An acoustic-force gradient collection method and sensor based on a self-excitation stress gradient field

Abstract

The application provides an acoustic-force gradient collection method and sensor based on self-excitation stress gradient field, and belongs to the technical field of collection methods and devices. The self-excitation acoustic-force gradient sensor is attached to the surface of a sealed part to realize acoustic coupling, a drive integrated piezoelectric unit emits a multi-frequency superimposed ultrasonic wave sequence, a self-excitation stress gradient field is formed at a target depth, a nonlinear modulation response is generated at a static defect interface, each scattering source component is obtained through Tucker-ICA joint framework blind source separation, a high signal-to-noise ratio defect characteristic signal is output through phase-locked amplification processing, and finally, the defect position and defect quantization result are output through the Steiner tree optimal path algorithm to fuse multi-sensor data, thereby solving the technical problem that the internal static defect of the sealed part cannot be reconstructed into a high signal-to-noise ratio and high-precision three-dimensional gradient force field through a traditional linear ultrasonic detection method.

Description

Technical Field

[0001] This invention belongs to the field of acquisition methods and devices, and specifically relates to an acoustic-force gradient acquisition method and sensor based on a self-excited stress gradient field. Background Technology

[0002] Sealing components are widely used in critical industrial structures such as pressure pipeline flange connections and bolt preload joints. Early identification of internal defects is crucial for structural safety. Traditional ultrasonic testing methods rely on linear echo signals and use single-frequency pulse transmission and delayed superposition beamforming algorithms to locate defects. In actual testing, there is a geometric misfit between the sensor probe and the irregular sealing surface, leading to acoustic impedance mismatch and acoustic energy loss. Linear echoes have low sensitivity to static interface defects such as microcracks and micro-delamination, making it difficult to distinguish defect-scattered echoes from structural noise. Under conditions of high aliasing at multiple interfaces, the resolution of single-frequency delayed superposition algorithms is severely insufficient, resulting in a high rate of missed defects. In existing technologies, traditional linear ultrasonic testing methods lack the ability to actively excite stress gradient fields, and static defect interfaces cannot generate nonlinear modulation responses. This causes effective defect characteristic signals to be submerged in broadband structural noise, and multi-channel aliased echoes cannot achieve effective blind source separation. Existing inversion algorithms also struggle to converge stably under low signal-to-noise ratio conditions, resulting in severely insufficient accuracy and reliability in reconstructing the three-dimensional gradient stress field distribution. In other words, existing technologies have a technical problem that static defects inside sealing components cannot be reconstructed with high signal-to-noise ratio and high precision using traditional linear ultrasonic testing methods. Summary of the Invention

[0003] In view of this, the present invention provides a sound-force gradient acquisition method and sensor based on a self-excited stress gradient field, which can solve the technical problem in the prior art that static defects inside sealed components cannot be reconstructed with high signal-to-noise ratio and high precision three-dimensional gradient force field through traditional linear ultrasonic detection methods.

[0004] The present invention is implemented as follows: The first aspect of the present invention provides a method for acquiring acoustic-force gradients based on a self-excited stress gradient field, comprising the following steps: A self-excited acoustic-force gradient sensor is attached to the surface of a sealed component, and the irregularly shaped acoustic lens array substrate is fully coupled with the irregularly shaped curved surface being measured to establish an acoustic coupling channel. The self-excited control module drives the integrated piezoelectric unit to emit a multi-frequency superimposed ultrasonic sequence into the sealed component, forming a self-excited stress gradient field at the target depth; The signal acquisition and analysis module performs high-speed sampling of the modulated acoustic signal and uses the Tucker-ICA joint framework to perform blind source separation of the multi-channel aliased echoes to obtain the components of each scattering source. The second harmonic component and the mixing component are extracted from each scattering source component, and after lock-in amplification, a high signal-to-noise ratio defect characteristic signal is output. Based on the comparison between the signal-to-noise ratio of the defect feature signal and the signal-to-noise ratio threshold, the acoustic-elastic coupling full waveform self-excited gradient inversion algorithm or the sealing contact stiffness gradient inversion algorithm based on acoustic admittance spectrum is selected to reconstruct the three-dimensional gradient stress field distribution. The array control host gathers the three-dimensional gradient stress field distribution of multiple self-excited acoustic-force gradient sensors, uses the Steiner tree optimal path algorithm to plan the fusion order of multi-sensor data, fuses them into an overall force field distribution map of the sealing component, and outputs the defect location and defect quantification results.

[0005] The irregular acoustic lens array substrate refers to an acoustic lens substrate pre-formed according to the surface geometry of the sealed component under test, with its bottom contour matching the irregular curved surface under test to ensure acoustic impedance matching between the integrated piezoelectric unit array and the irregular curved surface under test.

[0006] The integrated piezoelectric unit is embedded in the phase control surface array inside the irregular acoustic lens array matrix. Each chip is independently subjected to delayed phase excitation by the self-excitation control module. The sound field is focused at the target depth through the Huygens superposition principle to form a self-excited stress gradient field.

[0007] The multi-frequency superimposed ultrasonic sequence refers to a broadband ultrasonic pulse sequence synthesized by superimposing a fundamental frequency excitation signal and a high-frequency pump signal, which is used to simultaneously excite the fundamental frequency sound field and the high-order harmonic response within the sealed component.

[0008] Specifically, the Tucker-ICA joint framework involves constructing a third-order tensor from the multi-channel modulated acoustic signal, first using Tucker decomposition to extract the low-rank structure, and then applying independent component analysis to each mode factor matrix to separate each scattering source component.

[0009] Specifically, the phase-locked amplification process uses the excitation reference signal of the multi-frequency superimposed ultrasonic sequence as a reference to perform same-frequency narrowband lock-in demodulation on each scattering source component, suppress broadband structural noise and electrical noise, and extract the second harmonic component and the mixing component.

[0010] The signal-to-noise ratio threshold is determined by conducting no less than 30 sets of comparative experiments on prefabricated standard defect specimens, covering different defect types, different defect sizes, and different excitation parameters, and statistically analyzing the intersection points of the reconstruction errors of the two algorithms as a function of the signal-to-noise ratio.

[0011] Specifically, the acoustic-elastic coupling full-waveform self-excited gradient inversion algorithm is based on the joint forward modeling framework constructed by the elastic dynamics equation and the piezoelectric constitutive equation. It establishes the Green's function response matrix in the frequency domain, uses the conjugate gradient iteration method to solve the frequency domain linearized inverse scattering equations, and embeds a total variational regularization term to constrain the piecewise smoothness of the solution.

[0012] The regularization weight coefficient of the total variation regularization is determined by conducting no less than 20 independent experiments on standard defect specimens with different noise levels, using the minimum reconstruction error as the criterion, and employing the L-curve method to determine the optimal value.

[0013] Specifically, the sealed contact stiffness gradient inversion algorithm based on acoustic admittance spectrum treats the sealed contact interface as an equivalent spring-damped system, uses the Prony method to extract the resonant frequencies and damping ratios of each mode from the acoustic admittance spectrum, and employs a differential evolution global optimization algorithm to search for the optimal solution in the stiffness parameter space.

[0014] Specifically, the Steiner tree optimal path algorithm abstracts each excitation acoustic-force gradient sensor node and array control host node as vertices of a graph, uses data transmission delay as edge weight, solves for the minimum total weight connected subgraph connecting all nodes, and plans the data fusion and aggregation order.

