Piezoelectric guided wave online damage alarm method for aircraft flight test structure

Abstract

The application discloses a piezoelectric guided wave online damage alarm method for an aircraft test flight structure, and belongs to the field of aircraft structure health monitoring, and comprises the following steps: constructing a guided wave sensing network; obtaining a reference feature sample set by acquiring a reference feature through a variable step window sliding method; performing probability statistical modeling on the reference feature sample set, obtaining a probability statistical correlation feature and a reference feature sample set residual error, and taking an absolute value to obtain a reference alarm feature; establishing a monitoring feature sample set, multiplying the monitoring feature sample set and the probability statistical correlation feature to obtain a monitoring feature sample set residual error, and taking an absolute value to obtain a structure damage alarm feature; performing network weighted fusion on the structure damage alarm feature, determining a damage high-probability channel according to a channel screening threshold, calculating a structure damage alarm threshold of the damage high-probability channel, and realizing structure damage alarm. The method is simple and efficient in implementation process, and realizes reliable alarm of structure pit damage under the influence of time-varying factors.

Description

Technical Field

[0001] This invention relates to the field of aircraft structural health monitoring technology, and in particular to an online damage alarm method for piezoelectric guided waves for aircraft flight test structures. Background Technology

[0002] Structural health monitoring, as a typical interdisciplinary technical field, involves multiple research directions such as structural engineering, materials science, computer science, sensor technology, metrology, and signal processing. Structural health monitoring technology integrates sensing elements into structures to achieve real-time online monitoring of their health status. Its goal is to promptly detect internal damage, locate it, and assess its severity, thereby predicting the structure's remaining load-bearing capacity and service life. Ultimately, it aims to achieve autonomous diagnosis and condition management of structural health, ensuring safe and stable structural operation, reducing maintenance costs, and extending service life.

[0003] Among existing structural health monitoring technologies, guided wave structural health monitoring technology shows promising engineering application prospects due to its wide monitoring range and high sensitivity to minor damage. Multiple piezoelectric elements are arranged on the surface of a structure or embedded within it to form a guided wave sensor network. When guided waves generated by the piezoelectric elements propagate through the structure and encounter damage such as cracks, delamination, or debonding, phenomena such as scattering, energy attenuation, and amplitude and phase changes occur. Therefore, by analyzing these characteristic changes in the guided wave signal, damage monitoring and identification can be achieved. However, in practical engineering applications, analyzing guided wave signals using traditional baseline comparison methods faces significant challenges. Besides structural damage such as cracks, delamination, and corrosion altering the monitoring signal, time-varying factors such as temperature changes, vibration, load, and humidity in the structure and environment also cause signal changes. For example, temperature changes the elastic modulus and Poisson's ratio of the structure, and also affects parameters such as the piezoelectric constant of the piezoelectric elements. Therefore, the characteristics of the guided wave signal propagating in the structure will change accordingly with temperature variations. When the signal changes caused by time-varying factors are comparable in magnitude to, or even more significant than, the signal changes caused by structural damage, the presence of time-varying factors may mask the true damage information or lead to misjudgment.

[0004] Damage alarm is a fundamental step in damage monitoring, providing guidance for subsequent task decisions through a preliminary health assessment of the monitored object's condition. Achieving accurate damage alarms under time-varying factors requires overcoming the dual challenges of suppressing time-varying interference and designing alarm mechanisms. Limited by the insufficient sensitivity of time-invariant feature extraction and the lack of adaptability of alarm thresholds, the reliability of existing methods for damage monitoring under time-varying factors still needs further improvement. Summary of the Invention

[0005] The purpose of this application is to overcome the shortcomings of the prior art and provide a piezoelectric guided wave online damage alarm method for aircraft flight test structures.

[0006] In a first aspect, this application provides a piezoelectric guided wave online damage alarm method for aircraft flight test structures, comprising the following steps: A guided wave sensing network is constructed on the surface of the structure being monitored, and the guided wave sensing network includes multiple excitation sensing channels; When the monitored structure is in a healthy state, guided wave signals are collected, and reference features are obtained in each excitation sensing channel by using the variable step-size window sliding method to obtain a reference feature sample set. Probabilistic statistical modeling is performed on the benchmark feature sample set to obtain probabilistic statistical correlation features and benchmark feature sample set residuals. The absolute value of the benchmark feature sample set residuals is then processed to obtain benchmark alarm features. A monitoring feature sample set is established based on each excitation sensing channel. The monitoring feature sample set is multiplied with the probability statistical correlation feature to obtain the monitoring feature sample set residual. The absolute value of the monitoring feature sample set residual is processed to obtain the structural damage alarm feature of each excitation sensing channel. The structural damage alarm features of each excitation sensing channel are fused by network weighting, and the channel screening threshold is calculated using the structural damage alarm features to determine the channels with high probability of damage. The structural damage alarm threshold of the high-probability channel is calculated using the benchmark alarm characteristics. When the structural damage alarm characteristic of any excitation sensing channel is greater than the structural damage alarm threshold of the excitation sensing channel, a structural damage alarm is triggered.

