Multi-field coupling hydrogen storage bottle internal defect three-dimensional reconstruction method and system

Abstract

The invention discloses a multi-field coupling hydrogen storage bottle internal defect three-dimensional reconstruction method and system, and the method comprises the steps: deploying a multi-modal data collection and processing architecture system, and carrying out the collection and processing of multi-source data; a multi-field coupling model is constructed, and mixed strategy directional order reduction solution is carried out; extracting geometric feature data of the conversion defect, and reversely evolving defect parameters; a three-dimensional visual model is generated, and residual life prediction is completed; the reconstruction system comprises a multi-modal data acquisition device, a data processing integration module and a signal line. According to the electromagnetic-acoustic-thermal multi-field coupling inversion model provided by the invention, data modeling is carried out after multi-modal data is processed through a multi-field coupling algorithm, so that full-dimension and high-precision three-dimensional reconstruction of internal defects of the hydrogen storage bottle in a full-use scene is realized, a micro-seepage part can be accurately positioned when leakage occurs in the high-pressure hydrogen storage bottle, and the reliability of the high-pressure hydrogen storage bottle is improved. The detection precision and the response speed are effectively improved, and early warning is carried out in order to prolong the service life of the hydrogen storage bottle and improve the safety.

Description

Technical Field

[0001] The present invention relates to the field of hydrogen fuel cell technology, and in particular to a method and system for three-dimensional reconstruction of internal defects of a multi-field coupled hydrogen storage bottle. Background Art

[0002] Hydrogen energy is a clean, green, low-carbon, and environmentally friendly renewable secondary energy source. In recent years, hydrogen energy has been widely used in automotive power systems. High-purity hydrogen generates electricity and heat energy through electrocatalysis within the fuel cell. The electricity can directly power the drive motor to drive the vehicle smoothly, while the heat energy is transported to the passenger compartment through the thermal management system, providing a warm and comfortable driving environment for the occupants. Due to the physical properties of hydrogen's low density (only 0.0899g / L under standard conditions), hydrogen storage requires a large amount of storage space and has high material costs. To meet the energy needs of automotive fuel cells, hydrogen must be stored at high pressure in high-pressure carbon fiber hydrogen cylinders (for example, 35MPa or 70MPa). During use, the hydrogen supply is controlled by controlling the on / off value of the cylinder valve and pressure reducing valve.

[0003] However, for special working conditions (for example, when the vehicle encounters extreme conditions such as vibration, impact and collision), the existing monitoring and detection solutions are: internal detection methods rely on sensors (such as pressure sensors and hydrogen quality detectors, etc.), and their monitoring and detection accuracy is difficult to capture tiny leaks, and thus cannot meet the safety threshold requirements; while external detection methods (such as acoustic wave detection devices and stress-strain sensors) are subject to the anisotropic properties of carbon fiber materials, and the detection depth of the gas cylinder protective layer is limited. It is difficult to penetrate the multi-layer structure and there is a risk of missing deep defects; in addition, the existing detection methods can only provide two-dimensional projection or cross-sectional data, which cannot accurately find leaks and failure paths. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a three-dimensional reconstruction method and system for internal defects of a multi-field coupled hydrogen storage bottle.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for three-dimensional reconstruction of internal defects of a hydrogen storage bottle using multi-field coupling, comprising:

[0006] S100, deploying a multimodal data acquisition and processing architecture system to collect and process multi-source data;

[0007] S200: Construct a multi-field coupling model, solve the multi-field coupling equations simultaneously, and perform hybrid strategy direction-by-direction order reduction.

[0008] S300, extracting and transforming defect geometric feature data, outputting defect three-dimensional parametric data, and reversely evolving defect parameters;

[0009] S400 generates a three-dimensional visualization model, performs mechanical-diffusion coupling analysis, and completes the remaining life prediction.

[0010] In some embodiments, the S100 includes:

[0011] S110, corresponding to the deployment of electromagnetic-acoustic impedance-thermal conduction multi-modal data acquisition and processing architecture system;

[0012] S120 , synchronously triggering all acquisition devices to collect and receive multimodal monitoring data, and performing spatiotemporal coordinate calibration and multi-source data alignment preprocessing.

[0013] In some embodiments, the S200 includes:

[0014] S210, constructing a spatially discrete multi-field coupling model and simultaneously solving electromagnetic-acoustic impedance-heat conduction coupling equations;

[0015] S220, decompose the three-dimensional spatial coupling equations and perform time advancement alternately in one dimension;

[0016] S230, constructing the optimal basis function based on the proper orthogonal decomposition, and solving the coupled equations by reducing the order.

[0017] In some embodiments, the S300 includes:

[0018] S310, a deep priori network is designed to extract defect geometric features, input multimodal data tensors, and output a three-dimensional parametric description of the defect;

[0019] S320, based on the particle swarm optimization algorithm, introduces the quantum tunneling mechanism and data-driven reverse evolution of defect parameters.

[0020] In some embodiments, the S400 includes:

[0021] S410, constructing a defect voxel model and generating a three-dimensional visual mesh model;

[0022] S420, establish a hydrogen-induced damage evolution model, introduce stress intensity factor and hydrogen diffusion flux calculation to perform mechanical-diffusion coupling analysis;

[0023] S430, predict the remaining life of the hydrogen storage bottle and output the remaining life result.

