An analysis method for extracting dynamic characteristics of a slamming flow field

Abstract

The application discloses an analysis method for dynamic feature extraction of a slamming flow field, and the analysis steps comprise the following steps: S1, acquiring a slamming load response quantity; S2, obtaining time-aligned flow field snapshot data; S3, obtaining a flow field in a unified reference domain; S4, constructing a flow field joint snapshot vector; S5, constructing a first joint snapshot matrix and a second joint snapshot matrix; S6, solving a low-dimensional spectral feature vector; S7, screening a main control space-time feature structure; and S8, using the main control space-time feature structure to perform low-order approximation reconstruction on the slamming flow field, and inversely mapping the reconstruction result from the unified reference domain to a physical domain of each original time, so as to identify a flow field transient dynamic feature. The excellent effect of the application is that the background interference, mode aliasing and scale domination caused by fixed regions, absolute time and single flow field analysis can be reduced, and the accuracy and physical interpretability of main control dynamic structure identification and cross-condition comparison of the slamming flow field are improved.

Description

Technical Field

[0001] This invention belongs to the technical fields of marine structural material dynamics data analysis, unsteady flow field order reduction modeling, and multi-flow field data processing. Specifically, it relates to an analysis method that uses slamming flow field time series results data as input, and realizes dynamic feature extraction and low-order reconstruction of slamming flow field through time alignment of slamming events, dynamic target region mapping, and construction of multi-flow field joint snapshot vectors. Background Technology

[0002] During high-speed navigation, wave encounters, or large-amplitude movements, floating offshore structures may experience high-speed relative interactions between their bottom, bow, sides, or localized outer plates and the free liquid surface, resulting in a slamming phenomenon characterized by short duration, high pressure gradients, severe interface deformation, and rapid changes in local flow scales. The slamming process is typically accompanied by instantaneous loading of localized high pressure, rapid expansion of the wetted surface, free liquid surface rise and breakup, jet formation, flow separation, and the generation and decay of vortices. These flow behaviors directly affect peak pressure, slamming impulse, impact range, and load propagation path, and are closely related to the structure's local strength, fatigue damage, and safety.

[0003] Current methods for analyzing slamming flow fields typically employ instantaneous contour plots, time-domain or frequency-domain analysis of local measurement points, peak power statistics, and reduced-order analysis based on orthogonal decomposition or dynamic mode decomposition. Instantaneous contour plots rely on manual interpretation, and local measurement points struggle to reflect the overall spatial organization. Peak power statistics fail to reveal the formation, propagation, and attenuation processes of high-pressure zones and interface structures. Orthogonal decomposition based on energy criteria can extract the main spatial structures, but multiple time scales may be mixed within the same mode. Dynamic mode decomposition based on the evolution relationship of adjacent snapshots can provide frequency and growth / attenuation information, but its direct application to slamming flow fields still faces the following problems.

[0004] First, traditional methods typically construct snapshots within a fixed spatial region. During the slamming process, the free interface, wet surface, and high-pressure zone all move rapidly. The fixed region contains a large amount of background flow unrelated to the target event and may misidentify interface translation and target region migration as dynamic structures, resulting in modal aliasing.

[0005] Second, traditional methods often use equal-interval sampling and directly compare different operating conditions using absolute time. Since the initial contact time, pressure peak time, and unloading duration are different under different operating conditions, snapshots without event localization and time alignment are difficult to form comparable dynamic evolution relationships. At the same time, equal-interval sampling may result in insufficient sampling during the short, rapid loading phase and redundant samples during the slow decay phase.

[0006] Third, pressure, velocity, liquid volume fraction, and vorticity reflect pressure impact, momentum transport, interface morphology, and vortex evolution, respectively. Analyzing only a single flow field makes it difficult to uniformly represent the entire impact process; if flow fields with different dimensions are directly spliced ​​together, physical quantities with larger numerical scales will dominate singular value decomposition, causing the joint modes to lose their clear physical meaning.

[0007] Therefore, an analytical method is needed to address the strong transient characteristics of impacts, which can complete impact event localization, time normalization, dynamic target region extraction, and multi-flow field scale coordination before constructing a dynamic evolution model. This would improve the physical interpretability of the modes, the comparability between different operating conditions, and the accuracy of low-order reconstruction.

[0008] The information disclosed in this background section is only for understanding the background technology of the inventive concept, and therefore may include information that does not constitute prior art. Summary of the Invention

[0009] This invention aims to solve the problems of background interference and geometric drift caused by a fixed analysis area, inconsistent timing of different impact events, and inconsistencies in the dimensions and scales of multiple flow fields, and provides an analysis method for dynamic feature extraction of impact flow fields.

[0010] An analytical method for dynamic feature extraction of slamming flow fields includes the following steps:

[0011] S1: Obtain the flow field physical quantities at multiple moments during the impact of the floating marine structure, and simultaneously obtain the impact load response.

