A method for identifying cracks in a titanium alloy plate of a submersible based on a PVDF piezoelectric effect
By constructing a sensor network by arranging PVDF piezoelectric sheets on the titanium alloy plate of a submersible, and combining sparse dictionary and time-frequency envelope analysis, the problem of crack identification of titanium alloy plates in the deep-sea environment of a submersible was solved, achieving high-sensitivity and reliable crack identification and outputting crack level.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NAT DEEP SEA CENT
- Filing Date
- 2026-01-05
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies struggle to effectively identify minute cracks in titanium alloy plates in the deep-sea service environment of submersibles. In particular, under multimodal coupling, strong background noise, and complex strain fields, they cannot accurately separate crack-induced transient disturbances from stable components caused by normal operating loads, and lack quantitative analysis capabilities.
A distributed sensing network is constructed using PVDF piezoelectric sheets. Through sparse dictionary and time-frequency envelope extraction techniques, combined with support vector machine algorithms, adaptive sparse decomposition of piezoelectric signals and crack indication feature extraction are achieved. Local modal energy spectra are constructed to identify the presence and significance level of cracks.
It achieves highly sensitive, real-time crack response extraction and accurate crack identification in submersible operation, improving the accuracy and reliability of early crack identification, and can output crack saliency level, suitable for multimodal coupling environments.
Smart Images
Figure CN122171620A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of structural health monitoring technology, and in particular to a method for identifying cracks in titanium alloy plates of submersibles based on the PVDF piezoelectric effect. Background Technology
[0002] During deep-sea operation, the titanium alloy pressure hull and cabin components of submersibles are subjected to complex hydrostatic pressure, cyclic loads, and occasional impact loads over long periods. Even the development of minute cracks in the titanium alloy plate structure can lead to stress concentration, accelerated material fatigue, and potential instability and failure. Therefore, monitoring crack development in the titanium alloy plate structure of submersibles during operation is of significant safety importance.
[0003] The operating environment of submersibles is characterized by multimodal coupling, strong background noise, and complex strain field distribution, which imposes many limitations on traditional crack detection technologies. Static-based strain monitoring struggles to capture non-stationary disturbances during crack initiation; ultrasonic or acoustic emission methods rely on external excitation or high-energy signals, making them unsuitable for continuous use while the submersible is in operation; and signal processing methods based on single time-domain or frequency-domain analysis cannot effectively separate crack-induced transient disturbances from stationary components caused by normal operating loads. Furthermore, existing methods lack the quantitative analysis capability to combine crack dynamic response characteristics with modal energy distribution, making it impossible to accurately identify crack features such as spectral drift and energy abrupt changes in multimodal coupling environments, and they also lack stable classification models that can output crack significance levels. Summary of the Invention
[0004] This invention provides a method for identifying cracks in titanium alloy plates of submersibles based on the PVDF piezoelectric effect. This method enables high-sensitivity crack response extraction, feature separation, and grade identification of cracks in titanium alloy plates of submersibles during operation.
[0005] A method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect includes the following steps: S1: The PVDF piezoelectric sheet is attached to the test area of the titanium alloy plate of the submersible, and the corresponding piezoelectric signal sequence is obtained based on the strain response induced by the structural working load. S2: Input the piezoelectric signal sequence into the signal demixing model based on sparse dictionary and time-frequency envelope extraction, identify crack-induced non-stationary disturbance components, and extract crack indication features; S3: Construct local modal energy spectra based on crack indication characteristics, analyze their spectral drift and energy mutation modes in multimodal coupling environments, establish a crack identification model, and output the presence and significance level of cracks in the titanium alloy plates of the submersible.
[0006] Optionally, S1 includes: S11: Multiple PVDF piezoelectric sheets are arranged in a predetermined array on the surface of the test area of the titanium alloy plate of the submersible, and the PVDF piezoelectric sheets are ensured to be in close contact with the surface of the titanium alloy plate to form a distributed sensing network; when the titanium alloy plate of the submersible deforms under the action of structural working load, the distributed sensing network senses the change of strain field on its surface. S12: The strain field change sensed by the distributed sensing network is converted into a distributed charge signal proportional to the strain change rate through the piezoelectric effect of the PVDF piezoelectric sheet. This distributed charge signal is the direct response of the PVDF piezoelectric sheet to the dynamic strain field induced by the structural working load. S13: The distributed charge signals generated by all PVDF piezoelectric elements in the distributed sensor network are synchronously acquired through a multi-channel data acquisition system, and then impedance-converted and initially amplified through a signal conditioning circuit to form a multi-channel charge signal sequence. S14: Perform preprocessing operations such as standardization and filtering on the multi-channel charge signal sequence to eliminate environmental electromagnetic interference and system background noise, and finally obtain a pure and amplitude-normalized piezoelectric signal sequence, which serves as the input data for the subsequent signal demixing step.