[0007] Optionally, a guided wave sensing network is constructed on the surface of the monitored structure. The guided wave sensing network includes multiple excitation sensing channels, including deploying multiple sensors on the surface of the monitored structure, selecting one as the excitation end and another as the receiving end, and forming an excitation sensing channel between the excitation end and the receiving end. Multiple excitation sensing channels are deployed in this manner to obtain the guided wave sensing network.

[0008] Optionally, the variable step-size window sliding method includes: determining the window width based on the sampling rate and the center frequency of the guided wave signal; for each group of guided wave signals, the window slides from the initial sampling point, and before each slide, the deviation between the current window center point and the target value is calculated, and the sliding step size of the window is adaptively adjusted according to the deviation to obtain the adjusted step size; and the position of the next window is updated according to the adjusted step size.

[0009] Optionally, the adjusted step size The expression is: in, Indicates deviation; Indicates the window width; This indicates the initial step size.

[0010] Optionally, the variable step size window sliding method also includes resetting the step size to the initial step size after each update of the window position, and finally filtering out the window obtained after the step size reduction operation.

[0011] Optionally, probabilistic statistical modeling is performed on the benchmark feature sample set to obtain probabilistic statistical correlation features and benchmark feature sample set residuals. The absolute value of the benchmark feature sample set residuals is then processed to obtain benchmark alarm features, including: Probabilistic statistical modeling is performed on the benchmark feature sample set to obtain a vector autoregressive model for each benchmark feature sample; The coefficient matrix of each vector autoregressive model is determined by the maximum likelihood estimation method, and the corresponding probability density function is obtained. The probability density function is substituted into the white noise vector in the vector autoregressive model and the logarithm of both sides of the probability density function is taken to obtain the maximum likelihood function. The probabilistic statistical correlation features are determined based on the maximum likelihood function. Multiply the probabilistic statistical correlation feature with the benchmark feature sample set to obtain the benchmark residual for each benchmark feature sample in a single channel, thus obtaining the benchmark feature sample set residual; The absolute value of each element of the residual of the benchmark feature sample set is taken to obtain the corresponding benchmark alarm feature.

[0012] Optionally, the network weighted fusion includes setting the weight of the structural damage alarm features obtained from any set of monitoring data as the square of the data value.

[0013] Optionally, the structural damage alarm threshold is the mean of the baseline alarm features plus the standard deviation of the baseline alarm features.

[0014] Secondly, this application also provides an online piezoelectric guided wave damage alarm system for aircraft flight test structures, comprising: one or more processors; a storage device for storing one or more programs; and when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the online piezoelectric guided wave damage alarm method for aircraft flight test structures as described in any one aspect.

[0015] Thirdly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the piezoelectric guided wave online damage alarm method for aircraft flight test structures as described in any one of the first aspects.

[0016] This application provides an online damage alarm method for piezoelectric guided waves in aircraft flight test structures. By combining the construction of a guided wave sensing network with a variable step-size window sliding method, it can adaptively locate and accurately extract the target band reference features of the guided wave signal under healthy conditions. This effectively overcomes the sensitivity of traditional fixed-window methods to changes in signal wave velocity and propagation path, thus significantly improving the accuracy and efficiency of feature extraction. Probabilistic statistical modeling is used to mine feature correlation patterns to suppress interference from time-varying factors such as temperature and load, enhancing the model's ability to perceive subtle changes in the structural health status. A network weighted fusion strategy is used to fuse the structural damage alarm features of each channel and calculate the channel filtering threshold, which can adaptively highlight damage-sensitive channels and suppress environmental noise interference, achieving accurate location of high-probability damage channels. By simultaneously setting network filtering thresholds and channel alarm thresholds, reliable alarm for structural pit damage under the influence of time-varying factors is achieved. The algorithm is simple, accurate, and efficient, effectively improving the accuracy and reliability of structural damage alarms under the influence of time-varying factors. Furthermore, it has a certain robustness to new time-varying factors during monitoring, providing efficient technical support for safety monitoring and maintenance decisions for structures such as aircraft.