[0024] In a second aspect, the present invention further provides a multi-field coupled hydrogen storage bottle internal defect three-dimensional reconstruction system for executing the multi-field coupled hydrogen storage bottle internal defect three-dimensional reconstruction method as described in the first aspect, the reconstruction system comprising:

[0025] A multimodal data acquisition device, a data processing integrated module and a controller, wherein the multimodal data acquisition device is electrically connected to the data processing integrated module and the controller;

[0026] The controller is used to control the multimodal data acquisition device to collect and send multimodal monitoring data of the hydrogen storage bottle in real time;

[0027] The data processing integration module is used to receive multimodal monitoring data, perform time-space coordinate calibration and multi-source data alignment preprocessing, and complete three-dimensional monitoring of internal defects of hydrogen storage bottles.

[0028] In some embodiments, the multimodal data acquisition device includes a pulsed eddy current probe, an air-coupled ultrasonic transducer and a phase-locked infrared thermal imager, and the pulsed eddy current probe, air-coupled ultrasonic transducer and phase-locked infrared thermal imager are electrically connected to the data processing integrated module through a signal line.

[0029] In some embodiments, the pulsed eddy current probe is arranged on the outside of the hydrogen storage bottle, and the air-coupled ultrasonic transducer and the phase-locked infrared thermal imager are arranged on the outside of the pulsed eddy current probe.

[0030] In some embodiments, at least 6 eddy current array probes are arranged at equal angles, at least 3 groups of air-coupled ultrasonic transducers and phase-locked infrared thermal imagers are arranged at equal intervals in the axial direction of the hydrogen storage bottle, and 2-4 air-coupled ultrasonic transducers and phase-locked infrared thermal imagers are arranged at equal angles. The axial spacing between each group of air-coupled ultrasonic transducers and phase-locked infrared thermal imagers is 250-350 mm, and the axial spacing is 1-2 times the diameter of the hydrogen storage bottle.

[0031] In some embodiments, the data processing integration module includes a parameter setting subunit, a data receiving subunit, a data storage subunit and a data analysis subunit, the parameter setting subunit is used to set calculation parameters, the data receiving subunit is used to receive multimodal monitoring data, the data analysis subunit is used to analyze and process multimodal monitoring data, and the data storage subunit stores multimodal monitoring data.

[0032] The present invention has the following beneficial effects:

[0033] 1. This invention proposes an electromagnetic-acoustic-thermal multi-field coupled inversion model. Through a multi-field coupling algorithm, it processes multimodal data and then performs data modeling. This model realizes full-dimensional, high-precision three-dimensional reconstruction of internal defects of hydrogen storage bottles in all usage scenarios. It can accurately locate the micro-leakage site when a high-pressure hydrogen storage bottle leaks, effectively improving detection accuracy and response speed. It can also provide real-time online modeling to determine the health status of the fiber layer of the hydrogen bottle. Through an intelligent algorithm, it calculates health prediction data, providing early warning to extend the service life of the hydrogen storage bottle and improve safety.

[0034] 2. After deploying the multimodal data acquisition and processing architecture system, the present invention uses a multimodal data acquisition device composed of a pulsed eddy current probe, an air-coupled ultrasonic transducer, and a phase-locked infrared thermal imager to send the detected multimodal monitoring data to the data processing integration module for acquisition and preprocessing. This can quantify the three-dimensional morphology of defects, accurately locate leaks and failure paths, and then predict the remaining life of hydrogen storage bottles, thereby enabling all-round monitoring and protection of carbon fiber high-pressure hydrogen storage bottles. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The process of the three-dimensional reconstruction method of internal defects of the multi-field coupled hydrogen storage bottle proposed by the present invention Figure 1 ;

[0036] Figure 2 The process of the three-dimensional reconstruction method of internal defects of the multi-field coupled hydrogen storage bottle proposed by the present invention Figure 2 ;

[0037] Figure 3 The process of the three-dimensional reconstruction method of internal defects of the multi-field coupled hydrogen storage bottle proposed by the present invention Figure 3 ;

[0038] Figure 4 The process of the three-dimensional reconstruction method of internal defects of the multi-field coupled hydrogen storage bottle proposed by the present invention Figure 4 ;

[0039] Figure 5 The process of the three-dimensional reconstruction method of internal defects of the multi-field coupled hydrogen storage bottle proposed by the present invention Figure 5 ;

[0040] Figure 6 This is a schematic diagram of the multi-field coupled three-dimensional reconstruction system for internal defects of hydrogen storage bottles proposed by the present invention;

[0041] Figure 7 This is the main view of the multi-field coupling hydrogen storage bottle internal defect 3D reconstruction system proposed by the present invention;

[0042] Figure 8 This is a side view of the multi-field coupling hydrogen storage bottle internal defect three-dimensional reconstruction system proposed by the present invention.