[0007] Optionally, the predetermined array form in S11 is: a grid-like array with equal spacing on the surface of the test area of the titanium alloy plate of the submersible.
[0008] Optionally, the distributed charge signal is a dynamic response signal with a frequency range of 0.1 Hz to 10 kHz.
[0009] Optionally, S2 includes: S21: Construct an overcomplete sparse dictionary to characterize the health and crack states of titanium alloy plate structures in submersibles. This sparse dictionary consists of a parameterized waveform atom library and pre-acquired reference signals of crack-free states, providing a basis for sparse representation of subsequent signal demixing models. S22: Input the piezoelectric signal sequence into the signal demixing model, and use the sparse dictionary to perform adaptive sparse decomposition on the piezoelectric signal sequence to obtain the corresponding sparse representation coefficients. The sparse representation coefficients can separate the stationary components related to the normal operating load of the structure and the non-stationary components related to cracks in the piezoelectric signal sequence. S23: Based on the sparse representation coefficients, reconstruct and separate the non-stationary disturbance component mainly induced by crack initiation or propagation from the piezoelectric signal sequence. This non-stationary disturbance component is a transient or modulated signal that is distinct from background noise and normal operating strain response. S24: Perform time-frequency envelope extraction on the separated non-stationary disturbance components, obtain the time-varying energy envelope features of the non-stationary disturbance components through wavelet packet analysis algorithm, and form a time-frequency envelope matrix; S25: Calculate and extract a set of quantitative crack indication features from the time-frequency envelope matrix. The crack indication features include the envelope peak value, energy entropy, and energy concentration in a specific frequency band of the non-stationary disturbance component.
[0010] Optionally, the sparse dictionary constructed in S21 has a parameterized waveform atom library that is a Gaussian frequency modulated linear frequency modulated atom library.
[0011] Optionally, the wavelet packet analysis algorithm uses Daubechies wavelet basis functions to decompose the non-stationary disturbance components to complete the time-frequency envelope extraction.
[0012] Optionally, S3 includes: S31: Based on the crack indication features extracted from the piezoelectric signal sequence, a local modal energy spectrum characterizing the dynamic response characteristics of the crack is constructed through feature fusion and mode decomposition techniques. This local modal energy spectrum reflects the energy distribution related to the crack under different modes. S32: In the multimodal coupling environment of the titanium alloy plate of the submersible, the local modal energy spectrum is analyzed to identify and quantify the spectral drift and energy mutation modes caused by the presence of cracks. The spectral drift is manifested as the shift of characteristic frequencies, and the energy mutation mode is manifested as the discontinuous jump of specific modal energy levels. S33: Using the identified spectral line drift and energy mutation patterns as input feature vectors, a crack identification model based on threshold judgment and pattern classification is established. This crack identification model outputs Boolean judgment of crack existence and quantitative evaluation results of crack significance level of titanium alloy plate of submersible through support vector machine algorithm.
[0013] Optionally, the feature fusion and mode decomposition technique specifically involves: performing principal component analysis on multiple crack indication features to complete feature fusion, and using an empirical mode decomposition algorithm on the fused feature sequence to construct the local mode energy spectrum.
[0014] Optionally, the crack severity level is quantified into three discrete levels: slight, moderate, and severe.
[0015] The beneficial effects of this invention are: 1. This invention, by arranging multiple PVDF piezoelectric elements in the test area of a submersible's titanium alloy plate and constructing a distributed sensing network, can acquire dynamic piezoelectric signal sequences in real time within the strain field induced by structural working loads, ranging from 0.1 Hz to 10 kHz. This signal acquisition method can directly reflect the local non-stationary disturbances caused by the initiation stage of micro-cracks. From sampling, conditioning, filtering to normalization processing, signal purity is ensured, so that the crack-related dynamic response is not masked by background noise. Compared with traditional static strain or ultrasonic testing methods, this method can maintain high sensitivity detection capability while the submersible is in operation, improving the accuracy and real-time performance of early crack identification.