[0012] S2: Construct slamming event indices based on slamming load response, determine the characteristic stages of the entire slamming response process, and map the original sampling times within the characteristic stages of the entire slamming response process to the normalized slamming time axis to obtain time-aligned flow field snapshot data on the normalized slamming time axis;

[0013] S3: Determine the instantaneous free interface based on the liquid volume fraction field at each moment, determine the high gradient flow region based on the flow field physical quantities, merge the neighborhood of the instantaneous free interface with the high gradient flow region to form a dynamic target region that changes with time, and map the dynamic target region at each moment to a unified reference domain to obtain the flow field in the unified reference domain.

[0014] S4: Dimensionless transformation is performed on different flow fields mapped to a unified reference domain, and physical weights are determined based on the response of each flow field to the slamming load, and a joint snapshot vector of the flow fields is constructed.

[0015] S5: Construct the first joint snapshot matrix and the second joint snapshot matrix based on the joint snapshot vector of the flow field at adjacent normalized impact moments;

[0016] S6: Perform truncated singular value decomposition on the first joint snapshot matrix, construct a low-dimensional spectral evolution operator, and solve the spectral features of the low-dimensional spectral evolution operator to obtain the low-dimensional spectral eigenvectors.

[0017] S7: Backmap the low-dimensional spectral feature vectors to a unified reference domain to obtain a high-dimensional spatiotemporal feature structure of multi-flow field coupling, and select the main spatiotemporal feature structure from the high-temperature spatiotemporal feature structure.

[0018] S8: Use the master spatiotemporal feature structure to perform low-order approximate reconstruction of the slamming flow field, and inversely map the reconstruction results from the unified reference domain to the physical domain of each original moment to identify the transient dynamic characteristics of the flow field.

[0019] As an optional implementation, the slam event indicator It can be expressed by formula (1):

[0020] (1)

[0021] in, , and The weights of the slam events are non-negative, and Unused response values ​​have a weight of 0; the horizontal line indicates the normalized response value.

[0022] 3. The analytical method for dynamic feature extraction of slamming flow field according to claim 1, characterized in that any of the response quantities... The normalized expression is given by formula (2):

[0023] (2)

[0024] in, and These represent the minimum and maximum values ​​of the response within the target time window, respectively.

[0025] As an optional implementation, step S2, determining the characteristic stages of the entire slam response process, includes:

[0026] 1) Based on the impact event index E(t) and the fluid-structure contact data, the initial contact time tc is determined by formula (3):

[0027] (3)

[0028] in, The contact detection threshold, For event indicators The maximum value, The percentage of liquid phase in the neighborhood of the structural surface. This represents the threshold for the proportion of liquid phase.

[0029] 2) Determine the peak response time using formula (4) based on the maximum pressure or the maximum impact force. :

[0030] or (4)

[0031] 3) Determine the end time of unloading decay using formula (5) :

[0032] (5)

[0033] in, The threshold for ending the uninstallation process. To ensure the duration of the stable criterion.

[0034] As an optional implementation, the step of acquiring the flow field snapshot data includes:

[0035] 1) Map the original time t to the normalized slamming time τ in segments, and express it using formula (6):

[0036] (6)

[0037] in, The starting point of the pre-contact time window, , , and Preset stage nodes;

[0038] 2) The time-aligned flow field snapshot data obtained on the normalized time axis is expressed by formula (7):

[0039] (7)

[0040] in, For interpolation operators from raw time to normalized time, This represents the normalized number of sampling points.

[0041] As an optional implementation, the instantaneous free interface is defined by formula (8). It means that, among them, This is the value used for interface judgment, and it is generally set to 0.5;

[0042] The high gradient flow region is defined by formula (9). It means that, among them, , and These are the normalized pressure gradient, velocity gradient, and vorticity index, respectively. , and Non-negative weights High gradient threshold;

[0043] The dynamic target region is defined by formula (10). It means that, among them, This refers to the near-wall region of the structure. The scale is represented as Buffer expansion operation;

[0044] The mapping is given by formula (11). It means that, among them, The scaling transformation matrix is... As a reference point for the dynamic target area, To unify the corresponding reference points in the reference domain;

[0045] The flow field within the unified reference domain is given by formula (12). It means that, among them, It is a conserved mapping operator from the original physical domain to the unified reference domain.

[0046] As an optional implementation, the joint snapshot vector of the flow field is expressed by formula (18). It means that, among them, , , and This is the spatial weight matrix.

[0047] As an optional implementation, the impact event index is a weighted average of at least two of the following: the rate of change of maximum pressure on the structural surface, the rate of change of wetted surface area, and the rate of change of impact resultant force.

[0048] As an optional implementation, the normalized impact time axis includes a pre-contact phase, a rapid loading phase, a peak phase, and an unloading decay phase, and the original sampling time mapping includes a piecewise linear mapping or a monotonic interpolation mapping.

[0049] As an optional implementation, the dynamic target region includes at least one of the following: the free interface strip region on both sides of the liquid volume fraction isosurface, the near-wall region of the structure, and the high gradient flow region. The high gradient flow region satisfies at least one of the following: pressure gradient threshold, velocity gradient threshold, and vorticity threshold.