[0017] To make the above-mentioned features and advantages of the invention more apparent and understandable, specific embodiments are described below, and detailed descriptions are provided in conjunction with the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a piezoelectric guided wave online damage alarm method for aircraft flight test structures provided in one embodiment of this application.

[0020] Figure 2 This is a flowchart of step S3 in the piezoelectric guided wave online damage alarm method for aircraft flight test structures provided in one embodiment of this application.

[0021] Figure 3 This is a schematic diagram of the flight test structure and waveguide sensor network layout in a specific embodiment of this application.

[0022] Figure 4 This is a schematic diagram showing the load application and the location of the pit damage in another specific embodiment of this application.

[0023] Figure 5This is a schematic diagram of network weighted fusion for damage signals in another specific embodiment of this application.

[0024] Figure 6 This is a damage alarm result in another specific embodiment of this application. Detailed Implementation

[0025] To make the objectives and technical solutions of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the described embodiments of this application without creative effort are within the scope of protection of this application.

[0026] In one embodiment, see Figure 1 This application provides a piezoelectric guided wave online damage alarm method for aircraft flight test structures, which may include the following steps: steps S1 to S6.

[0027] Step S1: Construct a guided wave sensing network on the surface of the structure being monitored. The guided wave sensing network includes multiple excitation sensing channels.

[0028] Step S2: When the monitored structure is in a healthy state, the guided wave signal is acquired, and the reference features are obtained in each excitation sensing channel by using the variable step size window sliding method to obtain the reference feature sample set.

[0029] Step S3: Perform probabilistic statistical modeling on the benchmark feature sample set, obtain probabilistic statistical correlation features and benchmark feature sample set residuals, and perform absolute value processing on the benchmark feature sample set residuals to obtain benchmark alarm features.

[0030] Step S4: Establish a monitoring feature sample set based on each excitation sensing channel, multiply the monitoring feature sample set with the probability statistical correlation feature to obtain the monitoring feature sample set residual, and perform absolute value processing on the monitoring feature sample set residual to obtain the structural damage alarm feature of each excitation sensing channel.

[0031] Step S5: Perform network weighted fusion of the structural damage alarm features of each excitation sensing channel, and use the structural damage alarm features to calculate the channel screening threshold to determine the high-probability damage channel.

[0032] Step S6: Calculate the structural damage alarm threshold of the high-probability channel using the benchmark alarm features. When the structural damage alarm feature of any excitation sensing channel is greater than the structural damage alarm threshold of the excitation sensing channel, a structural damage alarm is triggered.

[0033] This application's piezoelectric guided wave online damage alarm method for aircraft flight test structures accurately extracts guided wave reference features using a variable step-size window sliding method and effectively isolates signal fluctuations caused by time-varying factors based on probabilistic statistical correlation modeling, thereby constructing reference alarm features robust to time-varying interference. By simultaneously setting network filtering thresholds and channel alarm thresholds, reliable alarm for structural pit damage under the influence of time-varying factors is achieved. The algorithm is simple, accurate, and efficient. It fully utilizes the overall information of the guided wave sensor network for preliminary screening and ensures the accuracy of the alarm through fine discrimination of a single channel. This significantly improves the reliability, sensitivity, and anti-interference capability of structural damage monitoring in complex time-varying environments, effectively enhancing the accuracy and reliability of structural damage alarms under the influence of time-varying factors, and exhibiting a certain degree of robustness to new time-varying factors during the monitoring process.

[0034] It is understood that the method of this application can be widely applied to various plate-shaped, shell-shaped or composite material structures that require health monitoring, including but not limited to aircraft test flight structures, ship structures, bridge structures or pressure vessel structures.

[0035] For ease of understanding, the following description uses a typical aerospace structure, such as an aircraft test flight structure, as the monitored structure. Steps S1 to S6 are explained in detail. The specific structure in this embodiment is merely an example and does not constitute any limitation on the scope of protection of this invention.

[0036] In step S1, please refer to Figure 1 In step S1, a guided wave sensing network is constructed on the surface of the monitored structure, the guided wave sensing network including multiple excitation sensing channels.

[0037] Specifically, E sensors are deployed on the surface of the monitored structure, where E is greater than 2. One sensor is selected as the excitation end to generate the guided wave signal, and the other is selected as the receiving end to receive the guided wave signal. An excitation sensing channel is formed between the excitation end and the receiving end, and the sensors are deployed in this manner. n A single excitation sensing channel is used to obtain a guided wave sensing network, which covers the monitored structure.

[0038] As an example, the sensors can be distributed in an array.