[0015] Specifically, the acoustic admittance spectrum is calculated by the ratio of the frequency domain value of the velocity response to the frequency domain value of the excitation force at each spatial node, wherein the reference acoustic admittance standard value, the reference velocity standard value, and the reference excitation force standard value are used for normalization processing.

[0016] A second aspect of the present invention provides an acoustic-force gradient acquisition sensor based on a self-excited stress gradient field, comprising an irregularly shaped acoustic lens array substrate, an integrated piezoelectric unit, a self-excitation control module, and a signal acquisition and analysis module; the self-excitation control module is electrically connected to the integrated piezoelectric unit and is used to output an excitation signal; the integrated piezoelectric unit is embedded in a phase control surface array inside the irregularly shaped acoustic lens array substrate and is used to transmit and receive ultrasonic signals; the signal acquisition and analysis module is electrically connected to the integrated piezoelectric unit and is used to acquire, process, and analyze the received modulated acoustic wave signal.

[0017] The bottom contour of the irregular acoustic lens array substrate matches the irregular curved surface being tested, eliminating acoustic energy loss caused by geometric mismatch and ensuring acoustic impedance matching between the integrated piezoelectric unit array and the irregular curved surface being tested.

[0018] The phase control surface array is a two-dimensional piezoelectric crystal array embedded in the matrix of the irregular acoustic lens array. Each crystal is independently subjected to delayed phase excitation by the self-excitation control module, and the sound field is focused at the target depth through the Huygens superposition principle.

[0019] Among them, the number of experimental groups for verifying the signal-to-noise ratio threshold shall not be less than 30, and the number of experimental groups for determining the regularization weight coefficient shall not be less than 20, both of which cover different defect types and different excitation parameter conditions.

[0020] This invention actively drives the static defect interface to generate nonlinear modulation through a self-excited stress gradient field, and combines it with the Tucker-ICA joint framework to achieve blind source separation of multi-channel aliased echoes. Then, it extracts high signal-to-noise ratio defect feature signals through lock-in amplification, and reconstructs the three-dimensional gradient stress field distribution by adaptively selecting an inversion algorithm based on the signal-to-noise ratio. This fundamentally changes the traditional linear ultrasonic detection mode that relies on passively receiving linear scattered echoes. Because the phase control array focuses the acoustic beam, inducing a non-uniform transient stress field at the target depth, the static defect interface undergoes micro-displacement under the drive of acoustic pressure difference. This transforms the originally undetectable static defect into a dynamic acoustic characteristic signal carrying second harmonics and mixing components, overcoming the fundamental limitation of insufficient sensitivity of linear detection to static interface defects. Tucker decomposition utilizes the low-rank structure of the third-order tensor to separate time-domain, spatial-domain, and frequency-domain modes. Independent component analysis further utilizes the statistical independence of each scattering source to achieve blind source separation, overcoming the problem of indistinguishable multi-interface echoes under high aliasing conditions. Phase-locked amplification uses the excitation reference signal as a reference for narrowband locked demodulation, effectively suppressing broadband noise and ensuring high signal-to-noise ratio output of the defect characteristic signal. The adaptive switching between acoustic elastic coupling full waveform inversion and stiffness gradient inversion based on acoustic admittance spectrum ensures stable convergence and accuracy of three-dimensional gradient stress field reconstruction under high and low signal-to-noise ratio conditions, respectively. In summary, this invention solves the technical problem mentioned in the background art that high signal-to-noise ratio and high-precision three-dimensional gradient force field reconstruction of internal static defects in sealed components cannot be achieved through traditional linear ultrasonic detection methods. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a schematic diagram of the overall structure of a single acoustic-force gradient sensor unit based on a self-excited stress gradient field provided in an embodiment of this application; Figure 3 This is a cross-sectional structural diagram of the sensor unit provided in the embodiments of this application; Figure 4 This is a schematic diagram of the array layout of the acoustic-force gradient sensor for the self-excited stress gradient field provided in the embodiments of this application; Figure 5 This is a schematic diagram of the acoustic-force gradient sensor array system for a self-excited stress gradient field provided in an embodiment of this application.

[0023] The annotations in the attached figures are explained as follows: 1-Irregularly shaped acoustic lens array substrate; 2-Irregularly shaped curved surface contact layer; 3-Phase control surface array; 4-Integrated piezoelectric unit; 5-Self-excitation control module; 6-Signal acquisition and analysis module; 41-Irregularly shaped acoustic lens array substrate; 42-Flange under test; 51-Computer; 52-Bolt; 53-Acoustic-force gradient sensor; 54-Flange; 55-Sealing gasket. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0025] like Figure 1 The diagram shown is a flowchart of an acoustic-force gradient acquisition method based on a self-excited stress gradient field provided by the present invention. This method includes the following steps: S01. The self-excited acoustic-force gradient sensor is attached to the surface of the sealed component, and the irregular acoustic lens array substrate is fully coupled with the irregular curved surface to be measured to establish an acoustic coupling channel. S02. The self-excited control module drives the integrated piezoelectric unit to emit a multi-frequency superimposed ultrasonic sequence into the sealed component, forming a self-excited stress gradient field at the target depth. S03, the signal acquisition and analysis module performs high-speed sampling of the modulated acoustic signal and uses the Tucker-ICA joint framework to perform blind source separation of the multi-channel aliased echoes to obtain the components of each scattering source. S04. Extract the second harmonic component and the mixing component from each scattering source component, and output a high signal-to-noise ratio defect characteristic signal after phase-locked amplification. S05. Based on the comparison result between the signal-to-noise ratio and the signal-to-noise ratio threshold of the defect feature signal, select to execute the acoustic-elastic coupling full waveform self-excited gradient inversion algorithm or the sealing contact stiffness gradient inversion algorithm based on acoustic admittance spectrum to reconstruct the three-dimensional gradient stress field distribution. S06. The array control host gathers the three-dimensional gradient stress field distribution of multiple self-excited acoustic-force gradient sensors, uses the Steiner tree optimal path algorithm to plan the fusion order of multi-sensor data, fuses them into an overall force field distribution map of the sealing component, and outputs the defect location and defect quantification results.

[0026] Among them, the irregular acoustic lens array substrate refers to the acoustic lens substrate pre-formed according to the surface geometry of the sealed component under test. Its bottom contour matches the irregular curved surface under test, ensuring acoustic impedance matching between the integrated piezoelectric unit array and the irregular curved surface under test, and eliminating acoustic energy loss caused by geometric misfit.

[0027] A phase control surface array refers to a two-dimensional piezoelectric crystal array embedded in the matrix of an irregularly shaped acoustic lens array. Each crystal is independently subjected to delayed phase excitation by a self-excitation control module. Through the Huygens superposition principle, the sound field is focused at the target depth inside the sealed component, forming a local transient pressure gradient, i.e., a self-excited stress gradient field.

[0028] A self-excited stress gradient field refers to a spatially non-uniform transient stress field induced by a focused acoustic beam from a phase control surface array within a sealed component. This self-excited stress gradient field causes micro-displacement of the static defect interface within the sealed component under the influence of acoustic pressure difference, thereby nonlinearly modulating the incident sound wave and transforming the static defect into a dynamic acoustic characteristic signal.