[0043] Legend:

[0044] 1. Multimodal data acquisition equipment; 11. Pulsed eddy current probe; 12. Air-coupled ultrasonic transducer; 13. Phase-locked infrared thermal imager; 2. Data processing integration module; 21. Parameter setting subunit; 22. Data receiving subunit; 23. Data storage subunit; 24. Data analysis subunit; 3. Controller; 4. Hydrogen storage bottle. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] The embodiment of the present application provides a method and system for three-dimensional reconstruction of internal defects of hydrogen storage bottles with multi-field coupling, which solves the problem that the monitoring detection accuracy of internal detection means in the prior art is difficult to capture tiny leaks, while the external detection means have the risk of missing deep defects and can only provide two-dimensional projection or cross-sectional data, and cannot accurately find leak points and failure paths. However, the present application uses a multi-field coupling algorithm to process multimodal data and then perform data modeling, achieving full-dimensional, high-precision three-dimensional reconstruction of internal defects of hydrogen storage bottles in all usage scenarios. It can accurately locate the micro-seepage site when a high-pressure hydrogen storage bottle leaks, effectively improving the detection accuracy and response speed, and providing early warning to extend the service life of the hydrogen storage bottle and improve safety.

[0047] Please refer to the following examples for details:

[0048] Reference Figure 1-Figure 5 The present invention provides an embodiment of a method for three-dimensional reconstruction of internal defects of a multi-field coupled hydrogen storage bottle, the specific steps of which include:

[0049] S100, deploying a multimodal data acquisition and processing architecture system to collect and process multi-source data;

[0050] S200: Construct a multi-field coupling model, solve the multi-field coupling equations simultaneously, and perform hybrid strategy direction-by-direction order reduction.

[0051] S300, extracting and transforming defect geometric feature data, outputting defect three-dimensional parametric data, and reversely evolving defect parameters;

[0052] S400 generates a three-dimensional visualization model, performs mechanical-diffusion coupling analysis, and completes the remaining life prediction.

[0053] Please continue reading Figure 2 In this embodiment, S100 includes:

[0054] S110, corresponding to the deployment of electromagnetic-acoustic impedance-thermal conduction multi-modal data acquisition and processing architecture system;

[0055] S120 , synchronously triggering all acquisition devices to collect and receive multimodal monitoring data, and performing spatiotemporal coordinate calibration and multi-source data alignment preprocessing.

[0056] It should be explained in detail that the electromagnetic-acoustic impedance-heat conduction multimodal data acquisition and processing architecture system consists of a multimodal data acquisition device 1, a data processing integration module 2 and a controller 3. Data interaction and command transmission are achieved between the modules through a high-speed communication bus. The controller sends a synchronous trigger signal to the data acquisition device and then performs detection synchronously, which can eliminate the time error caused by the device response delay and ensure the consistency of the multimodal data in the time dimension. After receiving the multimodal data, the data processing integration module performs data preprocessing (i.e., time synchronization and spatial coordinate calibration to achieve spatiotemporal alignment of multi-source data).

[0057] It can be understood that by using a multi-field coupling algorithm and then processing multimodal data for data modeling, a full-dimensional, high-precision three-dimensional reconstruction of the internal defects of the hydrogen storage bottle 4 in all usage scenarios is achieved. When a leak occurs in the high-pressure hydrogen storage bottle 4, the micro-seepage site can be accurately located, effectively improving the detection accuracy and response speed; and real-time online modeling can be used to give the health status of the fiber layer of the hydrogen cylinder. Through an intelligent algorithm, the health prediction data is calculated to provide early warning for extending the service life of the hydrogen storage bottle 4 and improving safety.

[0058] Please continue reading Figure 3 In this embodiment, S200 includes:

[0059] S210, constructing a spatially discrete multi-field coupling model and simultaneously solving electromagnetic-acoustic impedance-heat conduction coupling equations;

[0060] S220, decompose the three-dimensional spatial coupling equations and perform time advancement alternately in one dimension;

[0061] S230, constructing the optimal basis function based on the proper orthogonal decomposition, and solving the coupled equations by reducing the order.

[0062] It should be explained in detail that when constructing the electromagnetic-acoustic impedance-heat conduction coupling equation, the numerical solution adopts a hybrid solution:

[0063]

[0064] Specifically:

[0065] (1) is the electromagnetic field equation:

[0066] Where E is the electric field intensity (V / m), which describes the distribution of the electromagnetic field in the medium; μ is the magnetic permeability tensor (H / m), which characterizes the response characteristics of the material to the magnetic field. For carbon fiber composite materials, anisotropy needs to be considered (typical values ​​include: μ xx =1.2μ0,μ yy =μ zz =1.05μ0,μ0=4π×10 -7H / m); σ is the conductivity tensor (S / m), and the conductivity of the carbon fiber direction in the composite material can reach 10 4 S / m, resin matrix is ​​about 10 -12 °S / m; Je is the external excitation current density (A / m 2 ), generated by the eddy current probe (typical value is 5×10 6 A / m 2 , 10MHz excitation).

[0067] (2) is the acoustic field equation:

[0068] Where u is the displacement vector (m), which describes the vibration displacement of the particles inside the material; ρ is the material density (kg / m 3 ), the density of T700S carbon fiber / epoxy resin composite material is 1.6×10 3 kg / m 3 ; S is the stress tensor (Pa), through the constitutive relationship S=C ijkl∈k lCalculation; C ijkl is the fourth-order stiffness tensor (Pa). For orthotropic materials, the independent components are reduced to 9, such as C 1111 =145GPa (fiber direction), C 1212 =4.5GPa (shear modulus);∈ kl is the strain tensor, which is given by the displacement gradient ∈ kl =21(uk, l+ul, k) calculation; α is the thermal expansion coefficient tensor (1 / K), the carbon fiber longitudinal α 11 =-0.7×10 -6 K -1 , lateral α 22 =28×10 -6 K -1 .