[0039] As an example, the E sensors can be piezoelectric sensors.

[0040] In step S2, please refer to Figure 1 In step S2, when the monitored structure is in a healthy state, guided wave signals are acquired, and reference features are obtained in each excitation sensing channel by using the variable step-size window sliding method to obtain a reference feature sample set.

[0041] Specifically, when the monitored structure is in a healthy state, data is collected in the guided wave sensor network.n Group guided wave signals to obtain healthy state guided wave signals. , i =1,2,…, n ,based on n Group health status guided wave signal settings t Reference signal for the time period A variable step-size window sliding method is used to extract reference features within each excitation sensing channel, resulting in a reference feature sample set. ,in, is the dimension of the baseline feature sample set.

[0042] As an example, the reference signal can be set as the first set of guided wave signals.

[0043] As an example, the variable step-size window sliding method may include, based on the sampling rate and the center frequency of the guided wave signal Determine window width For each set of guided wave signals, the window slides from the initial sampling point. Before each slide, the deviation between the current window center point and the target value is calculated. According to the deviation The adaptive adjustment window's sliding step size is obtained. According to the adjusted step size Update the position of the next window.

[0044] As an example, window width The expression is: in, Sampling rate, The center frequency of the guided wave signal This represents the number of peaks.

[0045] As an example, deviation The expression is: in, and These represent the start and end points of the current window, respectively. Indicates the first i The center point of the target band of the guided wave signal.

[0046] As an example, the adjusted step size The expression is: in, Indicates deviation; Indicates the window width; This represents the initial step size, which can be flexibly adjusted according to the size of the target window to adapt to different search needs.

[0047] As an example, the adjustment based on the step size The expression to update the position of the next window is: in, and These represent the start and end points of the current window, respectively. and These represent the start and end points of the updated window, respectively.

[0048] Furthermore, the variable step-size window sliding method may also include resetting the step size to the initial step size after each update of the window position. l Finally, the window obtained after the step size reduction operation is selected to prevent the search range from getting out of control due to the step size increasing continuously.

[0049] Specifically, during the sliding window process, if the current window deviates... If the target band center point is too large, the next sliding step size will be increased to quickly approach it. However, if left uncontrolled, the step size may continue to increase, causing excessive window jumps and potentially exceeding the target band center point, or even oscillating throughout the entire signal range, making precise positioning impossible. Therefore, when approaching the target band center point, the step size will be reduced for fine-tuning, and a new adjusted step size will be calculated each time based on the deviation. After updating the window position, the step size is reset to the initial step size. l The window that is finally determined after the step size reduction operation is selected and recorded as the valid output, which is used for subsequent feature extraction.

[0050] As an example, after the step size reduction operation, M windows are finally determined.

[0051] In step S3, please refer to Figure 1 In step S3, probabilistic statistical modeling is performed on the benchmark feature sample set to obtain probabilistic statistical correlation features and benchmark feature sample set residuals. The absolute value of the benchmark feature sample set residuals is then processed to obtain benchmark alarm features.

[0052] For example, please refer to Figure 2 Step S3 may include the following steps: Step S31 to Step S35.

[0053] Step S31: Perform probabilistic statistical modeling on the benchmark feature sample set to obtain the vector autoregressive model for each benchmark feature sample.

[0054] Step S32: Use the maximum likelihood estimation method to determine the coefficient matrix of each vector autoregressive model and obtain the corresponding probability density function. Substitute the probability density function into the white noise vector in the vector autoregressive model and take the logarithm of both sides of the probability density function to obtain the maximum likelihood function. Determine the probabilistic statistical association features based on the maximum likelihood function.

[0055] Step S33: Multiply the probabilistic statistical correlation feature with the benchmark feature sample set to obtain the benchmark residual of each benchmark feature sample in a single channel, thus obtaining the benchmark feature sample set residual.

[0056] Step S34: Take the absolute value of each element of the residual of the benchmark feature sample set to obtain the corresponding benchmark alarm feature.

[0057] Specifically, in step S31, a variable step-size window sliding method is used to extract reference features within each excitation sensing channel to obtain a reference feature sample set. ,in, , Represents the benchmark feature sample set The Middle i column vectors, i =1,2,3…, N , N Indicates the number of guided wave signals collected; Indicates the first M Damage factor of a direct-to-wave band of a guided wave signal.

[0058] As an example, the first M Damage factor of a direct-to-wave band of guided wave signals The expression is: in, express t The reference signal for the time period express t The first segment of the time period d Direct waveband of guided wave signals, and These represent the start and end times of the direct waveband of the intercepted guided wave signal, respectively.