[0016] 2. This invention employs an overcomplete sparse dictionary composed of a parameterized waveform atom library and a crack-free state reference signal. Through a matching pursuit algorithm, it performs adaptive sparse decomposition of the piezoelectric signal sequence, effectively distinguishing between stationary components and crack-related non-stationary perturbation components at the signal level. The sparse representation coefficients are screened using a residual energy threshold setting criterion, and a time-frequency envelope matrix is formed using wavelet packet analysis, enabling precise characterization of crack-induced transient perturbations in the time-frequency dimension. A complete crack indication feature system is constructed by using envelope peak value, energy entropy, and energy concentration within a specific frequency band, allowing for the quantification of crack response in the time, frequency, and energy domains, significantly improving the interpretability and reliability of crack feature extraction.
[0017] 3. This invention achieves the fusion of multiple crack indication features through principal component analysis and constructs local modal energy spectra by combining empirical mode decomposition, realizing a modal hierarchical expression of the dynamic response of cracks. In a multimodal coupling environment, by quantitatively calculating the peak frequency shift and modal energy change rate of the dominant mode, effective spectral line drift and energy mutation modes can be accurately identified. Based on these crack response modes, a support vector machine crack identification model using radial basis function kernels is constructed. The Boolean judgment of crack existence is realized through the numerical range of the decision function, and the crack significance level of "slight", "moderate" and "severe" is output. The entire process realizes a complete link from the signal layer, feature layer to the decision layer, making the crack diagnosis results more stable, quantifiable and engineering applicable. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2This is a schematic diagram of the S2 process in an embodiment of the present invention. Detailed Implementation
[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.
[0021] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.
[0022] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.
[0023] like Figures 1-2 As shown, a method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect includes the following steps: S1: A PVDF piezoelectric sheet is attached to the test area of the titanium alloy plate of the submersible. The corresponding piezoelectric signal sequence is obtained based on the strain response induced by the structural working load. Specifically: S11: Arrange multiple PVDF piezoelectric elements in a predetermined array on the surface of the test area of the submersible's titanium alloy plate. The predetermined array is a grid pattern with equal spacing on the surface of the test area of the submersible's titanium alloy plate. To ensure the uniformity of the array geometry, the grid spacing is first determined according to the size of the test area, and the installation position of each PVDF piezoelectric element is marked on the surface of the titanium alloy plate using a positioning fixture.
[0024] To ensure a tight bond between the PVDF piezoelectric sheet and the titanium alloy plate surface, the titanium alloy plate surface must be cleaned, degreased, and slightly roughened before bonding to achieve a stable adhesion state. Then, the adhesive layer on the back of the PVDF piezoelectric sheet is aligned with the corresponding position on the titanium alloy plate surface, and uniform pressure is applied until adhesion and curing occur, ensuring that the bottom surface of each PVDF piezoelectric sheet forms a continuous contact surface with the titanium alloy plate surface.
[0025] A distributed sensing network is constructed by arranging multiple PVDF piezoelectric elements in an array, enabling the network to cover the area under test and fully sense the strain field changes generated by the titanium alloy plate of the submersible under structural working loads.
[0026] S12: When the titanium alloy plate of the submersible is subjected to structural working load, the distributed sensing network generates strain field changes. Multiple PVDF piezoelectric sheets convert the strain changes into distributed charge signals that are proportional to the strain changes through their piezoelectric effect.
[0027] To ensure stable output of the distributed charge signal, the electrodes of each PVDF piezoelectric element are led out through shielded wires and fixed along the surface of the submersible's titanium alloy plate using insulated clamps to avoid additional noise caused by plate vibration or wire swaying.
[0028] The frequency range of the distributed charge signal generated by the PVDF piezoelectric sheet is from 0.1 Hz to 10 kHz. This frequency range covers the main dynamic strain response of the titanium alloy plate of the submersible under structural working load, enabling the distributed charge signal to include crack-induced non-stationary disturbance components.
[0029] The distributed charge signal output by each PVDF piezoelectric element is transmitted to the subsequent data acquisition system through an independent signal channel, maintaining signal independence and ensuring a true reflection of the spatial distribution of strain field changes.