[0050] As an optional implementation, the unified reference domain is a structurally fixed reference domain, a free interface fixed reference domain, or a local reference domain composed of a combination of the two; the mapping method is one or more of the following: translation, rotation, scale normalization, and coordinate transformation based on mesh interpolation.

[0051] As an optional implementation, the joint snapshot vector of the flow field is composed of at least two of the following: dimensionless pressure field, dimensionless velocity field, liquid volume fraction field, and dimensionless vorticity field, spliced ​​together with preset weights.

[0052] As an optional implementation, the physical weights are determined based on the root mean square, characteristic scale, pressure impulse sensitivity, or load reconstruction sensitivity of each flow field within the target time window, and satisfy the condition that the sum of the squares of each physical weight is 1.

[0053] As an alternative implementation, the truncation order of singular values ​​is determined based on the cumulative singular value contribution rate and the slam response reconstruction error.

[0054] As an optional implementation, the master spatiotemporal characteristic structure is determined by ranking multiple evaluation indicators, including at least two of the following: structural amplitude, characteristic frequency, growth decay rate, contribution to the resultant force and impulse of the impact, and the change in reconstruction error.

[0055] As an optional implementation, the methods for obtaining the flow field physical quantities and the slamming load response quantities include one or more of the following: numerical simulation, particle image velocimetry, pressure-sensitive coating, and high-speed photography.

[0056] The present invention has at least the following effects:

[0057] (1) Constructing a dynamic target region based on the instantaneous free interface and high gradient flow region and mapping it to a unified reference domain can reduce modal aliasing caused by far-field background flow, overall interface translation and target region migration.

[0058] (2) By locating the contact, peak and unloading stages through the impact event index and performing normalized time mapping, the time resolution near the peak can be improved, making the joint snapshot vector of the flow field under different working conditions and different impact events comparable at the same physical stage.

[0059] (3) By constructing a joint snapshot vector through dimensionless, physical weight and spatial weight of different flow fields, pressure impact, momentum transport, free interface deformation and vortex structure evolution can be uniformly represented in the same dynamic feature structure, avoiding the scale problem caused by direct splicing.

[0060] (4) The identification and time alignment of slamming events, dynamic target region mapping, dimensionless multi-flow field and physical weighting are completed before the low-dimensional spectral evolution analysis. This makes the dynamic feature solution no longer just a direct application of traditional decomposition methods to slamming data, but an analysis based on a dedicated joint snapshot vector after event alignment, region correction and physical coordination, thereby improving the physical interpretability of the identification of the master spatiotemporal feature structure and low-order reconstruction. Attached Figure Description

[0061] Figure 1 For analysis of flowcharts;

[0062] Figure 2 A schematic diagram of the mapping between characteristic moments of a slam event and normalized time.

[0063] Figure 3 A schematic diagram of the dynamic target region and unified reference domain mapping;

[0064] Figure 4 A schematic diagram of constructing a joint snapshot vector for multiple flow fields;

[0065] Figure 5 This is a schematic diagram of the model and computational domain;

[0066] Figure 6 This is the average flow field diagram near the free interface;

[0067] Figure 7 This is a diagram showing the spatiotemporal structural spectral features and parameter distribution.

[0068] Figure 8 This is a spatial distribution map of spatiotemporal characteristic structures. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] This embodiment provides the following technical solution, and the analysis process is attached. Figure 1 As shown. This application does not limit the method of generating the result data, nor does it involve the flow field control equations, solvers, computational domains, meshes, or boundary conditions used to generate the result data.

[0071] This embodiment uses three-dimensional numerical simulation data of a strong transient impact process on a structure to extract the dominant dynamic features of the transient velocity field near the free interface, in order to identify the characteristic frequencies, growth and decay patterns, and dominant spatial structures in the strong transient unsteady flow field. The target structure has a total length of 2.3085 m, a maximum width of 0.9 m, and a depth of 0.649 m. The computational domain has a length of 8 m, a width of 4 m, and a water depth of 4.5 m to ensure that the deformation of the free interface and the evolution of the local flow field during the strong transient impact can be fully unfolded within the computational domain.

[0072] The numerical simulation employs an unsteady solver to describe the coupled evolution of the fluid and free interface. The fluid medium is considered an incompressible viscous fluid, and its governing equations are the continuity equation and the momentum conservation equation. The free interface is captured using the fluid volume fraction method. Mesh generation employs a partitioned refinement strategy combining background and overlapping meshes. Overlapping mesh regions are set near the target structure to cover the high-gradient flow region under strong transient effects, while boundary layer meshes are set near the wall to represent near-wall viscous effects. Boundary conditions are set with the top of the computational domain as a pressure outlet and the remaining boundaries as no-slip walls. The model and computational domain layout are detailed below. Figure 2 .