[0059] Furthermore, for the benchmark feature sample set Probabilistic statistical modeling is performed to obtain a vector autoregressive model for each baseline feature sample. (Baseline feature sample) The vector autoregression model is expressed as: in, Indicates the first p indivual The parameter matrix, Represents the benchmark feature sample set The Middle column vectors, Represents a white noise vector. express a constant vector, p This represents the lag order of the vector autoregressive model.

[0060] Furthermore, in step S32, both sides of the vector autoregression model equation are subtracted simultaneously. x i-1 Obtain the first-order difference sequence ∆ of the vector autoregressive model. x i The expression is: in, Indicates the first n The short-term dynamic matrix of a vector autoregressive model. Represents the long-term equilibrium matrix. Represents a white noise vector. Represents the first in the benchmark feature sample set column vectors, Indicates the first A first-order difference sequence, p This represents the lag order of the vector autoregressive model.

[0061] As an example, the first n The parameter matrix and coefficient matrix of a vector autoregressive model are defined. The expression is: in, p This represents the lag order of the vector autoregressive model. Indicates the first s A parameter matrix.

[0062] As an example, the coefficient matrix is ​​defined by the identity matrix and the parameter matrix. The expression is: in, Represents the identity matrix. Indicates the first n A parameter matrix, p This represents the lag order of the vector autoregressive model.

[0063] Furthermore, the coefficient matrix is ​​defined for the identity matrix and the parameter matrix. Perform full-rank decomposition, the expression is: Where A and B represent the first and second coefficient matrices obtained from the full-rank decomposition, respectively.

[0064] Furthermore, subtract at the same time x n-1 The first difference sequence ∆ of the post-vector autoregressive model x i The probability density function is The probability density function is Substitute the white noise vector into the vector autoregressive model and the probability density function Taking the logarithm of both sides yields the maximum likelihood function. The expression is as follows: in, Represents a white noise vector; Represents the covariance matrix in a multivariate normal distribution; Indicates the first n A first-order difference sequence; This indicates the number of features that can be extracted from a single guided wave signal, which is determined by the window filtering results; Indicates the number of guided wave signals collected; This represents the short-term dynamic matrix of the first vector autoregressive model; Indicates the first The short-term dynamic matrix of a vector autoregressive model; Represents the first coefficient matrix; This represents the second coefficient matrix.

[0065] Furthermore, according to the maximum likelihood function Estimate the second coefficient matrix , the second coefficient matrix The first column vector in the dataset is set as a probabilistic statistical association feature. .

[0066] Furthermore, in step S33, the probability statistics are associated with the features. Compared with the benchmark feature sample set Multiply to obtain the baseline residual for each baseline feature sample in a single channel, thus obtaining the baseline feature sample set residual. In a single channel, the first i The benchmark residuals of each benchmark feature sample , means as follows: in, This represents the probabilistic statistical correlation characteristics of the first guided wave signal. This represents the damage factor in the direct waveband of the first guided wave signal. Indicates the first M Probabilistic statistical correlation characteristics of individual guided wave signals Indicates the first M Damage factor of a direct-to-wave band of a guided wave signal.

[0067] Further, in step S34, the residuals of the benchmark feature sample set are... Perform absolute value calculations for each element to obtain the corresponding baseline alarm characteristics. .

[0068] In step S4, please refer to Figure 1 In step S4, a monitoring feature sample set is established based on each excitation sensing channel. The monitoring feature sample set is multiplied with the probability statistical correlation feature to obtain the monitoring feature sample set residual. The absolute value of the monitoring feature sample set residual is processed to obtain the structural damage alarm feature of each excitation sensing channel.

[0069] Specifically, a set of guided wave signals is collected in each excitation sensing channel to establish a monitoring feature sample set. D 1×M , will monitor feature sample set D 1×M Features associated with probability and statistics β Multiply to obtain the residual of the monitoring feature sample set. e d residuals of the monitoring feature sample set e d By performing absolute value processing, the structural damage alarm characteristics of each excitation sensing channel are obtained. DAF .

[0070] In step S5, please refer to Figure 1 In step S5, the structural damage alarm features of each excitation sensing channel are fused using network weighting, and the channel screening threshold is calculated using the structural damage alarm features to determine the high-probability damage channels.

[0071] Specifically, the structural damage alarm characteristics of each excitation sensing channel are... DAF Perform network weighted fusion. The purpose of network weighted fusion is to filter alarm channels rather than directly issue the final alarm. To achieve this, the weight of the affected areas needs to be increased to ensure that all possible damaged channels are covered.