[0030] S13: The distributed charge signal is synchronously acquired through a multi-channel data acquisition system. The sampling rate of the multi-channel data acquisition system is set to no less than 20kHz to ensure aliasing-free acquisition of distributed charge signals with a frequency range of 0.1Hz to 10kHz.
[0031] To ensure the signal meets the acquisition requirements, the distributed charge signal first enters the signal conditioning circuit for impedance conversion and preliminary amplification. The signal conditioning circuit consists of a charge conversion module, an impedance matching module, and a preamplifier module. The charge conversion module converts the distributed charge signal into a voltage signal with measurable amplitude. The impedance matching module ensures that the signal amplitude remains stable during transmission. The preamplifier module linearly amplifies the signal to match the input range of the multi-channel data acquisition system.
[0032] After signal conditioning, the charge signals of each channel are recorded synchronously with a unified time reference to form a multi-channel charge signal sequence, providing a continuous and stable data foundation for further signal processing.
[0033] S14: Preprocessing operation to standardize and filter the multi-channel charge signal sequence to eliminate environmental electromagnetic interference and system background noise.
[0034] The filtering and noise reduction operation employs a bandpass filter with a passband frequency of 1Hz to 1kHz. This bandpass filter processes all channel signals uniformly, completely filtering out low-frequency drift components below 1Hz and high-frequency noise above 1kHz, while retaining the effective frequency components involved in the crack response characteristics.
[0035] The filtered signal is normalized by standardization. The standardization process is based on the amplitude range of each channel signal, converting the signal of each channel to a uniform scale to avoid subsequent signal demixing errors caused by the difference in sensitivity of PVDF piezoelectric elements.
[0036] After processing, a clean and amplitude-normalized piezoelectric signal sequence is formed, which serves as the input to a signal demixing model based on sparse dictionary and time-frequency envelope extraction.
[0037] S2: Input the piezoelectric signal sequence into a signal demixing model based on sparse dictionary and time-frequency envelope extraction, identify crack-induced non-stationary disturbance components, and extract crack indication features, specifically: S21: Construct an overcomplete sparse dictionary to characterize the health and crack states of the titanium alloy plate structure of the submersible. This sparse dictionary consists of a parameterized waveform atom library and pre-acquired reference signals for a crack-free state, forming the sparse representation basis for the subsequent signal demixing model. The steps are as follows: When the titanium alloy plate of the submersible is in a crack-free state, the titanium alloy plate of the submersible is placed under multiple known structural working load conditions, and the piezoelectric signal sequence is collected through PVDF piezoelectric sheet.
[0038] Alignment, denoising, and amplitude normalization are performed on multiple sets of piezoelectric signal sequences under crack-free conditions to extract waveform components that are stable under different structural working load levels, forming a set of reference signals for crack-free conditions.
[0039] When constructing the parameterized waveform atom library, the Gaussian frequency modulation and linear frequency modulation atom library is used as the basis for the parameterized waveform atom library. A set of center frequency parameters, linear frequency modulation parameters and time extension parameters covering the main frequency range of piezoelectric signals are pre-set. A series of time-locally concentrated waveform atoms are constructed through Gaussian envelope and linear frequency modulation.
[0040] After normalizing each Gaussian-modulated linear-modulated atom to its unit energy, it is superimposed and combined with the crack-free state reference signal set in column vector form to obtain an overcomplete sparse dictionary containing a parameterized waveform atom library and the crack-free state reference signal.
[0041] S22: Input the piezoelectric signal sequence into the signal demixing model, and use a sparse dictionary to perform adaptive sparse decomposition on the piezoelectric signal sequence to obtain the corresponding sparse representation coefficients.
[0042] The signal demixing model consists of the following three functional modules: (1) Signal input module: used to receive the piezoelectric signal sequence after preprocessing by S1 and store it with a unified sampling time axis; (2) Sparse dictionary matching module: used to perform correlation calculations one by one between the piezoelectric signal sequence and all dictionary atoms in the sparse dictionary constructed by S21; (3) Residual signal update module: used to update the current residual signal and determine the iteration termination condition after each sparse decomposition iteration.
