Wavelet basis optimization-based near-fault multi-component ground motion pulse identification method
Patent Information
- Application Number
- CN202611013399.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-08
- Publication Date
- 2026-09-15
Smart Images
Figure CN122755084A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of earthquake engineering and ground motion signal processing technology, specifically to a method for identifying near-fault multi-component ground motion pulses based on wavelet basis optimization. Background Technology
[0002] Frequent earthquakes pose a significant threat to the safety of building structures. In near-fault regions, due to rupture directionality and slippage effects, seismic waves are often accompanied by highly destructive velocity pulses. These velocity pulses carry highly concentrated energy, easily causing severe damage or even instantaneous collapse of engineering structures. Accurately extracting and identifying these pulse components is the physical basis for developing seismic design codes. Engineers rely on reliable pulse parameter systems to scientifically conduct disaster risk assessments and structural dynamic response calculations.
[0003] Early identification of velocity pulses relied primarily on manual observation of the presence of obvious waveforms in velocity time histories, but this method lacked objective quantitative evidence. With technological advancements, researchers have gradually proposed methods based on mathematical models to characterize pulses and methods based on signal processing to identify them. Among numerous signal processing schemes, wavelet theory, with its advantage of localization analysis in both time and frequency domains, has been the most widely used (e.g., the existing Shahi method). Researchers utilize continuous wavelet transform to process raw acceleration and velocity time history data. This method can simultaneously perform two-dimensional unfolding and analysis of the signal in both the time and frequency domains. By calculating the wavelet coefficients of the one-dimensional signal, existing techniques can preliminarily locate low-frequency energy concentration areas in seismic waves. This mathematical processing approach breaks away from the early, crude method of relying on manual waveform observation, realizing the basic proceduralization of seismic motion record analysis and providing an effective computational tool for extracting local fluctuations in specific frequency bands.
[0004] However, existing wavelet transform-based signal processing techniques have significant technical flaws in their underlying computational logic. On one hand, traditional algorithms forcibly bind to a single, specific wavelet basis function. Near-fault rupture mechanisms are extremely complex, and the actual generated pulse waveforms vary greatly. Fixed-form basis functions cannot accommodate this physical diversity, resulting in low waveform matching and directly distorting the extracted pulses. This leads to a significant weakening of peak data and difficulty in controlling the reconstruction residuals. On the other hand, previous data extraction was mostly limited to a single component of parallel or perpendicular faults. Seismic energy essentially propagates as a spatial vector, and the direction of the strongest actual pulse usually deviates from the preset orthogonal coordinate axes. Performing only single-component calculations completely severs the spatial directional attribute of seismic motion, inevitably leading to the complete omission of hidden strong velocity pulses in space due to projection angle deviations. Furthermore, current verification processes generally rely on absolute velocity thresholds as cutoff conditions. Seismic events vary in size, and rigid numerical boundaries easily misidentify medium-intensity but characteristically standard real pulses. The energy of a real pulse must be concentrated in the early stages of seismic wave propagation. Existing discrimination systems lack consideration for the energy arrival time dimension. The algorithm is prone to mistaking slow, dragging ordinary low-frequency bands in earthquake coma for pulses, thus generating a large number of false identification results. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a near-fault multi-component ground motion pulse identification method based on wavelet basis optimization. It aims to solve the problems of pulse extraction distortion caused by poor matching degree of fixed wavelet basis functions, omission of strong pulses due to neglect of spatial direction effect in single-component calculation, and misjudgment and omission caused by the lack of energy arrival time assessment in the discrimination criteria.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying near-fault multi-component ground motion pulses based on wavelet basis optimization, comprising the following steps: S100: Collect orthogonal component data of ground motion to form orthogonal ground motion acceleration time history dataset, perform filtering preprocessing on orthogonal ground motion acceleration time history dataset, and establish a continuous wavelet transform wavelet basis function database. S200 performs multi-angle continuous wavelet transform on the preprocessed orthogonal ground motion acceleration time history dataset, calculates the maximum value of the wavelet coefficients of the linear combination under different spatial directions to determine the direction of the strongest velocity pulse; S300 uses the ground motion velocity time history under the direction of the strongest velocity pulse as the initial record, and traverses the wavelet basis function database to extract the potential pulse components under different wavelet basis functions. S400 calculates the energy ratio index of potential pulse components and extracts the cumulative energy arrival time parameter. It then uses a dual discrimination criterion to determine the effectiveness of potential pulse components under different wavelet basis functions. S500 calculates the improved dimensionless matching evaluation parameters between the potential pulse component velocity time history and the original ground motion velocity time history, determines the optimal wavelet basis function based on the similarity of the signal waveform, and outputs the corresponding optimal strong velocity pulse signal and pulse period.