[0069] (3) is the thermal field equation

[0070] T is the temperature field (K), which reflects the internal heat distribution of the material; k is the thermal conductivity tensor (W / (m·K)), the longitudinal k of the carbon fiber 11 =150W / (m·K), lateral k 22 =5W / (m·K);c p is the specific heat capacity (J / (kg·K)), and the epoxy resin is about 1.2×10 3 J / (kg·K);σ|E| 2 is the Joule heat source term (W / m 3 ), which is generated by the energy dissipation of the electromagnetic field; β is the thermoelastic coupling coefficient (Pa / K), which is 1.5×10 6 Pa / K (calibrated by DSC experiment); T0 is the reference temperature (K), usually the ambient temperature of 293K.

[0071] Furthermore, when constructing a spatially discrete multi-field coupling model:

[0072] (1) The electromagnetic field uses an unstructured tetrahedral grid (minimum size 0.1 mm);

[0073] It should be noted that the distribution of electromagnetic fields in a medium is significantly affected by the geometric structure and material properties, especially at interfaces (such as the boundary between conductors and insulators), where sudden changes in field strength will occur. Unstructured tetrahedral meshes can accurately fit complex geometric boundaries through irregular but flexible unit shapes (tetrahedrons), avoiding the step-like approximation errors produced by structured meshes (such as hexahedral meshes) when dealing with sharp edges or curved surfaces. The minimum size of 0.1 mm is set to meet the analytical requirements of high-frequency electromagnetic field effects (such as the skin effect) - high-frequency currents are concentrated on the surface of the conductor, and a sufficiently small mesh is required to capture the drastic changes in current density.

[0074] It can be understood that the unstructured tetrahedral mesh is suitable for hydrogen storage bottles 4 containing cracks, holes or complex surfaces; it ensures that a sufficient number of mesh layers are formed on the conductor surface to accurately calculate the eddy current loss and Joule heat distribution; the isotropic characteristics of the unstructured mesh avoid numerical oscillations caused by the mismatch between the mesh direction and the field direction.

[0075] (2) The acoustic field is encrypted using anisotropic meshes (3 times denser along the fiber direction);

[0076] It should be explained in detail that the propagation of sound waves in fiber-reinforced composites shows obvious anisotropy - the sound speed along the fiber direction is usually 2-3 times higher than that in the perpendicular direction, and the attenuation is lower; this characteristic allows the sound wave energy to propagate farther and more concentratedly in the fiber direction; traditional isotropic grids (such as uniform tetrahedral grids) cannot effectively capture this directional difference, which may cause the propagation of sound waves in the fiber direction to be excessively attenuated or scattered; by increasing the grid size by 3 times along the fiber direction, the ratio of the unit size of the grid in each direction to the wavelength of the sound wave can be kept consistent (usually the grid size is required to be ≤ wavelength / 10), thereby ensuring the stability of the numerical solution of the wave equation.

[0077] It is understandable that anisotropic mesh encryption can optimize the discrete accuracy of the sound wave propagation path and reduce dispersion errors; form a denser mesh at the fiber-matrix interface to accurately capture the refraction and reflection behavior of sound waves; and encrypt only in directions where high precision is required to avoid the waste of computational resources caused by global mesh refinement.

[0078] (3) Establishing an adaptive octree grid for thermal field:

[0079] It should be explained in detail that during heat conduction, temperature gradients are usually concentrated near the heat source or at the material interface, while the temperature changes in areas far away from the heat source are relatively gentle; the octree grid adopts a recursive segmentation strategy: starting from a cube covering the entire computational domain, the area with drastic temperature changes (such as the area where the temperature gradient exceeds the threshold) is divided into 8 sub-cubes, and this process is repeated until the preset minimum grid size is reached; while a coarse grid is maintained in areas where the temperature changes slowly; this dynamic adaptive strategy forms a grid distribution that is "dense in the heat source area and sparse in the far field area."

[0080] It can be understood that the adaptive octree grid automatically generates a high-density grid around the hot spot to ensure the convergence of the numerical solution of the heat conduction equation; compared with the uniform grid, it can reduce more than 90% of unnecessary grid cells, greatly reducing the amount of calculation; and it can handle both millimeter-level devices and micron-level local hotspots at the same time.

[0081] Furthermore, decomposing the three-dimensional space coupling equations, when time advancement is performed alternately in one-dimensional direction:

[0082] (1) Using the Alternating Direction Implicit (ADI) algorithm for time advancement:

[0083] It should be explained in detail that the ADI algorithm is an efficient time-marching method for multidimensional partial differential equations (such as the heat conduction equation and the wave equation). The equation contains unknown quantities at future times and needs to be solved through matrix inversion. Compared with the explicit format, its time step is not restricted by conditions, avoiding the numerical divergence caused by the excessive step size in the explicit format.

