is used to build a parameter identification model based on a deep learning algorithm, combined with a surface roughness compensation mechanism, to achieve accurate identification of characteristic parameters of thermal barrier coatings; The feedback correction module is used to achieve adaptive optimization and continuous iterative update of the system through error compensation analysis and parameter optimization; The practical application optimization module is used to analyze and evaluate actual blade coating detection data, and update system parameters and models based on the evaluation results to improve the applicability and robustness of the system.

Images