[0029] Multi-frequency superimposed ultrasonic sequence refers to a broadband ultrasonic pulse sequence synthesized by superimposing a fundamental frequency excitation signal and a high-frequency pump signal, output by a self-excitation control module. Its purpose is to simultaneously excite the fundamental frequency sound field and the higher harmonic response within a sealed component, providing a signal basis for the subsequent extraction of the second harmonic component and the mixing component.

[0030] Modulated acoustic signal refers to the echo signal generated after incident ultrasonic wave is nonlinearly modulated by the interface of internal defects in a sealed component. It includes a second harmonic component and a mixing component introduced by the nonlinearity of the interface stiffness at the defect. The second harmonic component and the mixing component have significantly higher response sensitivity to static defects such as microcracks and microdelamination than the linear echo component.

[0031] The Tucker-ICA joint framework involves constructing a third-order tensor from the received multi-channel modulated acoustic signals. Tucker decomposition is then used to decompose this third-order tensor into a kernel tensor and mode factor matrices, extracting the low-rank structure of each spatial, temporal, and frequency mode. Independent component analysis (ICA) is then applied to the extracted mode factor matrices, utilizing the prior constraint of statistical independence for each scattering source component to separate the independent scattering source components at each interface and defect location. The Tucker-ICA joint framework simultaneously utilizes the low-rank and statistical independence of the signal in the time, spatial, and frequency domains, overcoming the insufficient resolution capability of classical time-delay superposition beamforming algorithms under high aliasing conditions at multiple interfaces. This enables real-time blind source separation of multi-channel modulated acoustic signals, reducing the defect false negative rate and improving defect location accuracy.

[0032] Phase-locked amplification (PLA) refers to using the excitation reference signal of the multi-frequency superimposed ultrasonic sequence output by the self-excitation control module as a reference, performing same-frequency narrowband lock-in demodulation on each scattering source component acquired by the signal acquisition and analysis module, suppressing broadband structural noise and electrical noise, and retaining only the effective components that are coherent with the frequency and harmonics of the excitation reference signal, thereby stably extracting the second harmonic component and the mixing component, and outputting a high signal-to-noise ratio defect characteristic signal.

[0033] The defect feature signal refers to the signal output after phase-locked loop amplification, which contains the amplitude and phase information of the second harmonic component and the mixing component. The signal-to-noise ratio of the defect feature signal reflects the detectability of the effective defect response under the current acquisition environment.

[0034] The signal-to-noise ratio (SNR) threshold is the critical SNR value used to determine whether the quality of the defect feature signal meets the convergence requirements of the acoustic-elastic coupling full-waveform self-excited gradient inversion algorithm. When the SNR of the defect feature signal is higher than the SNR threshold, the acoustic-elastic coupling full-waveform self-excited gradient inversion algorithm is executed; when the SNR of the defect feature signal is lower than the SNR threshold, the sealing contact stiffness gradient inversion algorithm based on acoustic admittance spectrum is executed. The SNR threshold is determined by conducting no fewer than 30 sets of comparative experiments covering different defect types, sizes, and excitation parameters on prefabricated standard defect specimens under different SNR conditions. The intersection points of the reconstruction errors of the two algorithms with the SNR are statistically analyzed, and the SNR values ​​corresponding to these intersection points are taken and verified by no fewer than 30 sets of independent experimental data.

[0035] The principle and implementation of the acoustic-elastic coupling full-waveform self-excited gradient inversion algorithm are as follows. This algorithm constructs a joint forward modeling framework based on the elastic dynamics equation and the piezoelectric constitutive equation. The self-excited stress gradient field is considered as a known transient source term. A Green's function response matrix is ​​established for each discrete frequency point in the frequency domain to describe the contribution of stress disturbance at any location in the sealed component to the modulated acoustic signal of each receiving channel. The forward-modeled synthesized waveform is subtracted from the measured modulated acoustic signal to obtain the waveform residual vector. A frequency-domain linearized inverse scattering equation set is constructed with the stress distribution in the sealed contact area as the unknown quantity. The conjugate gradient iteration method is used to solve the equations round by round. In each iteration, the stress field estimate is updated synchronously and the acoustic field residual is recalculated. Simultaneously, a total variational regularization term is embedded to constrain the piecewise smoothness of the solution, suppressing ill-conditioned oscillations. After iterative convergence, the three-dimensional gradient stress field distribution is output. The objective function formula of the total variational regularization is expressed as follows: ; in The stress distribution to be determined is given in units of . This is a reference stress standard value, in units of... The response matrix of the Green's function This is the measured modulated acoustic signal data vector, in units of... These are regularization weight coefficients, dimensionless. Reference stress field, unit: Regularization weight coefficients By conducting no fewer than 20 independent experiments on standard defect specimens with different noise levels, and using the minimum reconstruction error as the criterion, the optimal values ​​were determined using the L-curve method and then fixed for use. The acoustic-elastic coupled full-waveform self-excited gradient inversion algorithm integrates the physical constraints of the acoustic field with the inversion of the force field, making full use of the defect scattering information carried in the broadband modulated acoustic signal. It overcomes the shortcomings of insufficient information and overly smooth reconstruction results in traditional single-frequency point inversion, thereby improving the spatial resolution and amplitude accuracy of the local stress concentration area of ​​the three-dimensional gradient stress field distribution, and realizing accurate three-dimensional positioning and quantitative characterization of defects such as bolt loosening and flange delamination.

[0036] The principle and implementation of the sealed contact stiffness gradient inversion algorithm based on acoustic admittance spectrum are as follows. This algorithm treats the sealed contact interface as an equivalent spring-damped system continuously distributed along the interface. An integrated piezoelectric unit is driven by a self-excited control module to perform frequency sweep excitation over a wide frequency range. The signal acquisition and analysis module synchronously acquires the velocity response at each spatial location and calculates the acoustic admittance spectrum of each spatial node. The formula for calculating the acoustic admittance spectrum is as follows: ;in The acoustic admittance spectrum of each spatial node, in units of For reference to the acoustic admittance standard value, the unit is... The frequency domain value of the velocity response, in units of This is a reference speed standard value, in units of The frequency domain value of the excitation force, in units of For reference to the standard value of incentive, the unit is... The Prony method is used to extract the modal resonant frequencies and damping ratios from the acoustic admittance spectrum. An analytical forward mapping model between the contact stiffness parameters and the acoustic admittance spectrum is established. A differential evolution global optimization algorithm is used to search for the optimal solution in the stiffness parameter space. The objective function is the weighted least squares error between the simulated and measured acoustic admittance spectra. Finally, the local contact stiffness gradient map distributed along the sealing contact interface is inverted, which intuitively quantifies the degree of non-uniform distribution of the compression state of the flange sealing surface and outputs the three-dimensional gradient stress field distribution. The sealing contact stiffness gradient inversion algorithm based on acoustic admittance spectrum uses the broadband acoustic admittance spectrum as the detection information carrier, avoiding the sensitivity of waveform inversion to noise under low signal-to-noise ratio conditions. The differential evolution global optimization algorithm avoids local minima traps, ensuring that the inversion of the local contact stiffness gradient map remains stable and convergent under noisy environments. It has a high ability to identify local areas of reduced contact stiffness caused by bolt loosening.

[0037] The Prony method involves establishing a linear combination model of exponential functions for time-domain or frequency-domain sampled sequences, and extracting the resonant frequencies and damping ratios of each mode by solving a system of linear equations. It is a parametric spectral estimation method and is applicable to the modal decomposition of acoustic admittance spectra.