[0072] As an example, the structural damage alarm characteristics obtained from any set of monitoring data To increase the relative importance of larger data points, the weight can be set to the square of the data value, expressed as: in, Indicates the first i The weights corresponding to each structural damage alarm feature To monitor the first feature sample set i Structural damage alarm characteristics of each excitation sensing channel n This indicates the number of excitation sensing channels.

[0073] Furthermore, the weights are normalized to obtain normalized weights. According to normalized weights Calculate structural damage alarm features DAF weighted mean and weighted standard deviation The expression is: in, n Indicates the number of excitation sensing channels. Indicates the first i The weights corresponding to each structural damage alarm feature To monitor the first feature sample set i Structural damage alarm features of each excitation sensing channel.

[0074] Furthermore, the weighted average Add weighted standard deviation The channel filtering threshold T is obtained by the following expression: .

[0075] As an example, channels that are greater than the channel filtering threshold T can be identified as channels with a high probability of damage.

[0076] In step S6, please refer to Figure 1 In step S6, the structural damage alarm threshold of the high-probability channel is calculated using the benchmark alarm features. When the structural damage alarm feature of any excitation sensing channel is greater than the structural damage alarm threshold of the excitation sensing channel, a structural damage alarm is triggered.

[0077] Specifically, for the selected high-probability-of-damage channels, the data collected from them is used... n Group reference alarm characteristics HAF Calculate the structural damage alarm threshold U, where the structural damage alarm threshold U is the mean of the baseline alarm characteristics. Add the standard deviation of the baseline alarm characteristics The expression is: This idealizes the structural damage alarm characteristics under the damage influence law as a normal distribution, and the probability that all structural damage alarm characteristics are within one standard deviation plus or minus the mean is 68.3%.

[0078] As an example, the mean of basic alarm characteristics The expression is: in, Indicates the first i Group baseline alarm characteristics, n This indicates the number of excitation sensing channels.

[0079] As an example, the standard deviation of the baseline alarm characteristics The expression is: in, n Indicates the number of excitation sensing channels. This represents the mean of basic alarm characteristics. Indicates the first i Group baseline alarm characteristics.

[0080] Furthermore, when any excitation sensing channel has a damage alarm characteristic DAF When the damage exceeds the structural damage alarm threshold U of the excitation sensing channel, a damage alarm is triggered.

[0081] In one specific embodiment, the test flight structure—an aluminum alloy wing skin—is used as the structure being monitored. Figure 3 This is a schematic diagram of the flight test structure and the arrangement of the guided wave sensor network. It can be seen that the dimensions of the flight test structure are 1200mm, 668mm, and 832mm. Thirty-two sensors, numbered 1 to 32, are arranged on the surface of the flight test structure. The spacing between two sensors in each horizontal row is 150mm, and the spacing between sensors in each vertical row is 170mm. One sensor serves as the receiving end, acting as the excitation element for the guided wave signal; the other sensor serves as the receiving end, acting as the response element for the guided wave signal. These two piezoelectric sensors form an excitation sensing channel, and the 32 sensors constitute a large-area piezoelectric sensor network.

[0082] Furthermore, during the 0~5kN dynamic load process, 30 sets of guided wave signals were acquired online when the test structure was in a healthy state, thus obtaining the healthy state guided wave signals. , i =1,2,…,30. The first group of signals is taken as the reference signal, and the sampling rate in the signal acquisition parameters is... center frequency of guided wave signal Number of peaks A pit can be applied to the area enclosed by sensors 21, 22, 29, and 30 in the test structure to create a hole. D ,hole DThe size is 5mm × 5mm. After applying the pit damage, a torsional load of 0-5kN is applied to the flight test structure. The pit damage of the structure is monitored under the varying load, and 19 sets of guided wave signals of the damaged state of the flight test structure are collected. , j =1,2,…,19, Guided wave signals for healthy status Add damaged state guided wave signal A total of 49 sets of mixed-state guided wave signals were obtained. The load application and pit damage locations are as follows: Figure 4 As shown, the arrow in the upper right corner indicates the direction of the torsional load.

[0083] Furthermore, it can be based on the sampling rate and the center frequency of the guided wave signal Calculate the window width This includes 150 sampling points; the initial step size can be set. According to the deviation Adjust the sliding step size to obtain the adjusted step size. According to the adjusted step size The position of the next window is updated. To prevent the search range from getting out of control due to the step size continuously increasing, the step size is reset to 100 after each window position update, and the deviation is calculated again. Finally, the windows obtained after reducing the step size are selected. In this embodiment, 8 windows are finally selected.