[0084] It is understandable that in the coupling of multiple physical fields, the time characteristic scales of different fields vary significantly (such as nanosecond changes in electromagnetic fields and millisecond changes in thermal fields). Time advancement through the ADI algorithm allows the thermal field to advance in millisecond steps without having to follow the nanosecond steps of the electromagnetic field, which improves the computational efficiency by more than several orders of magnitude. The splitting direction can be adjusted for different physical fields (such as the electromagnetic field is split according to the x / y / z axes, and the thermal field is split according to the material interface direction) to adapt to the dominant transmission direction of each field. When the Joule heat of the electromagnetic field is transferred to the thermal field and the temperature of the thermal field affects the propagation of the sound field, the implicit characteristics of the ADI algorithm can avoid the accumulation of coupling errors.

[0085] (2) Construct a reduced-order basis based on Proper Orthogonal Decomposition (POD):

[0086]

[0087] Where ui is the original high-dimensional solution vector (snapshot data), which is the solution vector collected at different times / parameters by the full-order model. The dimension is the same as the degree of freedom of the physical field (such as 1×10 6 ), with typical values ​​being the electromagnetic field solution vector and acoustic field displacement field data; φk is the POD modal basis function, which characterizes the dominant spatial mode of the system's dynamic behavior. It is an orthogonal unit vector (∥φk∥=1), with typical values ​​being the first 20 modal basis functions; αk is the modal coefficient, which describes the weight of each mode in the solution. It is a time-varying coefficient (αk(t)∈R), with typical values ​​being obtained through projection calculation; r is the number of retained modes, which determines the accuracy and efficiency of the reduced-order model. It is selected based on the energy proportion criterion (usually 95% of the energy is retained), with a typical value of r=20 (for N=200 snapshots); N is the number of snapshots, which is the amount of training data used to construct the POD basis. It must satisfy N≥3r to ensure the completeness of the basis, with a typical value of N=200 (time step sampling).

[0088] Please continue reading Figure 4 In this embodiment, S300 includes:

[0089] S310, a deep priori network is designed to extract defect geometric features, input multimodal data tensors, and output a three-dimensional parametric description of the defect;

[0090] S320, based on the particle swarm optimization algorithm, introduces the quantum tunneling mechanism and data-driven reverse evolution of defect parameters.

[0091] It should be explained in detail that when designing a deep prior network (DPN) to extract defect geometric features:

[0092] (1) Taking a multimodal data tensor as input (channel dimensions correspond to different physical quantities), the three-dimensional parameterized description of the defect (position, size, orientation probability distribution) is output:

[0093] The Deep Prior Network (DPN) is a model that integrates physical prior knowledge and deep learning architecture, and is specifically designed for defect detection and characterization problems: the multimodal tensor input is to splice the data of different physical quantities such as electromagnetic fields, acoustic fields, and thermal fields according to the channel dimension to form a multidimensional tensor (such as [Batch, Height, Width, Channels]), while capturing the response characteristics of defects in different physical fields; and the three-dimensional parameterized output is to directly output the probability distribution of the defect geometric parameters (such as Gaussian distribution of position and lognormal distribution of size), rather than deterministic numerical values, to quantify the uncertainty in the detection process.

[0094] It is understandable that electromagnetic fields are sensitive to surface defects in conductive materials, acoustic fields are more effective for internal cracks, and thermal fields are good at detecting areas with changing thermal resistance. DPN achieves complementary advantages by automatically learning the weights of each mode.

[0095] (2) Applying the improved particle swarm optimization algorithm (introducing the quantum tunneling mechanism) to solve the inverse problem:

[0096] Quantum tunneling mechanism:

[0097] When a particle falls into a local optimum, tunneling occurs with probability p = exp(-Δf / T).

[0098] Where: Δf is the change in fitness, and T is the temperature parameter (decays with the number of iterations).

[0099] Parameter update rules:

[0100]

[0101] in:

[0102] p best is the historical optimal position of the particle, evaluated by the fitness function F; β is the shrinkage-expansion coefficient, which has an initial value of 1.0 and decreases linearly to 0.3 to balance global search and local convergence; is the position of the i-th particle at the t-th iteration (a multidimensional vector representing defect parameters such as position, size, etc.); m best is the average value of the group’s historical optimal position, u is a uniformly distributed random number, u~U(0,1).

[0103] The fitness function is:

[0104] F=‖M sim (x)-M meas ‖2+λTV(x)

[0105] Where Msim is the simulated data; Mmeas is the measured data; λ is the regularization coefficient (λ = 0.1), determined by the L-curve method; TV(x) is the total variation regularization term, TV(x) = ∑|x i +1-x i |, suppress the noise in the inversion results.

[0106] It is understandable that particles in traditional particle swarm optimization (PSO) are constrained by "speed" and are prone to oscillation near the local optimum; while the quantum tunneling mechanism allows particles to probabilistically penetrate the "potential barrier" and jump directly to a new area for search. Through the probabilistic distribution characteristics of the quantum state, the particle swarm can simultaneously explore multiple potential solution areas and adapt to the multi-solution characteristics of the inverse problem. Compared with traditional global optimization algorithms (such as genetic algorithms), directional tunneling reduces invalid searches and can reduce the number of iterations by more than half in high-dimensional inverse problems (such as three-dimensional defect inversion).