[0043] After inputting the piezoelectric signal sequence into the signal demixing model, the sparse dictionary matching module first performs an inner product operation between the piezoelectric signal sequence and all dictionary atoms in the sparse dictionary. It then calculates the projection amplitude of the piezoelectric signal sequence along each dictionary atom direction and stores the absolute value of the projection amplitude as a correlation index. Finally, it searches all correlation indices and selects the dictionary atom with the highest correlation index that is closest to the piezoelectric signal sequence. The index of this dictionary atom is recorded as the best matching atom for the current iteration.
[0044] Then, the residual signal update module is entered, where the product of the current best matching atom and its corresponding projection amplitude is subtracted point by point from the piezoelectric signal sequence to calculate the new residual signal. The new residual signal is then re-inputted into the sparse dictionary matching module to perform the next round of inner product operation.
[0045] The matching pursuit algorithm achieves adaptive sparse decomposition through the iterative operation of the signal input module, sparse dictionary matching module, and residual signal update module. Each iteration generates a sparse representation coefficient corresponding to a specific dictionary atom, which is the projection amplitude of the piezoelectric signal sequence along the direction of that dictionary atom. As the number of iterations increases, the set of sparse representation coefficients is gradually formed, enabling the piezoelectric signal sequence to be sparsely represented with the minimum number of dictionary atoms.
[0046] The iterative loop stops when the energy of the residual signal is lower than the set termination threshold or the number of iterations reaches the preset upper limit. At this time, all the sparse representation coefficients obtained constitute the final sparse representation result of the piezoelectric signal sequence. Among them, the sparse representation coefficients with high energy concentration and corresponding transient or modulation characteristics indicate that the piezoelectric signal sequence contains crack-related non-stationary components, while other sparse representation coefficients mainly correspond to stationary components related to the normal operating load of the structure, thereby realizing the distinction between stationary and non-stationary components.
[0047] S23: Based on sparse representation coefficients, non-stationary disturbance components, mainly induced by crack initiation or propagation, are reconstructed and separated from piezoelectric signal sequences. These non-stationary disturbance components are transient or modulated signals that differ from background noise and normal operating strain responses.
[0048] To effectively separate crack-related non-stationary disturbances, a residual energy threshold setting criterion is first used to screen sparse representation coefficients. This criterion is established based on the sparse representation characteristics of the crack-free reference signal, using the amplitude distribution of sparse representation coefficients, dictionary atom response energy, and residual energy variation patterns in the crack-free state as benchmarks. The sparse representation coefficients of the actual piezoelectric signal sequence are categorized according to dictionary atom types, and the sparse representation coefficient energies belonging to stationary dictionary atoms and transient or modulation dictionary atoms are calculated separately.
[0049] Subsequently, the residual energy threshold setting criterion was applied to the sparse representation coefficient set to remove sparse representation coefficients whose energy level, temporal locality, or frequency locality did not match the crack-induced characteristics, while retaining sparse representation coefficients that satisfied the non-stationary perturbation characteristics. For the retained sparse representation coefficients, the dictionary atoms were searched by their numbers according to their corresponding dictionary atoms, and the sparse representation coefficients were multiplied point by point with the corresponding dictionary atoms and then superimposed item by item.
[0050] By linearly superimposing all sparse representation coefficients that meet the residual energy threshold setting criteria and their corresponding dictionary atoms, the non-stationary perturbation component separated from the piezoelectric signal sequence is reconstructed. This non-stationary perturbation component fully preserves the transient or modulated waveform characteristics caused by crack initiation or propagation, providing input data for subsequent time-frequency envelope extraction.
[0051] S24: Perform time-frequency envelope extraction on the separated non-stationary disturbance components. Obtain the time-varying energy envelope features of the non-stationary disturbance components using wavelet packet analysis algorithm, and form a time-frequency envelope matrix. The wavelet packet analysis algorithm uses Daubechies wavelet basis functions to decompose the non-stationary disturbance components to complete the time-frequency envelope extraction. The implementation steps are as follows: First, using the non-stationary disturbance component as the input signal for wavelet packet decomposition, the non-stationary disturbance component is decomposed layer by layer using the Daubechies wavelet basis function according to the preset number of wavelet packet decomposition layers. The non-stationary disturbance component is divided into multiple wavelet packet nodes layer by layer according to the frequency range, with each node representing a different subdivision frequency band.