[0073] Considering the depth of the computational water area, the visualization of the entire flow field is easily affected by the scale effect of the deep background flow, weakening the identification effect of strong transient flow field characteristics near the free interface. Therefore, this embodiment only extracts the near-field region within approximately 0.6 m below the free interface as the target analysis area. This region exhibits significant local pressure concentration and a large pressure gradient, which can effectively represent the main dynamic behavior of strong transient flows. Furthermore, the instantaneous velocity field at multiple consecutive moments within this region is extracted as sample data. The sampling time window covers the key stages before and after the occurrence of the strong transient peak, selecting a total of 300 consecutive snapshots (the snapshot matrix contains flow field physical quantities and slamming load response quantities). The total duration is 0.75 s, used to construct the time-series snapshot matrix. The target region extraction and average flow field distribution are shown in the attached figure. Figure 6 As shown.

[0074] The specific steps are as follows:

[0075] S1: Obtain the flow field physical quantities at multiple moments during the impact of the floating marine structure, and simultaneously obtain the impact load response.

[0076] First, read the result data at each moment, along with its spatial coordinates and time markers, and then smooth the maximum pressure, wetted surface area, and impact force on the structural surface.

[0077] Numerical simulations or experimental measurements were performed on the impact process of floating offshore structures to obtain the pressure field at multiple consecutive moments. Velocity field Liquid phase volume fraction field and vorticity field At least two of them. Simultaneously obtain the maximum pressure on the structural surface. wetted surface area , combined force of impact or slam impulse At least one synchronous response quantity is used. Each data point has unified or interconvertible spatial reference information and a clear time signature. It should be noted that in this embodiment, the flow field physical quantities are the velocity field and vorticity field, and the impact load response quantities are the structural surface pressure, wetted surface area, and impact resultant force.

[0078] S2: Construct slamming event indices based on slamming load response, determine the characteristic stages of the entire slamming response process, and map the original sampling times within the characteristic stages of the entire slamming response process to the normalized slamming time axis to obtain time-aligned flow field snapshot data on the normalized slamming time axis.

[0079] It should be explained that the impact event index is composed of at least two of the following weighted values: the rate of change of maximum pressure on the structural surface, the rate of change of wetted surface area, and the rate of change of resultant impact force. In this embodiment, the maximum pressure, wetted surface area, and resultant impact force are selected as the impact load response quantities for demonstration. First, the normalized rate of change of the maximum pressure, wetted surface area, and resultant impact force on the structural surface are calculated.

[0080] Bombing event indicators It can be represented as:

[0081] (1)

[0082] in, , and The weights of the slam events are non-negative, and Unused response values ​​have a weight of 0; the horizontal line represents the normalized response value. Any response value The normalized expression is:

[0083] (2)

[0084] in, and These represent the minimum and maximum values ​​of the response within the target time window, respectively. The initial contact time can be determined based on event indicators and fluid-structure contact data. :

[0085] (3)

[0086] in, The contact detection threshold, For event indicators The maximum value, The percentage of liquid phase in the neighborhood of the structural surface. This is the threshold for the liquid phase percentage. The peak response time can then be determined based on the maximum pressure or the maximum impact force. :

[0087] or (4)

[0088] Unloading at the end of decay It can be represented as:

[0089] (5)

[0090] in, The threshold for ending the uninstallation process. To stabilize the criterion duration. To ensure comparability of different operating conditions or different impact events at the same dynamic stage, the original time... Mapped to normalized slam time The normalized impact timeline includes the pre-contact stage, rapid loading stage, peak stage, and unloading decay stage. The original sampling time mapping includes piecewise linear mapping or monotonic interpolation mapping. This embodiment uses the pre-contact, rapid loading, peak, and unloading decay stages as examples, employing piecewise mapping:

[0091] (6)

[0092] in, The starting point of the pre-contact time window, , , and This sets the preset stage nodes. Then, sampling nodes are set on the normalized time axis. And the time-aligned flow field is obtained through interpolation:

[0093] (7)

[0094] in, For interpolation operators from raw time to normalized time, This represents the normalized number of sampling points. The sampling interval during the fast loading phase and near the peak response time is preferably smaller than the sampling interval during the unloading and decay phase. The mapping relationship between event characteristic moments and normalized time is shown in the appendix. Figure 3 As shown.

[0095] It should be explained that denser sampling points are set during the rapid loading phase and near the peak, while sparser sampling points are set during the unloading and decay phase. In this step, event indicators are constructed according to formulas (1) and (2), and the initial contact, peak response, and unloading decay end time are identified according to formulas (3)-(5). Subsequently, the original time of each working condition is mapped to a unified normalized time interval according to formula (6), and finally, the time-aligned pressure field, velocity field, liquid volume fraction field, and vorticity field are obtained at the normalized sampling node according to formula (7).

[0096] Step S3: Determine the instantaneous free interface based on the liquid volume fraction field at each moment, determine the high gradient flow region based on the flow field physical quantities, merge the neighborhood of the instantaneous free interface with the high gradient flow region to form a dynamic target region that changes with time, and map the dynamic target region at each moment to a unified reference domain to obtain the flow field of the unified reference domain.

[0097] It should be explained that the dynamic target region includes at least one of the following: the free interface strip region on both sides of the liquid volume fraction isosurface, the near-wall region of the structure, and the high gradient flow region. The high gradient flow region must satisfy at least one of the following: pressure gradient threshold, velocity gradient threshold, and vorticity threshold.