[0038] The differential evolution global optimization algorithm refers to maintaining a population of candidate solutions in the stiffness parameter space and evolving generation by generation through three operations: differential mutation, crossover, and selection. The objective function value guides the population to converge toward the global optimum. It has the characteristics of not requiring gradient information and strong resistance to local minima, and is suitable for nonlinear parameter inversion of acoustic admittance spectrum mapping models.

[0039] The Steiner tree optimal path algorithm is applied in step S06 as follows: Each self-excited acoustic-force gradient sensor node and the array control host node are abstracted as vertices of a graph. Using data transmission delay as the edge weight, the Steiner tree connecting all self-excited acoustic-force gradient sensor nodes and the array control host node is solved, i.e., the connected subgraph with the minimum total weight. The resulting path is the optimal convergence order for data fusion, minimizing the total transmission delay of multi-sensor synchronous data acquisition and ensuring the real-time performance of array-level force field reconstruction. The Steiner tree problem belongs to the classic graph theory combinatorial optimization problem and is suitable for multi-sensor data fusion path planning optimization in step S06.

[0040] like Figure 2 As shown, the second aspect of this invention provides an acoustic-force gradient acquisition sensor based on a self-excited stress gradient field, comprising an irregularly shaped acoustic lens array substrate 1, an integrated piezoelectric unit, a self-excitation control module, and a signal acquisition and analysis module. These components work together to achieve a complete process of sound field focusing, gradient force field self-excitation, defect signal modulation, force field reconstruction, and defect identification. The self-excitation control module is electrically connected to the integrated piezoelectric unit and is used to output an excitation signal. The integrated piezoelectric unit is embedded in a phase control surface array within the irregularly shaped acoustic lens array substrate and is used to transmit and receive ultrasonic signals. The signal acquisition and analysis module is electrically connected to the integrated piezoelectric unit and is used to acquire, process, and analyze the received modulated acoustic wave signal.

[0041] When using it, first put Figure 2 The irregularly shaped curved contact layer of the sensor unit is tightly bonded to the surface of the sealed component under test, and zero-gap acoustic coupling is achieved using a coupling agent to eliminate energy loss of acoustic waves at the interface. For complex sealing structures, a sensor array system is used for distributed bonding. The array control host or self-excitation control module outputs a phase-modulated high-frequency pulse sequence to the integrated piezoelectric unit, driving the piezoelectric unit to vibrate in sync. Inside the sensor, a multi-point focused sound field is formed through an array of acoustic lenses and phase control surfaces, generating resonance within the sensor, which in turn excites surface pulse acoustic waves on the surface of the test component. The surface pulse acoustic waves propagate along the interior of the test component, inducing an instantaneous self-excited stress gradient field within the material. This force field can cover defect areas such as bolt axial direction and the interface between flanges and gaskets.

[0042] At this point, the self-excited stress gradient field interacts nonlinearly with the existing sealing defects inside the test component. The defects modulate the gradient force field, causing the reflected sound wave to carry the mechanical characteristics of the defect, forming a modulated sound wave. The integrated piezoelectric unit then receives the modulated sound wave and converts it into an electrical signal. The signal acquisition and analysis module filters and amplifies the electrical signal, extracting key characteristic parameters such as the wake time shift of the reflected echo and the sound wave reflection coefficient. Finally, the signal acquisition and analysis module substitutes these characteristic parameters into the model to invert the contact stiffness distribution and force field gradient changes inside the test component, achieving high-precision reconstruction of the gradient force field inside the sealing component. Based on the force field distribution characteristics, the type, location, and extent of the sealing defect are quantitatively identified, and the detection results are output on the display terminal.

[0043] Optionally, the present invention also provides a method for implementing a sound-force gradient acquisition system based on a self-excited stress gradient field using a computer. The computer is equipped with a readable storage medium that stores program instructions, which execute the above-described method when the computer is run.

[0044] The specific implementation of step S01 is as follows: First, based on the surface geometry of the sealed component under test, a non-circular acoustic lens array substrate is pre-prepared to match it. The bottom contour of the substrate strictly follows the three-dimensional shape of the non-circular curved surface under test, ensuring a zero-gap geometric match between the bottom surface of the substrate and the surface under test. During bonding, an acoustic coupling agent is applied to the contact interface to eliminate the abrupt change in acoustic impedance caused by the air gap, ensuring efficient conduction of acoustic wave energy from the integrated piezoelectric unit array to the interior of the sealed component, establishing a stable acoustic coupling channel. The core purpose of this step is to eliminate acoustic energy loss caused by geometric misfit, providing an effective acoustic conduction basis for the subsequent formation of a self-excited stress gradient field.

[0045] The specific implementation of step S02 is as follows: The self-excitation control module independently applies a delayed phase-modulated excitation signal to each two-dimensional piezoelectric crystal in the phase control surface array. The sound waves emitted by each crystal are coherently superimposed and focused at the target depth inside the sealed component according to the Huygens superposition principle, forming a spatially uneven transient stress field, namely the self-excited stress gradient field. The excitation signal is a multi-frequency superimposed ultrasonic sequence synthesized by superimposing a fundamental frequency excitation signal and a high-frequency pump signal. The frequency combination of the two results in the simultaneous existence of a fundamental frequency sound field and high-order harmonic excitation conditions inside the sealed component, providing a physical basis for the subsequent generation of nonlinear modulation signals. The self-excited stress gradient field induces a local transient pressure gradient inside the sealed component. This gradient causes the static defect interface to undergo micro-displacement under the drive of the acoustic pressure difference, transforming the static defect into a dynamic nonlinear acoustic response source.

[0046] The specific implementation of step S03 is as follows: The signal acquisition and analysis module performs high-speed analog-to-digital sampling on the modulated acoustic signal received by the integrated piezoelectric unit, arranging the multi-channel sampled data into a third-order tensor, with its three modes corresponding to the spatial channel dimension, time dimension, and frequency dimension, respectively. The Tucker-ICA joint framework first performs Tucker decomposition on this third-order tensor, decomposing it into a kernel tensor and each mode factor matrix, extracting the low-rank structure in the time, spatial, and frequency domains to reduce the redundancy of the aliased signal. Subsequently, independent component analysis is applied to each mode factor matrix, and the separation matrix is ​​solved using the prior constraint that each scattering source component is statistically independent, separating the multi-channel aliased interface scattering echo into independent components of each scattering source. The Tucker-ICA joint framework utilizes low-rank and statistical independence in three dimensions simultaneously, overcoming the inherent defect of insufficient resolution capability of traditional time-delay superposition beamforming algorithms under high aliasing conditions at multiple interfaces, significantly reducing the defect false detection rate and improving defect location accuracy.

[0047] The specific implementation of step S04 is as follows: For each scattering source component obtained in step S03, the second harmonic component and the mixing component are extracted in the frequency domain. These two types of components are direct products of the modulation of the incident sound wave by the nonlinear stiffness of the defect interface. Their amplitude and phase carry the mechanical characteristic information of the defect, and their response sensitivity to static defects such as microcracks and micro-delamination is significantly higher than that of the linear echo component. The lock-in amplification process uses the multi-frequency superimposed ultrasonic sequence excitation reference signal output by the self-excitation control module as the phase reference, and performs same-frequency narrowband lock-in demodulation on each scattering source component. The demodulation bandwidth is strictly limited to a narrow band range coherent with the excitation frequency and its harmonics, thereby effectively suppressing broadband structural noise and electrical noise, and outputting a defect characteristic signal with a high signal-to-noise ratio. The signal-to-noise ratio of the defect characteristic signal reflects the detectability of the effective defect response under the current acquisition environment and serves as the criterion input for the subsequent inversion algorithm selection.