[0052] For each wavelet packet node, the instantaneous energy value of that node over the entire time range is calculated, and the energy values are arranged according to the time index to form a time-varying energy envelope for a single frequency band. To ensure the stability of the energy envelope, the energy of all nodes is subjected to summation of squares and normalization.
[0053] The time-varying energy envelopes of all wavelet packet decomposition nodes are combined and arranged according to frequency band order and time order to form a time-frequency envelope matrix with time as the horizontal axis, frequency band number as the vertical axis, and energy values as matrix elements. This time-frequency envelope matrix records the energy distribution of non-stationary disturbance components within each sub-band over time, thus comprehensively characterizing the time-frequency evolution features of crack-induced signals.
[0054] S25: Calculate and extract a set of quantitative crack indication features from the time-frequency envelope matrix. The crack indication features include the envelope peak value, energy entropy, and energy concentration within a specific frequency band of the non-stationary perturbation component. The steps are as follows: First, for all frequency bands of the time-frequency envelope matrix, the local peak values of the energy envelope of each frequency band are searched along the time dimension, and the global maximum energy point is determined from them, which is used as the envelope peak value of the non-stationary perturbation component. The envelope peak value is obtained by comparing the energy amplitude of each frequency band under the same time index and selecting the frequency band and time point with the most significant energy change, which is used to characterize the energy surge characteristics caused by crack initiation or propagation.
[0055] Subsequently, the energy entropy is calculated by dividing the energy of each time-frequency unit in the time-frequency envelope matrix by the total energy of the matrix to obtain the energy percentage. Using the energy percentage as a probability distribution, the energy percentages of each time-frequency unit are summed according to the definition of Shannon entropy to obtain the energy entropy value. Energy entropy is used to characterize the degree of dispersion of crack-induced non-stationary disturbance components in the time-frequency distribution. The higher the entropy value, the more dispersed the energy distribution; the lower the entropy value, the more obvious the energy concentration.
[0056] Finally, the energy concentration within a specific frequency band is calculated. Wavelet packet decomposition nodes corresponding to the 50Hz to 2kHz frequency band are selected from the time-frequency envelope matrix, and the energy of each node is accumulated along the time dimension to obtain the total energy of the specific frequency band. This total energy of the specific frequency band is then divided by the total energy of the time-frequency envelope matrix to obtain the energy concentration within the specific frequency band. The energy concentration is used to characterize the energy accumulation of non-stationary disturbance components in the 50Hz to 2kHz frequency band, which coincides with the crack-induced modulation waveform and the main concentration area of sudden disturbances.
[0057] By jointly extracting the envelope peak value, energy entropy, and energy concentration within a specific frequency band, a complete set of crack indication features is formed. This set of crack indication features can quantitatively describe the amplitude characteristics, energy distribution characteristics, and frequency band concentration characteristics of crack-induced non-stationary perturbation components, providing accurate input parameters for subsequent local modal energy spectrum construction and crack identification models.
[0058] S3: Based on crack indication characteristics, a local modal energy spectrum is constructed, and its spectral drift and energy mutation modes in a multimodal coupling environment are analyzed. A crack identification model is established, and the presence and significance level of cracks in the titanium alloy plate of the submersible are output. Specifically: S31: Based on crack indication features extracted from piezoelectric signal sequences, a local modal energy spectrum characterizing the dynamic response of cracks is constructed using feature fusion and mode decomposition techniques. The local modal energy spectrum reflects the energy distribution associated with cracks under different modes.
[0059] First, the crack indication features extracted from S25 are arranged in chronological order to form a crack indication feature matrix. To reduce redundant information among different crack indication features and improve the sensitivity of features to crack dynamic behavior, principal component analysis (PCA) is performed on the crack indication feature matrix to complete feature fusion. PCA calculates the covariance matrix of the feature matrix, finds the eigenvalues and eigenvectors of the covariance matrix, and selects several principal components based on the magnitude of the eigenvalues. The original crack indication features are then projected onto the selected principal component space to obtain the fused feature sequence.
[0060] Subsequently, the fused feature sequence is decomposed into multiple physically meaningful intrinsic mode functions (IMFs) using an empirical mode decomposition algorithm. EMF extracts vibration modes at different time scales from the fused feature sequence layer by layer through a successive sieving method, with each IMF corresponding to a specific response frequency band. For each IMF, its instantaneous energy is calculated along the time dimension, and the instantaneous energy is categorized according to mode number to form a modal energy sequence.