[0098] The unified reference domain can be a structurally fixed reference domain, a free interface fixed reference domain, or a local reference domain composed of a combination of the two; the mapping method can be one or more of the following: translation, rotation, scale normalization, and coordinate transformation based on mesh interpolation.

[0099] For each normalized time step Determine the instantaneous free interface based on the isosurface of liquid phase volume fraction. :

[0100] (8)

[0101] in, This is the value used for interface judgment, and is generally set to 0.5. Construct a free interface neighborhood of preset thickness around the center. Simultaneously, a high-gradient flow region is constructed based on the pressure gradient, velocity gradient, and vorticity. :

[0102] (9)

[0103] in, , and These are the normalized pressure gradient, velocity gradient, and vorticity index, respectively. , and Non-negative weights This is a high gradient threshold. And for dynamic target regions... It can be defined as:

[0104] (10)

[0105] in, This refers to the near-wall region of the structure. The scale is represented as The buffer expansion operation can also filter the merged region, retaining only the regions connected to the structural surface or transient free interface to remove far-field isolated disturbances.

[0106] Since the dynamic target region differs in position, orientation, and scale at different times, translation and rotation methods are used to map it onto a unified reference domain. Let the mapping be... ,but:

[0107] (11)

[0108] in, The scaling transformation matrix is... As a reference point for the dynamic target area, For the corresponding reference point in the unified reference domain, the flow field within the unified reference domain is:

[0109] (12)

[0110] in, This is a conservation mapping operator from the original physical domain to the unified reference domain. The mapping reduces the non-physical effects of global translation of free interfaces, rigid body motion of structures, wet surface expansion, and high-pressure region migration on dynamic eigenvalue decomposition. The mapping relationship is as follows: Figure 4 As shown.

[0111] It should be explained that in this step, at each normalization time, the instantaneous free interface is extracted according to formula (8), and the normalized pressure gradient, velocity gradient, and vorticity index are calculated. The high gradient flow region is obtained according to formula (9), and it is merged with the free interface neighborhood and the near-wall region of the structure. Connectivity filtering and buffer expansion are performed, and the dynamic target region is obtained according to formula (10). Based on the structural fixed coordinates and the direction of the local free interface, the dynamic target region at each time is mapped to the unified reference domain according to formula (11), and the flow field is interpolated according to formula (12). During the mapping process, the corresponding positions of the structural surface and the order of the unified grid nodes remain unchanged.

[0112] S4: Dimensionless transformation is performed on different flow fields mapped to a unified reference domain, and physical weights are determined based on the response of each flow field to the slamming load, thus constructing a joint snapshot vector of the flow fields.

[0113] The pressure field, velocity field, and vorticity field can be expressed as follows:

[0114] (13)

[0115] (14)

[0116] (15)

[0117] in, Indicates a unified reference domain coordinate system. and These are background pressure and background velocity, respectively. , and These are characteristic pressure, characteristic velocity, and characteristic length, respectively; the liquid volume fraction itself is a dimensionless quantity, which can be used directly or its time average can be subtracted.

[0118] To prevent a single flow field from dominating the joint decomposition due to its large numerical scale, physical weights are set for pressure, velocity, phase distribution, and vorticity. , , and And satisfy:

[0119] (16)

[0120] Physical weights can be determined based on characteristic scale balance, root mean square balance, or slam response reconstruction sensitivity. One approach is based on the decrease in response error. The weights are determined as follows:

[0121] (17)

[0122] in, and Indicates different flow fields, For small positive numbers. The discrete values ​​of each flow field on a unified grid are expanded in a preset order, the first... The joint snapshot vector of multiple flow fields is:

[0123] (18)

[0124] in, , , and This is the spatial weighting matrix, used to enhance the contribution of the near-wall region, free interface neighborhood, or high-gradient region; unused flow field corresponding blocks are removed from the joint snapshot vector. The joint snapshot vector is constructed as follows: Figure 5 As shown.

[0125] It should be explained that the joint snapshot vector of the flow field is composed of at least two of the following: dimensionless pressure field, dimensionless velocity field, liquid volume fraction field, and dimensionless vorticity field, concatenated with preset weights. The joint snapshot vector of the flow field can be obtained through one or more of the following methods: numerical simulation, particle image velocimetry, pressure-sensitive coating, and high-speed photography. In this embodiment, numerical simulation is selected as the method for obtaining the joint snapshot vector of the flow field.

[0126] The pressure characteristic scale can be the peak pressure on the structural surface or the preset dynamic pressure scale within the target time window; the velocity characteristic scale can be the relative velocity between the structure and the water surface; and the length characteristic scale can be the characteristic width of the local impact region. Dimensionlessness is achieved according to formulas (13)-(15). In this embodiment, a dimensionless pressure field, a dimensionless velocity field, and a liquid volume fraction field are selected to form a joint snapshot vector; when a vorticity field is introduced, the vorticity field is further dimensionless using characteristic length and characteristic velocity. First, the physical weights are determined through characteristic scale balancing or single-field pre-analysis. It should be explained that the physical weights are determined based on the root mean square, characteristic scale, pressure impulse sensitivity, or load reconstruction sensitivity of each flow field within the target time window, and the sum of the squares of each physical weight is 1. Then, the weights are normalized according to formulas (16) and (17), and higher spatial weights are set in the near-wall region of the structure and the neighborhood of the free interface. A joint snapshot vector for each normalized time is formed according to formula (18).