[0048] The specific implementation of step S05 is as follows: The signal-to-noise ratio (SNR) of the defect feature signal output in step S04 is compared with a pre-calibrated SNR threshold. The SNR threshold is determined by no fewer than 30 sets of comparative experiments on prefabricated standard defect specimens, covering different defect types, different defect sizes, and different excitation parameters. The crossover point of the reconstruction error of the two algorithms with the change of SNR is taken, and the value is fixed after verification by no fewer than 30 sets of independent experiments. When the SNR is higher than the SNR threshold, the acoustic-elastic coupling full waveform self-excited gradient inversion algorithm is executed: This algorithm establishes a Green's function response matrix in the frequency domain based on the elastic dynamics equation and the piezoelectric constitutive equation, constructs a frequency domain linearized inverse scattering equation set with the stress distribution in the sealed contact area as the unknown, solves it round by round using the conjugate gradient iteration method, and embeds a total variational regularization term to constrain the piecewise smoothness of the solution. The regularization weight coefficient is determined by the L-curve method through no fewer than 20 sets of independent experiments. After the iteration converges, the three-dimensional gradient stress field distribution is output. When the signal-to-noise ratio (SNR) is below the SNR threshold, a sealing contact stiffness gradient inversion algorithm based on acoustic admittance spectrum is executed. This algorithm treats the sealing contact interface as an equivalent spring-damped system, collects the velocity response of each spatial node through frequency sweep excitation, calculates the acoustic admittance spectrum, extracts the resonant frequencies and damping ratios of each mode using the Prony method, establishes an analytical forward mapping model between the contact stiffness parameters and the acoustic admittance spectrum, and then uses a differential evolution global optimization algorithm with weighted least squares error as the objective function to search for the optimal stiffness parameters, inverting the local contact stiffness gradient distribution, i.e., the three-dimensional gradient stress field distribution. The differential evolution global optimization algorithm evolves generationally through three operations: differential mutation, crossover, and selection. It does not require gradient information, has strong resistance to local minima, and ensures stable convergence of the inversion under low SNR conditions.

[0049] The specific implementation of step S06 is as follows: The array control host abstracts each excitation-type acoustic-force gradient sensor node and its own node as vertices of a graph. Using the data transmission delay between nodes as edge weights, the Steiner tree optimal path algorithm is applied to solve for the minimum total weight connected subgraph connecting all nodes, obtaining the optimal convergence order for multi-sensor data fusion and minimizing the total transmission delay of multi-sensor synchronous data acquisition. According to the obtained order, the array control host sequentially converges the three-dimensional gradient stress field distribution output by each sensor, performs spatial registration and weighted fusion of each local force field, and generates an overall force field distribution map of the sealing component. Based on the location and amplitude of the stress concentration area in the overall force field distribution map, combined with preset defect criteria, the defect location coordinates and defect quantification results are output, realizing the global location and quantitative characterization of defects such as bolt loosening and flange delamination.

[0050] It should be noted that the key technologies of this invention include: an active nonlinear excitation mechanism of a self-excited stress gradient field, a multidimensional blind source separation method within the Tucker-ICA joint framework, and a signal-to-noise ratio adaptive dual-track inversion system. The self-excited stress gradient field induces spatially non-uniform transient stress at the target depth through Huygens focusing of a phase control surface array. This causes the static defect interface, which originally did not produce detectable echoes, to undergo micro-displacement due to pressure gradient drive, actively generating a nonlinear modulation response. This overcomes the physical limit of insufficient sensitivity of traditional linear ultrasonic testing to static interface defects. The Tucker-ICA joint framework organizes multi-channel echo signals into a third-order tensor, simultaneously applying low-rank constraints and statistical independence constraints in the time, spatial, and frequency domains. This mathematically guarantees the uniqueness of blind source separation for multi-interface aliasing signals, fundamentally overcoming the inherent limitations of single-dimensional signal processing methods under high aliasing conditions. The signal-to-noise ratio adaptive dual-track inversion system utilizes a quality dynamic matching algorithm based on defect characteristic signals. Under high signal-to-noise ratio conditions, the acoustic-elastic coupling full-waveform inversion fully leverages broadband scattering information to obtain high spatial resolution 3D reconstruction. Under low signal-to-noise ratio conditions, the acoustic admittance spectrum inversion uses broadband modal information as a carrier and employs differential evolution global optimization to avoid local minima, ensuring stable convergence in noisy environments. These three key technologies work synergistically to form a complete technical closed loop from active excitation and signal separation to adaptive inversion, systematically improving the sensitivity, resolution, and noise robustness of the 3D gradient force field reconstruction of static defects inside sealed components.

[0051] It should be noted that this invention also solves the following technical problem: In multi-sensor array-level detection scenarios, a large amount of three-dimensional gradient stress field data generated by the synchronous acquisition of various sensors needs to be converged to the array control host for fusion in real time. As the number of sensor nodes increases, the planning of the data transmission path directly affects the real-time performance of the overall force field reconstruction. Traditional multi-sensor data fusion usually uses fixed broadcast or polling methods to converge data, without optimizing the transmission path, resulting in a significant increase in total transmission delay as the number of nodes increases, making it difficult to meet the requirements of array-level real-time reconstruction. This invention abstracts each sensor node and the array control host node as vertices of a graph, uses the data transmission delay as the edge weight, and applies the Steiner tree optimal path algorithm to solve for the minimum total weight connected subgraph, obtaining the theoretically optimal data fusion convergence order and minimizing the total transmission delay of the synchronously acquired data from multiple sensors. The optimal connected subgraph structure of the Steiner tree problem ensures that the total edge weight is minimized while satisfying the connectivity constraints of all nodes, effectively guaranteeing the real-time performance of array-level three-dimensional force field reconstruction even as the number of sensor nodes increases, thus solving the technical problem of minimizing the total transmission delay in the multi-sensor array data fusion path planning.

[0052] Specifically, the principle of this invention is as follows: The core reason why this invention can solve the above-mentioned technical problems lies in transforming the passive linear receiving mode into an active nonlinear excitation-modulation-demodulation closed-loop detection mode, and introducing an algorithmic framework that matches the physical mechanism in both signal processing and inversion reconstruction stages. The formation of the self-excited stress gradient field depends on the Huygens superposition focusing principle of the phase control surface array. The local transient pressure gradient generated by focusing causes a nonlinear response at the static defect interface. This physical mechanism ensures the excitationability of the defect characteristic signal. The Tucker-ICA joint framework superimposes statistical independence constraints on the basis of tensor low-rank decomposition, which mathematically guarantees the uniqueness of blind source separation of multidimensional aliased signals, conforming to the signal structure characteristics of multi-interface scattering problems. The narrowband coherent demodulation of lock-in amplification strictly corresponds to the frequency of the excitation signal, thus eliminating broadband interference components unrelated to the defect in principle. The acoustic-elastic coupling full-waveform inversion establishes a linearized inverse scattering equation using the elastic dynamics Green's function and the unknown force field. The total variational regularization constraint ensures the piecewise smoothness of the solution, providing theoretical assurance for the iterative convergence of 3D reconstruction under high signal-to-noise ratio conditions. Meanwhile, the stiffness gradient inversion based on acoustic admittance spectrum uses broadband modal information as its carrier, and the differential evolution global optimization algorithm avoids local minima, ensuring stable reconstruction under low signal-to-noise ratio conditions. The adaptive switching logic of the two algorithms conforms to the matching relationship between signal quality and the applicable domain of the algorithm, and the overall scheme is logically complete and self-consistent.