[0107] Please continue reading Figure 5 In this embodiment, S400 includes:

[0108] S410, constructing a defect voxel model and generating a three-dimensional visual mesh model;

[0109] S420, establish a hydrogen-induced damage evolution model, introduce stress intensity factor and hydrogen diffusion flux calculation to perform mechanical-diffusion coupling analysis;

[0110] S430, predict the remaining life of the hydrogen storage bottle 4 and output the remaining life result.

[0111] It should be explained in detail that when a defect voxel model is used to generate a 3D mesh, and then a mechanical-diffusion coupling analysis is performed through stress intensity factor and hydrogen diffusion flux calculation to achieve life prediction:

[0112] (1) Defect voxel modeling, using the marching cube algorithm to generate a visual 3D mesh:

[0113] Scalar field construction:

[0114] The defect probability distribution P(x, y, z) obtained by inversion is a continuous field function and needs to be discretized into a three-dimensional grid. The octree structure recursively divides the space: starting from the root node (covering the entire computational domain), the space is divided into 8 sub-cubes; if the probability changes drastically within the sub-cube (such as the variance exceeds the threshold), it is further subdivided until the preset minimum resolution (such as 0.1mm) is reached; finally, each leaf node stores a single probability value, forming a three-dimensional grid with adaptive resolution, which can adaptively refine the defect boundary and accurately capture features such as crack tips and hole edges; subsequent isosurface extraction only needs to process the leaf nodes, reducing the amount of calculation and representing both macrostructures and micro defects at the same time.

[0115] Isosurface extraction:

[0116] Set the probability threshold P th=0.7, extract all point sets that satisfy P(x, y, z) = 0.7 to form the three-dimensional surface of the defect; use the cube algorithm (MarchingCubes, MC) to traverse the cube mesh composed of octet tree leaf nodes, and judge the relationship between the vertex probability value and the threshold voxel by voxel. Each cube has 256 vertex state combinations (each vertex has two states: "above the threshold" or "below the threshold"). The facet connection method is determined by looking up the table, and the surface normal vector is calculated based on the probability field gradient for subsequent lighting rendering and mechanical analysis. The defect size (such as crack length and hole volume) is calculated by the facet area / volume. The generated triangular mesh can be directly imported into mechanical analysis software.

[0117] Topology Optimization:

[0118] Apply smoothing and edge collapse algorithms (Laplacian) to optimize and reduce the number of mesh faces, accelerating subsequent analyses (such as fluid / heat conduction simulations).

[0119] (2) Mechanical-diffusion coupling analysis and establishment of hydrogen-induced damage evolution model:

[0120] Stress intensity factor calculation:

[0121]

[0122] Where Y is the shape factor, which is obtained by the J-integral method;

[0123] Calculate the hydrogen diffusion flux and solve Fick's second law:

[0124]

[0125] Among them, V H is the partial molar volume of hydrogen, C is the hydrogen concentration field (mol / m 3 ), the initial value is determined by gas phase permeation experiment; D is the diffusion coefficient (m 2 / s), carbon fiber longitudinal D 11 =1.5×10 -12 °m 2 / s, horizontal D 22 =3.0×10 -14 °m 2 / s;V H is the partial molar volume of hydrogen (m 3 / mol), take 2.0×10 -6 m 3 / mol; R is the gas constant 8.314 J / (mol·K); T is the absolute temperature (K); σ h is the hydrostatic stress (Pa),

[0126] (3) Remaining life prediction, improved Paris formula:

[0127]

[0128] Adaptive step-size Runge-Kutta method is used to solve differential equations;

[0129] Wherein, β is the hydrogen concentration factor, β = 0.15 (calibrated by molecular dynamics simulation); da / dN is the crack growth rate (m / cycle), which is fitted by experimental data; ΔK is the stress intensity factor range (MPa / m), Y is the geometric correction factor (dimensionless), which is 1.12 for surface cracks and 0.728 for internal cracks; Δσ is the stress amplitude (MPa), obtained by finite element analysis; a is the crack half-length (m);

[0130] C and m are material constants. For T700S carbon fiber / epoxy resin, C = 1.2 × 10-11, m = 3.2 (calibrated by fatigue test); β is the hydrogen enhancement factor 1 / (mol / m 3 )), β = 0.15, determined by molecular dynamics simulation; C H is the hydrogen concentration (mol / m 3 ), which is obtained by solving the diffusion equation.

[0131] Take the detection of 0.8mm×3.2mm interface debonding defect as an example:

[0132] The multi-field coupling model calculates the local stress concentration factor SCF = 2.3, corresponding to Δσ = 85 MPa; the hydrogen diffusion equation is solved to obtain the hydrogen concentration CH at the defect front = 12.5 mol / m 3 ;

[0133] The crack growth rate dN / da predicted by the improved Paris formula is calculated as follows:

[0134] dN / da=1.2×10 -11 ×(1.12×85π×0.004)×3.2×e0.15×12.5=4.7×10 -8 m / cycle;

[0135] When the critical defect size ac = 5.0 mm, the remaining life N is calculated as follows:

[0136] N = ∫ 0.85.04.7 × 10 -8 da≈2000 cycles.