[0007] The steps for implementing step S100 are as follows: S110: Collect orthogonal component data of the target ground motion record and reconstruct it to generate a three-dimensional orthogonal ground motion acceleration time history dataset; S120 performs filtering on the three-dimensional orthogonal ground motion acceleration time history dataset to achieve baseline correction and high-frequency noise elimination; S130, Establish a wavelet basis function database for continuous wavelet transform matching.
[0008] In step S130, the wavelet basis function database is divided into the following five categories: The first set of wavelets includes db2, db3, db4, db5, db6, db7, db8, coif1, coif2, sym2, sym3, and sym4; The second set of wavelets includes gaus1, gaus2, gaus3, gaus4, gaus5, gaus6, gaus7 and gaus8; The third set of wavelets includes fk6, fk8 and fk14; The fourth type of wavelet basis function is specified as rbio2.4; The fifth type of wavelet basis function is specified as bior1.3.
[0009] The steps for implementing step S200 are as follows: S210, based on the linear combination of bidirectional orthogonal ground motions at any angle in the horizontal plane, derive and calculate the corresponding wavelet coefficients and extreme directions; The analytical expression for the wavelet coefficients of horizontal combined ground motion is: ;in, The first wavelet coefficient in the horizontal direction, The wavelet coefficients of the second horizontal component ; For wavelet scaling parameters, corresponding to frequency or scaling; These are the wavelet shift parameters, corresponding to the time position; For time; It is the angle between two orthogonal ground motions in the horizontal plane, i.e., the horizontal rotation angle; The maximum value of the horizontal combined ground motion wavelet coefficients is: ; The direction angle of the strongest velocity pulse corresponding to this extreme value The calculation formula is: ; S220 extends the orientation recognition dimension from a two-dimensional horizontal plane to a three-dimensional space, and calculates the wavelet coefficients of the linear combination of three-dimensional orthogonal ground motion under arbitrary spatial azimuth and pitch angles. ; Among them, among them, The first wavelet coefficient in the horizontal direction, The wavelet coefficients are the second component in the horizontal direction. These are the wavelet coefficients for the vertical ground motion component. In specific low-level signal processing, a discretization approximation method can be used to achieve numerical integration. The angle between two orthogonal ground motions in the horizontal plane; The pitch angle; S230 extracts the extreme values of the combined wavelet coefficients and locks the direction of the strongest velocity pulse in three-dimensional space; The maximum value of the wavelet coefficient of the three-component combined ground motion is: ; The formulas for calculating the coordinates of the direction of the strongest velocity pulse in space corresponding to this maximum value are as follows: ; .
[0010] The steps for implementing step S300 are as follows: S310, Extract the seismic velocity time history in the direction of the strongest velocity pulse calculated in step S200, and construct it as the initial input record; S320, Traverse the wavelet basis function database established in step S100, and perform continuous wavelet transform and pulse separation on the initial input record; S330, reconstruct and extract the latent pulse signal under the current wavelet basis function.