[0053] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0054] The specific implementation of step S01 is as follows: the irregular acoustic lens array substrate of the self-excited acoustic-force gradient sensor is attached to the surface of the sealing component. The bottom contour of the irregular acoustic lens array substrate is pre-formed according to the geometry of the irregular curved surface to be measured, so as to ensure the acoustic impedance matching between the integrated piezoelectric unit array and the irregular curved surface to be measured, eliminate the acoustic energy loss caused by geometric misfit, and establish an acoustic coupling channel.

[0055] The specific implementation of step S02 is as follows: The self-excitation control module drives each piezoelectric crystal embedded in the phase control surface array inside the irregularly shaped acoustic lens array substrate. Each crystal independently applies delayed phase excitation, and the sound field is focused at the target depth through the Huygens superposition principle, forming a self-excited stress gradient field. The excitation signal is synthesized by superimposing the fundamental frequency excitation signal and the high-frequency pump signal to form a multi-frequency superimposed ultrasonic sequence, the time domain expression of which is as follows: ; In the formula, The excitation signal is a multi-frequency superimposed ultrasonic sequence, with units of . This is the reference voltage standard value, in units of , These represent the amplitudes of the fundamental frequency component and the high-frequency pump component, respectively, in units of... , These are the fundamental excitation frequency and the high-frequency pump frequency, respectively, in units of... , These represent the initial phases of the fundamental frequency component and the high-frequency pump component, respectively, in units of... For time, the unit is The delay phase excitation of each chip is designed according to the array aperture and focusing depth, so that the focused sound beam forms a transient stress field with non-uniform spatial distribution at the target depth, which drives the static defect interface to produce micro-displacement of the interface, thereby generating nonlinear modulation of the incident sound wave.

[0056] The specific implementation of step S03 is as follows: the signal acquisition and analysis module performs high-speed sampling of the modulated acoustic signal to obtain multi-channel discrete time-domain data. One receiving channel, Each time sampling point The multi-channel data at each frequency point is constructed as a third-order tensor. The third-order tensor is decomposed into a kernel tensor and mode factor matrices using Tucker decomposition, as expressed in the following formula: ; In the formula, The original multi-channel sampled signal tensor, in units of This is the reference voltage standard value, in units of For kernel tensors, the unit is . , , These are the factor matrices for spatial, temporal, and frequency modes, respectively, and are all dimensionless matrices. , , These are the reference standard values ​​corresponding to the spatial, temporal, and frequency mode factor matrices, respectively, all of which are dimensionless quantities. , , This represents the low-rank dimension of each mode, typically taken as 30% to 50% of the actual number of channels. For residual tensors, the unit is . Indicates along the first Modular product of patterns Independent component analysis is applied to the obtained mode factor matrices. With the statistical independence of each scattering source component as a constraint, the independent scattering source components of each interface and each defect location are separated, thereby realizing blind source separation of multi-channel modulated acoustic signals and reducing the defect false detection rate.

[0057] The specific implementation of step S04 is as follows: extract the second harmonic component (frequency ) from each scattering source component. ) and mixing components (frequency of The signal is referenced by the excitation reference signal output from the self-excited control module and then subjected to phase-locked amplification. The phase-locked amplification process performs narrowband lock-in demodulation on each scattering source component and the reference signal at the same frequency, suppressing broadband structural noise and electrical noise, and outputting a high signal-to-noise ratio defect characteristic signal. After demodulation, the second harmonic amplitude of each scattering source component is... With mixing amplitude Extracted by the following formula: ; ; In the formula, The amplitude of the second harmonic is given in units of 1. This represents the mixing amplitude, in units of... For reference sound pressure level standards, the unit is... The time-domain signal of the scattering source component after blind source separation is given in units of . The length of the integration time window, in units of imaginary unit This is the integral variable, and the unit is... . and The amplitude and phase information together constitute the defect feature signal, and its signal-to-noise ratio reflects the detectability of the effective defect response.

[0058] The specific implementation of step S05 is as follows: Based on the comparison result between the signal-to-noise ratio (SNR) of the defect feature signal and the SNR threshold, the corresponding inversion algorithm is selected for execution. The SNR threshold is determined by statistically analyzing the intersection points of the reconstruction errors of the two algorithms with the SNR through comparative experiments on no less than 30 sets of prefabricated standard defect specimens covering different defect types, sizes, and excitation parameters. This threshold is then used after verification through no less than 30 independent experiments. When the SNR is higher than the SNR threshold, the acoustic-elastic coupling full-waveform self-excited gradient inversion algorithm is executed. Its total variational regularization objective function is as follows: ; In the formula, The three-dimensional stress distribution is to be determined, with units of . This is a reference stress standard value, in units of... The Green's function response matrix is ​​a dimensionless matrix, with each element being a normalized Green's function response coefficient, obtained through forward modeling. Specifically, a joint forward modeling framework of elastic dynamics and piezoelectric constitutive equations is established for each discrete frequency point in the frequency domain. The frequency domain response of unit stress disturbance at any location within the sealed component to each receiving channel is calculated, and the results are aggregated into a response matrix. Normalization This is the measured modulated acoustic signal data vector, in units of... The regularization weighting coefficient is dimensionless and its optimal value is determined by using the L-curve method on no fewer than 20 sets of noisy standard defect specimens, and then fixed for use. Reference stress field, unit: Spatial gradient operator for norm for norm These are the spatial coordinates within the sealing component, in units of... The objective function consists of two terms: the first term is the normalized data fitting term, and the second term is the normalized total variation regularization term, which constrains the piecewise smoothness of the solution and suppresses ill-conditioned oscillations. A conjugate gradient iterative method is used to solve the problem round by round. In each iteration, the stress field estimate is updated synchronously and the acoustic field residual is recalculated. After convergence, the three-dimensional gradient stress field distribution is output.

[0059] When the signal-to-noise ratio is lower than the signal-to-noise ratio threshold, the sealing contact stiffness gradient inversion algorithm based on the acoustic admittance spectrum is executed. The formula for calculating the acoustic admittance spectrum is as follows: ; In the formula, The acoustic admittance spectrum of each spatial node, in units of For reference to the acoustic admittance standard value, the unit is... The frequency domain value of the velocity response, in units of This is a reference speed standard value, in units of The frequency domain value of the excitation force, in units of For reference to the standard value of incentive, the unit is... For frequency variables, the unit is... .

[0060] The Proni method is used to extract the modal resonant frequencies and damping ratios from the acoustic admittance spectrum. The Proni method establishes the following exponential function linear combination model for the acoustic admittance spectrum: ; In the formula, This represents the total number of modes, with an empirical value of 5 to 20. For the first Complex amplitude coefficients of the first mode, in units of For the first First-order modal resonance frequency, in units of For the first First-order modal damping ratio, dimensionless For the first The undamped natural angular frequency of the first mode, in units of This is the reference frequency standard value, in units of The default value is 1. The modal index has a value of [value]. Each parameter will be determined by... Substituting each frequency sampling point into the above formula, we establish... The solution is obtained by directly solving the system of linear equations.