[0127] S5: Construct the first joint snapshot matrix and the second joint snapshot matrix based on the joint snapshot vector of the flow field at adjacent normalized impact times.

[0128] Construct the first joint snapshot matrix using the joint snapshot vectors of adjacent normalized time steps. Second joint snapshot matrix :

[0129] (19)

[0130] (20)

[0131] S6: Perform truncated singular value decomposition on the first joint snapshot matrix to construct a low-dimensional spectral evolution operator, and solve the spectral features of the low-dimensional spectral evolution operator to obtain the low-dimensional spectral eigenvector.

[0132] Perform truncated singular value decomposition on the first joint snapshot matrix:

[0133] (twenty one)

[0134] in, For the front Left singular vector matrix of order 1 For the front A singular value diagonal matrix of order 1. For the front Right singular vector matrix of order, superscript This represents the conjugate transpose. It should be explained that the truncation order of singular values ​​is determined based on the cumulative singular value contribution rate and the slam response reconstruction error. Singular value truncation order Determined based on the cumulative singular value contribution rate:

[0135] (twenty two)

[0136] in, The total number of non-zero singular values. The cumulative contribution rate threshold.

[0137] Furthermore, it is possible to Constructing low-dimensional spectral evolution operators within the resulting low-dimensional subspace:

[0138] (twenty three)

[0139] Eigenvalue decomposition of low-dimensional spectral evolution operators:

[0140] (twenty four)

[0141] in, For discrete spectrum eigenvalue matrix, It is a low-dimensional spectral eigenvector matrix.

[0142] It should be explained that in this step, the joint snapshot matrix of adjacent time moments is constructed according to formulas (19) and (20), and the low-dimensional spectral evolution operator and its spectral characteristics are obtained according to formulas (21-)-(24). The truncation order simultaneously satisfies the requirements of cumulative singular value contribution rate and slam response reconstruction error.

[0143] S7: Backmap the low-dimensional spectral feature vectors to a unified reference domain to obtain a high-dimensional spatiotemporal feature structure of multi-flow field coupling, and then select the dominant spatiotemporal feature structure from the high-temperature spatiotemporal feature structure.

[0144] By back-mapping the low-dimensional spectral feature vectors to a unified reference domain, we obtain a high-dimensional joint spatiotemporal feature structure:

[0145] (25)

[0146] Discrete spectral eigenvalues Convert to continuous spectrum parameters :

[0147] (26)

[0148] in, This represents the interval between adjacent normalized sampling times. For non-uniform sampling, corresponding intervals can be used at each stage, or the continuous spectrum parameters can be obtained through a least-squares continuous-time model. Growth and decay rates of spatiotemporal characteristic structures and characteristic frequencies They are respectively:

[0149] (27)

[0150] (28)

[0151] The structural magnitude is obtained from the initial joint snapshot vector:

[0152] (28)

[0153] in, for The Moore-Penrose generalized inverse can also be used to screen the dominant spatiotemporal feature structure, and a comprehensive response contribution index can be defined:

[0154] (29)

[0155] in, To join the The decrease in target response reconstruction error after the first-order spatiotemporal feature structure. This represents the weighted strength of the structure in the near-wall region and the neighborhood of the free interface. , and The weights are non-negative and their sum is 1. According to... Select the main spatiotemporal characteristic structure from large to small until the joint field reconstruction error or the bang response reconstruction error meets the preset conditions.

[0156] It should be explained that the dominant spatiotemporal characteristic structure is determined by ranking multiple evaluation indicators, including at least two of the following: structural amplitude, characteristic frequency, growth and decay rate, contribution to the resultant force and impulse of the impact, and the change in reconstruction error. This step follows formula (25). , formula (26) , formula (27) , formula (28) , formula (29) We obtain high-dimensional spatiotemporal feature structures, and then select the dominant spatiotemporal feature structures from these high-dimensional spatiotemporal feature structures.

[0157] S8: The slamming flow field is reconstructed using the master spatiotemporal feature structure, and the reconstruction results are inversely mapped from the unified reference domain to the physical domain of each original moment to identify the transient dynamic characteristics of the flow field.

[0158] It should be explained that the transient dynamic characteristics of the flow field include propagation in high-pressure areas, expansion of wetted surfaces, lifting or breaking up of free interfaces, development of local jets, and generation and decay of vortex structures.

[0159] Select the set of master spatiotemporal feature structures Then, the joint flow field is reconstructed within the unified reference domain:

[0160] (30)

[0161] The reconstructed joint snapshot vector is split into flow field blocks and dedimensionalized, then based on the first... The coordinate correspondences and interpolation relationships saved at time step 1 will restore the reconstruction results in the unified reference domain to the time step 2. The original physical domain at each moment.