[0084] Furthermore, using the first 20 sets of acquired health-guided wave signals as modeling signals, the baseline feature sample set of any excitation sensing channel can be represented as follows: Monitoring feature sample set ,in, , Probabilistic statistical modeling is performed on the baseline feature sample set to ultimately obtain probabilistic statistical correlation features. and the residuals of the benchmark feature sample set .

[0085] As an example, using 10 sets of health signals and 19 sets of damage signals as monitoring data, a monitoring feature sample set was obtained. Take the i-th monitoring data Features directly related to probability and statistics β Multiply to obtain the residual of the monitoring feature sample set e d This is used to determine the changes in the long-term equilibrium relationship between the monitored feature sample set and the benchmark feature sample set. The residuals of the benchmark feature sample set are then analyzed. e and monitoring feature sample set residuals e dThe absolute values ​​of each sensor channel are processed separately to obtain the reference alarm characteristics. HAF and structural damage alarm features DAF .

[0086] Furthermore, the structural damage alarm features of each excitation sensing channel are fused using network weighting, and the channel screening threshold is calculated using the structural damage alarm features to determine the channels with high probability of damage. Figure 5 This diagram illustrates network weighted fusion of damage signals in a specific embodiment of this application. The horizontal axis represents channels, the vertical axis represents damage features, and the dashed line represents channel screening thresholds. It can be seen that when the damage alarm feature of a certain channel in a damage monitoring sample exceeds the channel screening threshold T, it is identified as a high-probability damage channel. Channels 22-29 in the diagram are close to the damage and are significantly affected by it; therefore, their damage alarm feature values ​​are the highest, and channels 22-29 can be identified as alarm channels.

[0087] Furthermore, based on the selected channels 22-29, 20 sets of baseline alarm features were collected from them. HAF Calculate the structural damage alarm threshold U. When the structural damage alarm characteristic of any excitation sensing channel is greater than the structural damage alarm threshold of the excitation sensing channel, a structural damage alarm is triggered. Figure 6 This is a damage alarm result in another specific embodiment of this application, where the horizontal axis represents the number of samples, the vertical axis represents the damage alarm features, and the bottom horizontal line represents the structural damage alarm threshold. From Figure 6 As can be seen, the alarm features can include health modeling data, health monitoring data, and damage monitoring data. Only the channels surrounding the damage are locked, the boundary between health data and damage data feature values ​​is clear, and the anti-time-varying interference capability is strong. This application deploys a guided wave sensor network on the monitored structure and achieves accurate alarm for 5mm×5mm pit damage under the influence of 0-5kN dynamic load through network weighted fusion and channelized alarm coordination. The algorithm is simple, accurate, and efficient, and can effectively improve the reliability of structural damage alarm under the influence of time-varying factors.

[0088] This application's piezoelectric guided wave online damage alarm method for aircraft flight test structures constructs a full-coverage guided wave sensing network using an array of piezoelectric sensors. Combined with a variable-step-size window sliding method, it accurately extracts the reference features of guided wave signals. This method can adaptively locate and accurately extract the target band reference features of guided wave signals under healthy conditions, effectively overcoming the sensitivity of traditional fixed-window methods to changes in signal velocity and propagation path, thus significantly improving the accuracy and efficiency of feature extraction. By establishing a vector autoregressive model of the reference feature sample set through probabilistic statistical modeling and extracting probabilistic statistical correlation features, it can deeply explore the inherent statistical dependencies between multi-channel signals, enhancing the model's ability to perceive subtle changes in structural health. Based on this, it calculates residuals to obtain stable and reliable reference alarm features. Through network weighted fusion, it highlights the channel weights in the damaged area, accurately locking high-probability damaged channels and avoiding interference from irrelevant channels. By simultaneously setting network filtering thresholds and channel alarm thresholds, it achieves reliable alarm for structural pit damage under the influence of time-varying factors. The algorithm is simple, accurate, and efficient, effectively improving the accuracy and reliability of online piezoelectric guided wave damage alarms under the influence of time-varying factors, and exhibits a certain degree of robustness to new time-varying factors during monitoring. The method described in this application has a simple and efficient implementation process. It can achieve real-time and reliable alarm for minor dent damage to aircraft and other aviation structures without complex calculations. It significantly improves the accuracy and robustness of structural health monitoring under time-varying environments, provides strong protection for the safe operation of structures, and reduces maintenance costs.

[0089] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0090] In another embodiment, this application also provides a piezoelectric guided wave online damage alarm system for aircraft flight test structures, comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the piezoelectric guided wave online damage alarm method for aircraft flight test structures described in the above embodiment.