[0137] Reference Figure 6-Figure 8The present invention further provides an embodiment of a multi-field coupling hydrogen storage bottle internal defect 3D reconstruction system, which is used to execute the multi-field coupling hydrogen storage bottle internal defect 3D reconstruction method described in the above embodiment. The reconstruction system includes:

[0138] Multimodal data acquisition equipment 1, data processing integrated module 2 and controller 3, the controller 3 is used to control the multimodal data acquisition equipment 1 to collect and send multimodal monitoring data of the hydrogen storage bottle 4 in real time; the data processing integrated module 2 is used to receive multimodal monitoring data, perform time and space coordinate calibration and multi-source data alignment preprocessing, and complete three-dimensional monitoring of internal defects of the hydrogen storage bottle 4; and the multimodal data acquisition equipment 1 is electrically connected to the data processing integrated module 2 and the controller 3.

[0139] Among them, the multimodal data acquisition equipment 1 includes a pulsed eddy current probe 11, an air-coupled ultrasonic transducer 12 and a phase-locked infrared thermal imager 13. The pulsed eddy current probe 11, the air-coupled ultrasonic transducer 12 and the phase-locked infrared thermal imager 13 are electrically connected to the data processing integrated module 2 through signal lines, and are respectively used to capture the electromagnetic response, sound wave propagation characteristics and thermal radiation characteristics of the object under test.

[0140] It can be understood that after deploying the multimodal data acquisition and processing architecture system, the multimodal data acquisition device 1 composed of a pulsed eddy current probe 11, an air-coupled ultrasonic transducer 12 and a phase-locked infrared thermal imager 13 sends the detected multimodal monitoring data to the data processing integration module 2 for acquisition and preprocessing, which can quantify the three-dimensional morphology of defects, accurately find leakage points and failure paths, and then predict the remaining life of the hydrogen storage bottle 4, and can carry out all-round monitoring and protection of the carbon fiber high-pressure hydrogen storage bottle 4.

[0141] It should be explained in detail that the pulsed eddy current probe 11 is arranged in a ring around the outside of the hydrogen storage bottle 4, and the air-coupled ultrasonic transducer 12 and the phase-locked infrared thermal imager 13 are arranged in a ring around the outside of the pulsed eddy current probe 11. Each acquisition device is arranged independently and does not affect each other. Specifically, there are at least 6 eddy current array probes arranged at equal angles, at least 3 groups of air-coupled ultrasonic transducers 12 and phase-locked infrared thermal imagers 13 are arranged in an axially equal spacing around the hydrogen storage bottle 4, and 2-4 air-coupled ultrasonic transducers 12 and phase-locked infrared thermal imagers 13 are arranged in an axially equal spacing (the ring angle is 60-180°, and each group is axially distributed according to the characteristics of the equipment). The axial spacing between each group of air-coupled ultrasonic transducers 12 and phase-locked infrared thermal imagers 13 is 250-350mm, and the axial spacing is 1-2 times the diameter of the hydrogen storage bottle 4.

[0142] For example, the operating frequency of the pulsed eddy current probe 11 can be flexibly adjusted within the range of 0.1-10 MHz: the low frequency band (such as 0.1-1 MHz) can produce a stronger skin effect, increase the eddy current penetration depth, and is suitable for detecting defects at deeper locations; the high frequency band (such as 1-10 MHz) has a higher spatial resolution and can accurately capture tiny surface cracks and shallow surface defects.

[0143] The air-coupled ultrasonic transducer 12 has a center frequency of 5 MHz, achieving a good balance between penetration and resolution when propagating through air. To overcome the interference of anisotropic materials such as carbon fiber on ultrasonic propagation, the transducer matrix transmits ultrasonic waves at an angle of incidence of 45° to the fiber direction. This special angle of incidence effectively suppresses problems such as sound wave propagation direction deviation and velocity differences caused by material anisotropy, allowing ultrasonic waves to propagate more evenly within the material, thereby more accurately capturing interlayer reflection signals.

[0144] The high-speed lock-in infrared thermal imager 13 has a frame rate of 1kHz and combines laser thermal excitation technology to emit a periodically modulated laser beam to the surface of the object being measured. This causes the surface to absorb the laser energy, generating a periodic temperature rise, forming a thermal wave signal that propagates in the depth direction. During the propagation of the thermal wave, if it encounters defects within the material (such as voids or debonded areas), heat will accumulate due to differences in thermal resistance, resulting in temperature anomalies on the object's surface.

[0145] Furthermore, the data processing integration module 2 includes a parameter setting subunit 21, a data receiving subunit 22, a data storage subunit 23, and a data analysis subunit 24. The parameter setting subunit 21 is used to set calculation parameters and supports user customization of multiple key parameters, including data acquisition frequency, sampling duration, algorithm threshold, etc. The data receiving subunit 22 is used to receive multimodal monitoring data and can simultaneously access multimodal data collected by various devices such as the annular array pulsed eddy current probe 11, the air-coupled ultrasonic transducer 12 matrix, and the high-speed phase-locked infrared thermal imager 13. The data analysis subunit 24 is used to analyze and process multimodal monitoring data, including time synchronization, spatial coordinate calibration, noise filtering, etc., to eliminate temporal and spatial differences and interference signals between multi-source data. The data storage subunit 23 stores multimodal monitoring data.