[0011] The steps for implementing step S400 are as follows: S410 extracts characteristic parameters from the original ground motion and candidate pulse signals, calculates combined parameters consisting of the peak ground velocity ratio and energy ratio, and constructs a combined parameter that comprehensively reflects the characteristics of the velocity pulse. The calculation formula is as follows: ; in, These are combined parameters representing the characteristics of the velocity pulse. This is the ratio of peak ground speed; This is the energy ratio; S420 introduces a peak ground velocity variable, calculates a new pulse discrimination index, and performs extreme value verification. The peak ground velocity of the original seismic motion Combined parameters with velocity pulse characteristics Perform joint mapping and calculate the new pulse discrimination index. The calculation formula is as follows: ; in, For new pulse discrimination indicators; The peak ground velocity of the original seismic motion; S430 performs dual verification of the velocity time history of the original ground motion signal and the cumulative energy arrival time of the velocity time history of the candidate pulse signal.
[0012] In step S430, the dual verification step is as follows: The cumulative energy function sequence is constructed by integrating the velocity time history ug(t) of the original ground motion signal and the velocity time history v(t) of the candidate pulse signal with squares respectively. The moment when the cumulative energy function of the extracted raw ground motion signal reaches 17% of the total energy is denoted as . Simultaneously, the moment when the cumulative energy function of the candidate pulse signal reaches 5% of the total energy is extracted and denoted as... ; The execution time comparison logic, when the following conditions are met: When the time condition is met, confirm that the time has elapsed; simultaneously execute the above. The determination result will output the candidate signal that simultaneously meets both verification conditions as the effective velocity pulse component.
[0013] The steps for implementing step S500 are as follows: S510, derive and calculate the improved dimensionless matching evaluation parameters for evaluating candidate pulse signals, the calculation formula is as follows: ; in, To improve the dimensionless matching evaluation parameters; The velocity-time history of the original ground motion signal; The velocity-time history of the extracted pulse signal; For time; S520, to improve dimensionless matching evaluation parameters The minimum value of the objective function is used to lock the optimal wavelet basis function of the current record; S530 uses the zero-point method to calculate and output the pulse period of the target ground motion record.
[0014] This invention provides a near-fault multi-component ground motion pulse identification method based on wavelet basis optimization. It has the following beneficial effects: 1. This invention employs a technical solution of spatial combination and continuous wavelet transform optimization of multi-dimensional orthogonal ground motion components, achieving the technical effect of adaptively locking the direction of the strongest velocity pulse in the entire space. Compared with the existing technology that relies on a single fixed component for conventional identification, this invention overcomes the deficiency of ignoring the seismic wave rupture direction effect, which leads to the easy omission of destructive strong pulses.
[0015] Specifically, this invention overcomes the limitations of traditional methods that rely solely on the fault normal component or a single azimuth within the horizontal plane. The energy of near-fault ground motions is essentially a spatial vector—velocity pulses generated by rupture directionality and slippage effects do not only propagate within the horizontal plane; the vertical component can also carry significant pulse energy. This invention, by introducing a vertical ground motion component, expands the direction identification dimension from a two-dimensional horizontal plane to three-dimensional space, establishing a full-space direction scanning mechanism that includes horizontal azimuth and elevation angles. By calculating the linear combination of the three orthogonal components in any spatial direction and their wavelet coefficient extrema, this invention can accurately pinpoint the direction of the strongest velocity pulse in three-dimensional space, fundamentally eliminating the risk of missed detection due to projection angle deviations or neglecting the vertical component, ensuring that the pulse identification results truly reflect the spatial vector propagation characteristics of ground motion. Compared to existing technologies that rely solely on a single fixed component or two-dimensional horizontal plane scanning, this invention overcomes the shortcomings of neglecting seismic wave rupture direction effects and vertical ground motion components, which easily lead to the omission of highly destructive pulses, achieving full-space, blind-zone-free identification of near-fault velocity pulses.
[0016] 2. This invention employs a technical solution of constructing a multi-type wavelet basis function library and combining it with dimensionless evaluation parameters for inversion screening, achieving the technical effect of dynamic matching and high-purity extraction of velocity pulses for different ground motion records. Compared with the existing technology that forcibly uses a single fixed wavelet basis to extract pulses, this invention solves the shortcomings of pulse extraction peak distortion and excessive reconstruction residuals caused by poor waveform mathematical morphology adaptability.