[0061] An analytical forward modeling model is established between contact stiffness parameters and acoustic admittance spectrum. The weighted least squares error between the simulated and measured acoustic admittance spectra is used as the objective function. A differential evolution global optimization algorithm is employed to search for the optimal solution in the stiffness parameter space. The objective function is as follows: ; In the formula, The weighted least squares error is dimensionless. Total number of discrete frequency points For the first Discrete frequency points, in units of For the first The weighting coefficients for each frequency point are dimensionless, and their empirical values ​​are the normalized signal-to-noise ratio values ​​for each frequency point. The reference weight standard value is dimensionless and defaults to 1. The simulated acoustic admittance spectrum values ​​are in units of The measured acoustic admittance spectrum value is given in units of... The discrete frequency point index has a value of [value]. .

[0062] The differential evolution global optimization algorithm maintains the population size in the stiffness parameter space as follows: candidate solution set ,in For the first The generation There are candidate solution vectors, in units of _ . The differential mutation operation is as follows: ; In the formula, The mutation vector, in units of For reference stiffness standard values, the unit is... , , Three distinct and not equal to 3 randomly selected from the current population The candidate solution vector, in units of This is the difference scaling factor, dimensionless, with an empirical value of 0.5–0.9. This is the individual index of the population, with a value of [value]. This is an iterative algebra. The crossover operation is as follows: ; In the formula, The first crossover test vector Each component, in units of The first of the variant vectors Each component, in units of The first candidate solution Each component, in units of For interval Uniformly random numbers, dimensionless This represents the crossover probability, which is dimensionless and has an empirical value of 0.1–0.9. Randomly selected dimension index This refers to the component indices of the stiffness parameter vector. The selection process is as follows: ; In the formula, For the first The generation There are candidate solution vectors, in units of _ . The test vector is in units of The objective function is the aforementioned weighted least squares error. The above three steps are iterated one after another until the objective function converges, and the local contact stiffness gradient map distributed along the sealed contact interface is inverted, outputting the three-dimensional gradient stress field distribution.

[0063] The specific implementation of step S06 is as follows: The array control host gathers the three-dimensional gradient stress field distribution of multiple self-excited acoustic-force gradient sensors, abstracts each sensor node and the array control host node as vertices of a graph, uses the data transmission delay as the edge weight, and uses the Steiner tree optimal path algorithm to solve the minimum total weight connected subgraph connecting all nodes, determines the optimal convergence order of multi-sensor data fusion, minimizes the total transmission delay of multi-sensor synchronous data acquisition, ensures the real-time performance of array-level force field reconstruction, and outputs the overall force field distribution map of the sealing component and the defect location and defect quantification results after fusion.

[0064] To better understand and implement this invention, a specific application scenario of the invention is provided below as Example 2: To verify the effectiveness of the invention, technicians set up a test environment and conducted a sealing defect detection experiment on the flange bolt connection structure of an industrial pressure pipeline. This systematically verified the detection accuracy and reliability of the acoustic-force gradient acquisition method based on a self-excited stress gradient field described in this invention. The tested object was a flange sealing assembly containing bolts with different preload states. Five gradient groups were preset, with five parallel samples in each group, totaling 25 groups of bolt samples with known actual preload forces. This covered four typical intervals: normal sealing, critical loosening, significant loosening, and severe loosening, thus forming a complete test sample set.

[0065] First, such as Figure 2As shown, the irregularly shaped curved contact layer of the sensor unit is tightly bonded to the surface of the sealed component being tested, and zero-gap acoustic coupling is achieved using a couplant, eliminating interface acoustic wave energy loss. For the complex sealing structure of this flange sealing assembly, a sensor array system is used for distributed bonding, as shown in the sensor array layout. Figure 4 As shown, the sensor nodes are evenly arranged according to the geometric distribution of the flange sealing surface, ensuring acoustic coverage of the entire sealing surface area. The array control host outputs a phase-modulated high-frequency pulse sequence to the integrated piezoelectric unit through the self-excitation control module, driving the piezoelectric unit to vibrate in sync. Inside the sensor, a multi-point focused sound field is formed through the acoustic lens phase control surface array, generating resonance within the sensor and thus exciting surface pulse sound waves on the surface of the workpiece. The surface pulse sound waves propagate along the interior of the workpiece, inducing an instantaneous self-excited stress gradient field within the material. This force field covers defect areas such as the bolt axial direction and the flange-gasket interface.

[0066] The self-excited stress gradient field interacts nonlinearly with existing sealing defects inside the test piece. The defects modulate the gradient force field, causing the reflected sound waves to carry the mechanical characteristics of the defects, forming modulated sound waves. An integrated piezoelectric unit receives the modulated sound waves and converts them into electrical signals. The signal acquisition and analysis module filters and amplifies the electrical signals, then uses the Tucker-ICA joint framework to perform blind source separation on the multi-channel aliased echoes, extracting each scattering source component. After lock-in amplification, the second harmonic component and mixing component are extracted, outputting a high signal-to-noise ratio defect characteristic signal. For example... Figure 3 As shown, the cross-sectional structure of the sensor unit clearly demonstrates the synergistic relationship between the phase control surface array and the acoustic lens substrate. The phase control surface array is embedded inside the irregularly shaped acoustic lens array substrate, with each crystal independently controlled to ensure the spatial accuracy of the focused sound beam. The signal acquisition and analysis module substitutes the characteristic parameters into the inversion model and adaptively selects the inversion algorithm based on the comparison between the signal-to-noise ratio and the signal-to-noise ratio threshold. It reconstructs the contact stiffness distribution and force field gradient changes inside the tested component, achieving high-precision reconstruction of the gradient force field inside the sealing component. Based on the force field distribution characteristics, it quantitatively identifies the type, location, and degree of sealing defects and outputs the detection results on the display terminal. The signal acquisition and data fusion process of the entire array system is as follows: Figure 5 As shown, the array control host uses the Steiner tree optimal path algorithm to plan the data aggregation order of each sensor node, minimize the total transmission delay, and ensure real-time reconstruction.

[0067] This experiment was conducted with a torque coefficient of 0.12 and a bolt diameter of 27mm, using the formula... The tightening torque corresponding to each preload is calibrated in reverse. The actual preload and corresponding tightening torque of each group are shown in Table 1.

[0068] Table 1. Preset Actual Preload and Corresponding Tightening Torque for Each Group

[0069] All 25 parallel samples were tested using a self-excited stress gradient field acoustic-force gradient sensor, ensuring consistent testing conditions for each sample, including coupling layer thickness, excitation parameters, and ambient temperature. After testing, the sensor-detected preload for each sample was compared with the preset actual preload, and the absolute and relative errors were calculated. Detailed comparison results for each sample are shown in Table 2.

[0070] Table 2 Comparison of Preload Test Results for 25 Groups of Samples

[0071] As shown in Table 2, the absolute errors of all 25 sample groups were within the range of 0.5–0.9 MPa, fully meeting the preset detection accuracy requirements. No samples exceeded the error limit. The overall average absolute error was 0.68 MPa, and the overall average relative error was 3.76%, indicating stable detection accuracy. The error statistics for each group are summarized in Table 3.