[0061] Finally, the modal energy sequences of all intrinsic modal functions are combined and arranged according to modal number and time order to form a local modal energy spectrum. The local modal energy spectrum fully records the energy distribution of the fused crack indication characteristics in each mode, so that the crack-induced dynamic response characteristics are clearly presented in different modes.
[0062] S32: In the multimodal coupled environment of the titanium alloy plate of the submersible, analyze the local modal energy spectrum, identify and quantify the spectral drift and energy mutation modes caused by the presence of cracks.
[0063] First, using the reference local modal energy spectrum acquired under healthy conditions as a reference, the peak frequencies of the dominant modes under the same mode number are compared to calculate the offset relative to the healthy state reference frequency. This offset is the spectral drift. When the spectral drift offset exceeds a preset 5% threshold, the spectral drift is determined to be a valid spectral drift.
[0064] When identifying energy mutation modes, the local modal energy spectrum is first segmented according to time windows, and the rate of change of energy amplitude for each mode within adjacent time windows is calculated. This rate of change is then compared with the statistical distribution of the rate of change of energy amplitude under healthy conditions. If the rate of change of energy amplitude for a certain mode within adjacent time windows exceeds three times the standard deviation of the baseline rate of change under healthy conditions, then the energy of that mode is determined to have a valid energy mutation mode within that time window.
[0065] Through the above process of identifying spectral line drift and energy mutation modes, the changes in local modal energy distribution caused by cracks can be accurately captured in a multimodal coupling environment, and the dynamic response anomalies caused by cracks can be characterized quantitatively.
[0066] S33: Using the identified spectral line drift and energy mutation patterns as input feature vectors, a crack identification model based on threshold judgment and pattern classification is established. This crack identification model outputs Boolean judgment of crack existence and quantitative evaluation results of crack significance level of titanium alloy plate of submersible through support vector machine algorithm.
[0067] When establishing the crack identification model, the offset of spectral line drift and the rate of change of energy mutation mode are jointly used to construct the input feature vector, which is then fed into a support vector machine for training. The support vector machine uses radial basis functions as kernel functions, and by mapping the input feature vector to a high-dimensional space and constructing an optimal classification hyperplane, it achieves the classification of crack presence and non-crack states. The classification output of the support vector machine can directly provide a Boolean judgment on the existence of cracks: true when the input feature vector is located in the crack classification region, and false otherwise.
[0068] The significance level of cracks is quantified based on the numerical range of the decision function of a support vector machine. First, a first threshold and a second threshold are determined, which correspond to the boundaries of the significance level of cracks, respectively.
[0069] The decision function value is directly calculated from the decision function of the support vector machine.
[0070] When the decision function value is less than the first threshold, the significance level of the crack is determined to be "slight". When the decision function value is greater than the second threshold, the crack significance level is determined to be "severe". When the decision function value is between the first threshold and the second threshold, the crack significance level is determined to be "moderate".
[0071] The first and second thresholds are determined by the output range of the decision function of typical crack samples during the training phase. By statistically analyzing the decision function values of minor, moderate, and severe crack samples, the boundary points of their decision function values are selected as the first and second thresholds, respectively, so that the quantitative range of crack significance level can accurately reflect the degree of influence of crack on the response of the titanium alloy plate structure of the submersible.
[0072] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.
[0073] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect, characterized in that, Includes the following steps: S1: The PVDF piezoelectric sheet is attached to the test area of the titanium alloy plate of the submersible, and the corresponding piezoelectric signal sequence is obtained based on the strain response induced by the structural working load. S2: Input the piezoelectric signal sequence into the signal demixing model based on sparse dictionary and time-frequency envelope extraction, identify crack-induced non-stationary disturbance components, and extract crack indication features; S3: Construct local modal energy spectra based on crack indication characteristics, analyze their spectral drift and energy mutation modes in multimodal coupling environments, establish a crack identification model, and output the presence and significance level of cracks in the titanium alloy plates of the submersible.
2. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 1, characterized in that, S1 includes: S11: Multiple PVDF piezoelectric sheets are arranged in a predetermined array on the surface of the test area of the titanium alloy plate of the submersible, and the PVDF piezoelectric sheets are ensured to be in close contact with the surface of the titanium alloy plate to form a distributed sensing network; when the titanium alloy plate of the submersible deforms under the action of structural working load, the distributed sensing network senses the change of strain field on its surface. S12: The strain field change sensed by the distributed sensing network is converted into a distributed charge signal through the piezoelectric effect of the PVDF piezoelectric sheet. The distributed charge signal is the direct response of the PVDF piezoelectric sheet to the dynamic strain field induced by the structural working load. S13: The distributed charge signals generated by all PVDF piezoelectric elements in the distributed sensor network are synchronously acquired through a multi-channel data acquisition system, and then impedance-converted and initially amplified through a signal conditioning circuit to form a multi-channel charge signal sequence. S14: Perform preprocessing operations such as standardization and filtering on the multi-channel charge signal sequence to eliminate environmental electromagnetic interference and system background noise, and finally obtain a pure and amplitude-normalized piezoelectric signal sequence.
3. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 2, characterized in that, The predetermined array form in S11 is: a grid-like array with equal spacing on the surface of the test area of the titanium alloy plate of the submersible.
4. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 2, characterized in that, The distributed charge signal is a dynamic response signal with a frequency range of 0.1 Hz to 10 kHz.
5. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 4, characterized in that, S2 includes: S21: Construct an overcomplete sparse dictionary to characterize the health and crack states of titanium alloy plate structures in submersibles. This sparse dictionary consists of a parameterized waveform atom library and pre-acquired reference signals of crack-free states, providing a basis for sparse representation of subsequent signal demixing models. S22: Input the piezoelectric signal sequence into the signal demixing model, and use the sparse dictionary to perform adaptive sparse decomposition on the piezoelectric signal sequence to obtain the corresponding sparse representation coefficients. The sparse representation coefficients can separate the stationary components related to the normal operating load of the structure and the non-stationary components related to cracks in the piezoelectric signal sequence. S23: Based on the sparse representation coefficients, reconstruct and separate the non-stationary disturbance component mainly induced by crack initiation or propagation from the piezoelectric signal sequence. This non-stationary disturbance component is a transient or modulated signal that is distinct from background noise and normal operating strain response. S24: Perform time-frequency envelope extraction on the separated non-stationary disturbance components, obtain the time-varying energy envelope features of the non-stationary disturbance components through wavelet packet analysis algorithm, and form a time-frequency envelope matrix; S25: Calculate and extract a set of quantitative crack indication features from the time-frequency envelope matrix. The crack indication features include the envelope peak value, energy entropy, and energy concentration in a specific frequency band of the non-stationary disturbance component.
6. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 5, characterized in that, The sparse dictionary constructed in S21 has a parameterized waveform atom library that is a Gaussian frequency modulated linear frequency modulated atom library.
7. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 5, characterized in that, The wavelet packet analysis algorithm uses Daubechies wavelet basis functions to decompose the non-stationary disturbance components to complete the time-frequency envelope extraction.
8. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 7, characterized in that, S3 includes: S31: Based on the crack indication features extracted from the piezoelectric signal sequence, a local modal energy spectrum characterizing the dynamic response characteristics of the crack is constructed through feature fusion and mode decomposition techniques. This local modal energy spectrum reflects the energy distribution related to the crack under different modes. S32: In the multimodal coupling environment of the titanium alloy plate of the submersible, the local modal energy spectrum is analyzed to identify and quantify the spectral drift and energy mutation modes caused by the presence of cracks. The spectral drift is manifested as the shift of characteristic frequencies, and the energy mutation mode is manifested as the discontinuous jump of specific modal energy levels. S33: Using the identified spectral line drift and energy mutation patterns as input feature vectors, a crack identification model based on threshold judgment and pattern classification is established. This crack identification model outputs Boolean judgment of crack existence and quantitative evaluation results of crack significance level of titanium alloy plate of submersible through support vector machine algorithm.
9. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 8, characterized in that, The feature fusion and mode decomposition technique specifically involves performing principal component analysis on multiple crack indication features to complete feature fusion, and then using an empirical mode decomposition algorithm on the fused feature sequence to construct the local mode energy spectrum.
10. The method for identifying cracks in titanium alloy plates for submersibles based on the PVDF piezoelectric effect according to claim 8, characterized in that, The severity level of the crack was quantified into three discrete levels: slight, moderate, and severe.