[0017] This invention constructs a database encompassing 25 wavelet basis functions across five major categories, covering basis functions with different mathematical properties such as orthogonality, bioorthogonality, compact support, and symmetry, providing rich matching templates for diverse pulse waveforms. More importantly, this invention establishes a complete technical chain of "extraction through traversal → dual verification → dimensionless evaluation → adaptive optimization"—for each wavelet basis function, candidate pulse signals are independently extracted. After validity assessment, the optimal wavelet basis function with the lowest possible match to the current seismic motion record waveform is automatically selected, using the minimum value of the improved dimensionless matching evaluation parameter e′e′ as the objective function. This mechanism ensures that regardless of the actual pulse waveform's form (narrowband single peak, broadband multi-peak, asymmetric oscillation, etc.), the system can adaptively select the wavelet basis function with the best mathematical morphology from the database, thereby achieving high-fidelity reconstruction of the pulse signal. Compared with the existing technology that forces the use of a single fixed wavelet basis to extract pulses, this invention solves the shortcomings of pulse extraction peak distortion and reconstruction residuals caused by poor waveform mathematical morphology adaptability, and significantly improves the waveform fidelity, peak accuracy and reconstruction stability of pulse extraction.
[0018] 3. This invention employs a new discrimination index calculated by fusing energy ratios and introduces a dual verification technique using accumulated energy time nodes, achieving the technical effect of accurately removing background clutter and adaptively quantizing and screening effective pulse components. Compared to the existing technology that sets an absolute velocity physical threshold as a rigid cutoff condition, this invention overcomes the shortcomings of its weak anti-interference capability and its susceptibility to misjudgment and missed judgment when facing multi-scale source mechanisms. Attached Figure Description
[0019] Figure 1 This is a schematic diagram of the system architecture of a near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention; Figure 3 This is a distribution map of the earthquake magnitude and fault distance selected for this invention; Figure 4 This is a schematic diagram of the horizontal component velocity pulse; Figure 5 This is a schematic diagram of a three-component velocity pulse; Figure 6 This is a horizontal bidirectional orthogonal acceleration record of the seismic motion RSN77 in Example 1; Figure 7 The optimal wavelet basis result for seismic ground motion RSN77 in Example 1 is shown. Figure 8 The results of RSN77 velocity pulse identification based on the optimal wavelet basis in Example 1; Figure 9This is a comparison chart of velocity pulse identification of ground motion RSN77 under different wavelet bases in Example 1; Figure 10 This is a triaxial orthogonal acceleration record of the seismic motion RSN161 in Example 2; Figure 11 The optimal wavelet basis result of the seismic ground motion RSN161 in Example 2 is shown in the figure. Figure 12 The results of RSN161 velocity pulse identification based on the optimal wavelet basis in Example 2; Figure 13 This is a comparison chart of velocity pulse identification of ground motion RSN161 under different wavelet bases in Example 2; Figure 14 This is a comparison chart of the horizontal velocity pulse recognition results in Example 2; Figure 15 This is a comparison chart of the three-component velocity pulse recognition results in Example 2.
[0020] Among them, 10 is the data preprocessing module; 20 is the pulse direction determination module; 30 is the latent pulse extraction module; 40 is the pulse feature discrimination module; and 50 is the optimal wavelet basis function selection module. Detailed Implementation
[0021] The technical solutions in 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.
[0022] Reference Figure 1 The system architecture of a near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to the present invention is shown, including: a data preprocessing module 10, a pulse direction determination module 20, a potential pulse extraction module 30, a pulse feature discrimination module 40, and an optimal wavelet basis function selection module 50.
[0023] The data preprocessing module 10 is used to collect orthogonal component data of ground motion, establish a wavelet basis function database, and filter the input ground motion signal to eliminate interference terms.
[0024] The pulse direction determination module 20 is used to perform continuous wavelet transform calculations on ground motion signals with different direction combinations, and to lock the direction of the strongest velocity pulse in space by comparing the extreme values of the combined wavelet coefficients.