[0072] Table 3 Summary of Error Statistics for Each Group

[0073] As shown in Table 3, the average relative error of the normal sealing range (groups 1 and 2) is relatively small. This is because the contact stiffness is stable in this range, the modal characteristics of the acoustic admittance spectrum are clear, and the force field reconstruction accuracy is high. The critical loosening range (group 3) has the smallest error. The linear correlation between the preload and contact stiffness is optimal in this range, making the convergence condition of the inversion model most favorable. The average relative error of the obviously loosening and severely loosening ranges (groups 4 and 5) is about 5.6%. This is because the reflection and scattering path of the sound wave at the interface becomes more complex after the gap at the bolt contact interface increases, and the degree of echo aliasing increases. However, the multidimensional blind source separation capability of the Tucker-ICA joint framework can still effectively separate the components of each scattering source, and the error is still within the acceptable range for engineering.

[0074] The technological advancements brought by this invention compared to traditional methods are reflected in the following aspects. Traditional linear ultrasonic testing relies on passively receiving linearly scattered echoes. Static defect interfaces do not produce nonlinear modulation responses under conditions without external stress gradient excitation, and the defect characteristic signals are submerged in structural noise. This invention, through a phase-controlled surface array focusing the acoustic beam, actively induces a self-excited stress gradient field at the target depth, causing the static defect interface to generate micro-displacements driven by the pressure gradient. From a physical mechanism perspective, this transforms undetectable static defects into detectable dynamic nonlinear responses, overcoming the fundamental limitation of insufficient sensitivity of traditional methods to static interface defects. Traditional single-frequency delay superposition beamforming algorithms lack multi-dimensional joint separation capabilities under conditions of high aliasing across multiple interfaces. However, the Tucker-ICA joint framework simultaneously applies low-rank constraints and statistical independence constraints in the time, spatial, and frequency domains, ensuring the mathematical uniqueness of blind source separation of aliased echoes from multiple interfaces, significantly improving defect location accuracy and reducing the false negative rate. The signal-to-noise ratio adaptive dual-track mechanism in the inversion process enables the detection system to match the inversion algorithm most suitable for the current signal quality under different noise environments. This overcomes the problem of convergence performance degradation of traditional single inversion methods when noise conditions change, and systematically ensures the reliability of three-dimensional gradient stress field reconstruction.

[0075] It should be noted that the variables involved in this invention are explained in detail in Tables 4 and 5.

[0076] Table 4. Variable Explanation Table (Part 1)

[0077] Table 5. Variable Explanation Table (Part Two)

[0078] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for acquiring acoustic-mechanical gradients based on a self-excited stress gradient field, characterized in that, Includes the following steps: A self-excited acoustic-force gradient sensor is attached to the surface of a sealed component, and the irregularly shaped acoustic lens array substrate is fully coupled with the irregularly shaped curved surface being measured to establish an acoustic coupling channel. The self-excited control module drives the integrated piezoelectric unit to emit a multi-frequency superimposed ultrasonic sequence into the sealed component, forming a self-excited stress gradient field at the target depth; The signal acquisition and analysis module performs high-speed sampling of the modulated acoustic signal and uses the Tucker-ICA joint framework to perform blind source separation of the multi-channel aliased echoes to obtain the components of each scattering source. The second harmonic component and the mixing component are extracted from each scattering source component, and after lock-in amplification, a high signal-to-noise ratio defect characteristic signal is output. Based on the comparison between the signal-to-noise ratio of the defect feature signal and the signal-to-noise ratio threshold, the acoustic-elastic coupling full waveform self-excited gradient inversion algorithm or the sealing contact stiffness gradient inversion algorithm based on acoustic admittance spectrum is selected to reconstruct the three-dimensional gradient stress field distribution. The array control host gathers the three-dimensional gradient stress field distribution of multiple self-excited acoustic-force gradient sensors, uses the Steiner tree optimal path algorithm to plan the fusion order of multi-sensor data, fuses them into an overall force field distribution map of the sealing component, and outputs the defect location and defect quantification results.

2. The acoustic-force gradient acquisition method based on a self-excited stress gradient field according to claim 1, characterized in that, The irregular acoustic lens array substrate refers to an acoustic lens substrate pre-formed according to the surface geometry of the sealed component under test, with its bottom contour matching the irregular curved surface under test to ensure acoustic impedance matching between the integrated piezoelectric unit array and the irregular curved surface under test.

3. The acoustic-force gradient acquisition method based on a self-excited stress gradient field according to claim 2, characterized in that, The integrated piezoelectric unit is embedded in the phase control surface array inside the irregular acoustic lens array matrix. Each chip is independently subjected to delayed phase excitation by the self-excitation control module. The sound field is focused at the target depth through the Huygens superposition principle, forming a self-excited stress gradient field.

4. The acoustic-force gradient acquisition method based on a self-excited stress gradient field according to claim 3, characterized in that, The aforementioned multi-frequency superimposed ultrasonic sequence refers to a broadband ultrasonic pulse sequence synthesized by superimposing a fundamental frequency excitation signal and a high-frequency pump signal, which is used to simultaneously excite the fundamental frequency sound field and the high-order harmonic response within a sealed component.

5. The acoustic-force gradient acquisition method based on a self-excited stress gradient field according to claim 4, characterized in that, The Tucker-ICA joint framework specifically involves constructing a third-order tensor from the multi-channel modulated acoustic signal, first using Tucker decomposition to extract the low-rank structure, and then applying independent component analysis to each mode factor matrix to separate the scattering source components.

6. The acoustic-mechanical gradient acquisition method based on a self-excited stress gradient field according to claim 5, characterized in that, The phase-locked amplification process specifically uses the excitation reference signal of the multi-frequency superimposed ultrasonic sequence as a reference to perform same-frequency narrowband lock-in demodulation on each scattering source component, suppress broadband structural noise and electrical noise, and extract the second harmonic component and the mixing component.

7. A self-excited acoustic-force gradient sensor, characterized in that, It includes an irregularly shaped acoustic lens array substrate, an integrated piezoelectric unit, a self-excitation control module, and a signal acquisition and analysis module. The self-excitation control module is electrically connected to the integrated piezoelectric unit and is used to output an excitation signal. The integrated piezoelectric unit is embedded in the phase control surface array inside the irregularly shaped acoustic lens array substrate and is used to transmit and receive ultrasonic signals. The signal acquisition and analysis module is electrically connected to the integrated piezoelectric unit and is used to acquire, process, and analyze the received modulated acoustic wave signal.

8. The self-excited acoustic-force gradient sensor according to claim 7, characterized in that, The bottom contour of the irregular acoustic lens array substrate matches the irregular curved surface being tested, eliminating acoustic energy loss caused by geometric mismatch and ensuring acoustic impedance matching between the integrated piezoelectric unit array and the irregular curved surface being tested.

9. The self-excited acoustic-force gradient sensor according to claim 8, characterized in that, The phase control surface array is a two-dimensional piezoelectric crystal array embedded in the matrix of the irregular acoustic lens array. Each crystal is independently subjected to delayed phase excitation by the self-excitation control module, and the sound field is focused at the target depth through the Huygens superposition principle.

10. The self-excited acoustic-force gradient sensor according to claim 9, characterized in that, The number of experimental groups for verifying the signal-to-noise ratio threshold shall not be less than 30, and the number of experimental groups for determining the regularization weight coefficient shall not be less than 20, both of which shall cover different defect types and different excitation parameter conditions.

Images