[0146] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A three-dimensional reconstruction method for internal defects of a multi-field coupled hydrogen storage bottle, characterized in that: include: S100, deploying a multimodal data acquisition and processing architecture system to collect and process multi-source data; S200: Construct a multi-field coupling model, solve the multi-field coupling equations simultaneously, and perform hybrid strategy direction-by-direction order reduction. S300, extracting and transforming defect geometric feature data, outputting defect three-dimensional parametric data, and reversely evolving defect parameters; S400 generates a three-dimensional visualization model, performs mechanical-diffusion coupling analysis, and completes the remaining life prediction.

2. The method for three-dimensional reconstruction of internal defects of a multi-field coupled hydrogen storage bottle according to claim 1 is characterized in that: The S100 includes: S110, corresponding to the deployment of electromagnetic-acoustic impedance-thermal conduction multi-modal data acquisition and processing architecture system; S120 , synchronously triggering all acquisition devices to collect and receive multimodal monitoring data, and performing spatiotemporal coordinate calibration and multi-source data alignment preprocessing.

3. The method for three-dimensional reconstruction of internal defects of a multi-field coupled hydrogen storage bottle according to claim 1 is characterized in that: The S200 includes: S210, constructing a spatially discrete multi-field coupling model and simultaneously solving electromagnetic-acoustic impedance-heat conduction coupling equations; S220, decompose the three-dimensional spatial coupling equations and perform time advancement alternately in one dimension; S230, constructing the optimal basis function based on the proper orthogonal decomposition, and solving the coupled equations by reducing the order.

4. The method for three-dimensional reconstruction of internal defects of a multi-field coupled hydrogen storage bottle according to claim 1 is characterized in that: The S300 includes: S310, a deep priori network is designed to extract defect geometric features, input multimodal data tensors, and output a three-dimensional parametric description of the defect; S320, based on the particle swarm optimization algorithm, introduces the quantum tunneling mechanism and data-driven reverse evolution of defect parameters.

5. The method for three-dimensional reconstruction of internal defects of a multi-field coupled hydrogen storage bottle according to claim 1 is characterized in that: The S400 includes: S410, constructing a defect voxel model and generating a three-dimensional visual mesh model; S420, establish a hydrogen-induced damage evolution model, introduce stress intensity factor and hydrogen diffusion flux calculation to perform mechanical-diffusion coupling analysis; S430, predict the remaining life of the hydrogen storage bottle and output the remaining life result.

6. A multi-field coupled three-dimensional reconstruction system for internal defects of hydrogen storage bottles, characterized by: The reconstruction system is used to run the multi-field coupling hydrogen storage bottle internal defect three-dimensional reconstruction method according to any one of claims 1 to 5, and the reconstruction system includes: A multimodal data acquisition device, a data processing integrated module and a controller, wherein the multimodal data acquisition device is electrically connected to the data processing integrated module and the controller; The controller is used to control the multimodal data acquisition device to collect and send multimodal monitoring data of the hydrogen storage bottle in real time; The data processing integration module is used to receive multimodal monitoring data, perform time-space coordinate calibration and multi-source data alignment preprocessing, and complete three-dimensional monitoring of internal defects of hydrogen storage bottles.

7. The multi-field coupled hydrogen storage bottle internal defect 3D reconstruction system according to claim 6 is characterized in that: The multimodal data acquisition device includes a pulsed eddy current probe, an air-coupled ultrasonic transducer and a phase-locked infrared thermal imager, and the pulsed eddy current probe, the air-coupled ultrasonic transducer and the phase-locked infrared thermal imager are electrically connected to the data processing integrated module through a signal line.

8. The multi-field coupled hydrogen storage bottle internal defect 3D reconstruction system according to claim 7 is characterized in that: The pulsed eddy current probe ring is arranged on the outside of the hydrogen storage bottle, and the air-coupled ultrasonic transducer and the phase-locked infrared thermal imager ring are arranged on the outside of the pulsed eddy current probe.

9. The multi-field coupled hydrogen storage bottle internal defect 3D reconstruction system according to claim 7, characterized in that: There are at least 6 eddy current array probes arranged at equal angles, at least 3 groups of air-coupled ultrasonic transducers and phase-locked infrared thermal imagers are arranged at equal intervals in the axial direction of the hydrogen storage bottle, and 2-4 air-coupled ultrasonic transducers and phase-locked infrared thermal imagers are arranged at equal angles. The axial spacing between each group of air-coupled ultrasonic transducers and phase-locked infrared thermal imagers is 250-350 mm, and the axial spacing is 1-2 times the diameter of the hydrogen storage bottle.

10. The multi-field coupled hydrogen storage bottle internal defect 3D reconstruction system according to claim 6, characterized in that: The data processing integrated module includes a parameter setting subunit, a data receiving subunit, a data storage subunit and a data analysis subunit. The parameter setting subunit is used to set calculation parameters, the data receiving subunit is used to receive multimodal monitoring data, the data analysis subunit is used to analyze and process multimodal monitoring data, and the data storage subunit stores multimodal monitoring data.

Images