[0162] (31)

[0163] in, To unify the inverse interpolation operator from the reference domain to the original physical domain, This indicates that in the unified reference domain, at the normalized time... The reconstructed flow field results are obtained. Based on the reconstruction results, the propagation of local high-pressure areas, wet surface expansion, free interface lifting or breaking, jet development, and vortex structure generation and attenuation processes are identified.

[0164] This step follows formula (30). and formula (31) Complete the low-order reconstruction and the original physical domain back mapping.

[0165] The results of the spectral characteristic parameter analysis are attached. Figure 7 As shown, the discrete spectral eigenvalues ​​corresponding to each dominant spatiotemporal characteristic structure are mainly distributed within or near the unit circle, indicating that within the selected time window, the strong transient unsteady flow field response is mainly dominated by decaying or near-neutral stable dynamic structures, and no obvious divergent unstable spatiotemporal characteristic structures appear. Further analysis of the growth rate of the spatiotemporal characteristic structures reveals that the growth rate of each order of spatiotemporal characteristic structure is generally negative, and the decay rates of different spatiotemporal characteristic structures differ, indicating that flow structures at different scales have different dissipation characteristics after strong transient excitation.

[0166] Based on this, analysis of the relationship between structural amplitude and frequency distribution reveals that the large-amplitude spatiotemporal characteristic structures are mainly concentrated in the low-frequency range and exhibit a paired distribution. This indicates that the dominant dynamic behavior of the strong transient unsteady flow field is primarily controlled by the low-frequency spatiotemporal characteristic structures. Simultaneously, the real-valued flow field, after low-dimensional spectral mapping, displays the mathematical characteristics of conjugate characteristic structures. These results demonstrate that the method of this invention can effectively reveal the low-frequency dominant characteristics and energy attenuation mechanism of the strong transient flow field from a frequency domain perspective.

[0167] A representative spatiotemporal feature structure covering the low-frequency to high-frequency range was selected for spatial distribution analysis. The analysis results are shown in the appendix. Figure 8It can be seen that the significant structures at different frequencies are mainly concentrated near the target structure and the neighborhood of the free interface, and their overall spatial topology remains basically consistent. The differences are mainly reflected in the local intensity distribution and fluctuation details. This indicates that within a very short timescale, the dominant structure of the strong transient flow field is constrained by both the geometric boundary and the free interface, and its dynamic differences are mainly manifested as the oscillation and decay behavior of the same spatial structure under different frequency conditions.

[0168] Further, a low-order approximate reconstruction of the flow field in the target region is performed using the dominant spatiotemporal feature structure to obtain an approximate expression of the main spatiotemporal evolution processes during strong transient flow. This low-order approximate reconstruction result can be used to identify major dynamic behaviors such as the propagation of local high-pressure areas, free interface deformation, and local vortex evolution, while reducing the influence of secondary disturbances and weak noise components on the results. This allows for a low-order expression and identification of dominant features in high-dimensional transient flow field data while preserving key physical information. In particular, within the strong transient time window, the large-amplitude spatiotemporal feature structure is concentrated in the low-frequency range, and the overall growth rate is negative, verifying that the flow field exhibits obvious low-frequency dominance and multi-scale attenuation characteristics.

[0169] In summary, this invention extracts time-series snapshot data from the target region of a strong transient unsteady flow field, constructs a time-shifted snapshot matrix of adjacent time points, performs dimensionality reduction decomposition and dynamic feature solving, obtains the spatial structure, characteristic frequencies, and growth and decay parameters corresponding to each order of spatiotemporal feature structure, and reconstructs the target flow field using the dominant spatiotemporal feature structure. This effectively identifies the dominant structure and its evolution law in complex strong transient flows. Compared with directly observing the instantaneous flow field or analyzing only local response signals, the method of this invention can simultaneously represent the spatial organization and temporal dynamic behavior of the flow field, improving the accuracy, completeness, and interpretability of strong transient unsteady flow field analysis.

[0170] The process and principle of this invention:

[0171] First, slamming flow field and synchronous response data are acquired. Then, slamming event indices are constructed and event-driven sampling and time alignment are performed to obtain time-aligned flow field snapshots. Next, a target region that dynamically changes with the slamming process is constructed and mapped to a unified reference domain. Then, a joint snapshot vector of multiple flow fields with slamming physical meaning is constructed. Then, the dimensions of each flow field in the unified reference domain are made dimensionless. Then, a joint snapshot matrix is ​​constructed and a low-dimensional spectral evolution operator is solved. Then, the high-dimensional spatiotemporal feature structure is obtained and the dominant spatiotemporal feature structure is selected. Finally, low-order reconstruction and inverse mapping are performed to reduce background interference, mode mixing, and scale dominance problems caused by fixed analysis region, absolute time sampling, and single flow field analysis, thereby improving the accuracy of identifying the dominant dynamic structure of the slamming flow field, cross-condition comparison, and transient dynamic interpretation.