[0091] In another embodiment, this application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, it implements the various steps of the piezoelectric guided wave online damage alarm method for aircraft flight test structures provided in the above embodiments.

[0092] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0093] Although this application has been disclosed above with reference to embodiments, it is not intended to limit this application. Anyone skilled in the art can make some modifications and refinements without departing from the spirit and scope of this application.

Claims

1. A piezoelectric guided wave online damage alarm method for aircraft flight test structures, characterized in that, Includes the following steps: A guided wave sensing network is constructed on the surface of the structure being monitored, and the guided wave sensing network includes multiple excitation sensing channels; When the monitored structure is in a healthy state, guided wave signals are collected, and reference features are obtained in each excitation sensing channel by using the variable step-size window sliding method to obtain a reference feature sample set. Probabilistic statistical modeling is performed on the benchmark feature sample set to obtain probabilistic statistical correlation features and benchmark feature sample set residuals. The absolute value of the benchmark feature sample set residuals is then processed to obtain benchmark alarm features. A monitoring feature sample set is established based on each excitation sensing channel. The monitoring feature sample set is multiplied with the probability statistical correlation feature to obtain the monitoring feature sample set residual. The absolute value of the monitoring feature sample set residual is processed to obtain the structural damage alarm feature of each excitation sensing channel. The structural damage alarm features of each excitation sensing channel are weighted and fused by the network, and the channel screening threshold is calculated using the structural damage alarm features to determine the channels with high probability of damage. The structural damage alarm threshold of the high-probability channel is calculated using the benchmark alarm characteristics. When the structural damage alarm characteristic of any excitation sensing channel is greater than the structural damage alarm threshold of the excitation sensing channel, a structural damage alarm is triggered.

2. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 1, characterized in that, A guided wave sensing network is constructed on the surface of the monitored structure. The guided wave sensing network includes multiple excitation sensing channels. This includes deploying multiple sensors on the surface of the monitored structure, selecting one as the excitation end and another as the receiving end, with the excitation end and the receiving end forming an excitation sensing channel. Multiple excitation sensing channels are deployed in this manner to obtain the guided wave sensing network.

3. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 1, characterized in that, The variable step-size window sliding method includes: determining the window width based on the sampling rate and the center frequency of the guided wave signal; for each group of guided wave signals, the window slides from the initial sampling point; before each slide, the deviation between the current window center point and the target value is calculated; the sliding step size of the window is adaptively adjusted based on the deviation to obtain the adjusted step size; and the position of the next window is updated based on the adjusted step size.

4. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 3, characterized in that, Adjusted step size The expression is: in, Indicates deviation; Indicates the window width; This indicates the initial step size.

5. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 4, characterized in that, The variable step size window sliding method also includes resetting the step size to the initial step size after each update of the window position, and finally filtering out the window obtained after the step size reduction operation.

6. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 1, characterized in that, Probabilistic statistical modeling is performed on the benchmark feature sample set to obtain probabilistic statistical correlation features and benchmark feature sample set residuals. The absolute value of the benchmark feature sample set residuals is then processed to obtain benchmark alarm features, including... Probabilistic statistical modeling is performed on the benchmark feature sample set to obtain a vector autoregressive model for each benchmark feature sample; The coefficient matrix of each vector autoregressive model is determined by the maximum likelihood estimation method, and the corresponding probability density function is obtained. The probability density function is substituted into the white noise vector in the vector autoregressive model and the logarithm of both sides of the probability density function is taken to obtain the maximum likelihood function. The probabilistic statistical correlation features are determined based on the maximum likelihood function. Multiply the probabilistic statistical correlation feature with the benchmark feature sample set to obtain the benchmark residual for each benchmark feature sample in a single channel, thus obtaining the benchmark feature sample set residual; The absolute value of each element of the residual of the benchmark feature sample set is taken to obtain the corresponding benchmark alarm feature.

7. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 1, characterized in that, The network weighted fusion includes setting the weight of the structural damage alarm features obtained from any set of monitoring data as the square of the data value.

8. The piezoelectric guided wave online damage alarm method for aircraft flight test structures according to claim 1, characterized in that, The structural damage alarm threshold is the mean of the baseline alarm features plus the standard deviation of the baseline alarm features.

9. A piezoelectric guided wave online damage alarm system for aircraft flight test structures, characterized in that, include: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the piezoelectric guided wave online damage alarm method for aircraft flight test structures as described in any one of claims 1 to 8.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the piezoelectric guided wave online damage alarm method for aircraft flight test structures as described in any one of claims 1 to 8.

Images