[0172] Although embodiments of the invention have been shown and described (see the detailed description above), it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An analytical method for dynamic feature extraction of slamming flow fields, characterized in that, The steps of the analytical method include: S1: Obtain the flow field physical quantities at multiple moments during the impact of the floating marine structure, and simultaneously obtain the impact load response. S2: Construct slamming event index based on the slamming load response, determine the characteristic stages of the entire slamming response process, and map the original sampling time within the characteristic stages of the entire slamming response process to the normalized slamming time axis, and obtain time-aligned flow field snapshot data on the normalized slamming time axis; S3: Determine the instantaneous free interface based on the liquid volume fraction field at each moment, determine the high gradient flow region based on the flow field physical quantities, merge the neighborhood of the instantaneous free interface with the high gradient flow region to form a dynamic target region that changes with time, and map the dynamic target region at each moment to a unified reference domain to obtain the flow field within the unified reference domain. S4: Dimensionless transformation is performed on different flow fields mapped to the unified reference domain, and physical weights are determined based on the response of each flow field to the slamming load, and a joint snapshot vector of the flow fields is constructed. S5: Construct a first joint snapshot matrix and a second joint snapshot matrix based on the joint snapshot vector of the flow field at adjacent normalized impact moments; S6: Perform truncated singular value decomposition on the first joint snapshot matrix to construct a low-dimensional spectral evolution operator, and solve the spectral features of the low-dimensional spectral evolution operator to obtain a low-dimensional spectral feature vector. S7: Backmap the low-dimensional spectral feature vector to the unified reference domain to obtain a high-dimensional spatiotemporal feature structure of multi-flow field coupling, and select the master spatiotemporal feature structure from the high-temperature spatiotemporal feature structure. S8: The slamming flow field is reconstructed using the master spatiotemporal feature structure, and the reconstruction result is inversely mapped from the unified reference domain to the physical domain of each original moment to identify the transient dynamic characteristics of the flow field.

2. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The slamming event indicator It can be expressed by formula (1): （1） in, , and The weights of the slam events are non-negative, and Unused response values ​​have a weight of 0; the horizontal line indicates the normalized response value.

3. The analytical method for dynamic feature extraction of slamming flow field according to claim 1, characterized in that, Any of the aforementioned response quantities The normalized expression is given by formula (2): （2） in, and These represent the minimum and maximum values ​​of the response within the target time window, respectively.

4. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, Step S2, the step of determining the characteristic stages of the entire slam response process, includes: 1) Based on the impact event index E(t) and the fluid-structure contact data, the initial contact time tc is determined by formula (3): （3） in, The contact detection threshold, For event indicators The maximum value, The percentage of liquid phase in the neighborhood of the structural surface. This represents the threshold for the proportion of liquid phase. 2) Determine the peak response time using formula (4) based on the maximum pressure or the maximum impact force. : or (4) 3) Determine the end time of unloading decay using formula (5) : （5） in, The threshold for ending the uninstallation process. To ensure the duration of the stable criterion.

5. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The steps for obtaining the flow field snapshot data include: 1) Map the original time t to the normalized slamming time τ in segments, and express it using formula (6): （6） in, The starting point of the pre-contact time window, , , and Preset stage nodes; 2) The time-aligned flow field snapshot data obtained on the normalized time axis is expressed by formula (7): （7） in, For interpolation operators from raw time to normalized time, This represents the normalized number of sampling points.

6. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The instantaneous free interface is defined by formula (8). It means that, among them, This is the value used for interface judgment, and it is generally set to 0.5; The high gradient flow region is defined by formula (9). It means that, among them, , and These are the normalized pressure gradient, velocity gradient, and vorticity index, respectively. , and Non-negative weights High gradient threshold; The dynamic target region is defined by formula (10). It means that, among them, This refers to the near-wall region of the structure. The scale is represented as Buffer expansion operation; The mapping is given by formula (11). It means that, among them, The scaling transformation matrix is... As a reference point for the dynamic target area, To unify the corresponding reference points in the reference domain; The flow field within the unified reference domain is given by formula (12). It means that, among them, It is a conservation mapping operator from the original physical domain to the unified reference domain.

7. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The joint snapshot vector of the flow field is given by formula (18). It means that, among them, , , and This is the spatial weight matrix.

8. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The dynamic target region includes at least one of the following: the free interface strip region on both sides of the liquid volume fraction isosurface, the near-wall region of the structure, and the high gradient flow region. The high gradient flow region satisfies at least one of the following: pressure gradient threshold, velocity gradient threshold, and vorticity threshold.

9. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The physical weights are determined based on the root mean square, characteristic scale, pressure impulse sensitivity, or load reconstruction sensitivity of each flow field within the target time window, and satisfy the condition that the sum of the squares of each physical weight is 1.

10. The analytical method for dynamic feature extraction of slamming flow fields according to claim 1, characterized in that, The master spatiotemporal feature structure is determined by sorting multiple evaluation indicators, including the contribution of the impact resultant force and impact impulse, as well as the change in reconstruction error.

Images