[0025] The latent pulse extraction module 30 is used to traverse the wavelet basis function database in the direction of the strongest velocity pulse and extract the latent pulse components corresponding to different wavelet basis functions.
[0026] The pulse feature discrimination module 40 is used to quantitatively determine the extracted potential pulse components based on the energy ratio index and pulse energy arrival time characteristics in order to screen effective pulses.
[0027] The optimal wavelet basis function selection module 50 is used to construct waveform matching evaluation parameters, optimize the optimal wavelet basis function from the pulse components that have passed the discrimination, and output the identified pulse feature parameters.
[0028] Reference Figure 2 The diagram below illustrates the method flow of this invention. This invention provides a method for identifying near-fault multi-component ground motion pulses based on wavelet basis optimization, comprising the following steps: S100: Collect orthogonal component data of ground motion to form orthogonal ground motion acceleration time history dataset, perform filtering preprocessing on orthogonal ground motion acceleration time history dataset, and establish a continuous wavelet transform wavelet basis function database. This step is performed by data preprocessing module 10.
[0029] S200, performs multi-angle continuous wavelet transform on the preprocessed orthogonal ground motion acceleration time history dataset, calculates the maximum value of the wavelet coefficients of the linear combination under different spatial directions to determine the direction of the strongest velocity pulse, this step is performed by the pulse direction determination module 20.
[0030] S300 uses the ground motion velocity time history under the direction of the strongest velocity pulse as the initial record, and traverses the wavelet basis function database to extract the potential pulse components under different wavelet basis functions. This step is performed by the potential pulse extraction module 30.
[0031] S400 calculates the energy ratio index of the potential pulse components and extracts the cumulative energy arrival time parameter. It then uses a dual discrimination criterion to determine the effectiveness of the potential pulse components under different wavelet basis functions. This step is performed by the pulse feature discrimination module.
[0032] S500 calculates the improved dimensionless matching evaluation parameters between the potential pulse component velocity time history and the original ground motion velocity time history, determines the optimal wavelet basis function based on the similarity of the signal waveform, and outputs the corresponding optimal strong velocity pulse signal and pulse period. This step is performed by the optimal wavelet basis function selection module 50.
[0033] The technical details of the near-fault multi-component ground motion pulse identification method based on wavelet basis optimization provided by this invention will be elaborated below in conjunction with specific working steps.
[0034] Figure 3This embodiment illustrates the distribution of selected ground motion magnitudes and fault distances. Specific technical implementation details are provided for the data collection, processing, and low-level comparison database construction stages of the near-fault multi-component ground motion pulse identification method based on wavelet basis optimization. Step S100 involves collecting orthogonal ground motion component data to form an orthogonal ground motion acceleration time history dataset. The dataset is then filtered and preprocessed, and a continuous wavelet transform wavelet basis function database is established. The specific implementation method is as follows: S110: Collect orthogonal component data of the target ground motion record and reconstruct it to generate a three-dimensional orthogonal ground motion acceleration time history dataset.
[0035] Data preprocessing module 10 acquires actual ground motion acceleration records in the near-fault region and extracts the first horizontal ground motion components that are perpendicular to each other. Second horizontal ground motion component and vertical seismic components .
[0036] To ensure the synchronization of spatial calculations, the data preprocessing module 10 aligns the acquired three-dimensional independent components along a unified time series, thereby constructing a three-dimensional orthogonal seismic ground acceleration time history dataset containing a complete three-dimensional spatial coordinate system. For the data reading and format parsing of original records from publicly available strong earthquake databases, those skilled in the art can employ conventional methods.
[0037] S120 uses a bandpass filter to filter the three-dimensional orthogonal ground motion acceleration time history dataset to achieve baseline correction and high-frequency noise elimination.
[0038] The data preprocessing module 10 calls a 4th-order Butterworth bandpass filter to filter the acquired acceleration time history records. The passband frequency parameter of the bandpass filter is set from 0.05Hz to 25Hz. By filtering out low-frequency components with frequencies below 0.05Hz, it effectively eliminates the baseline drift error and low-frequency trend terms inherent in the instrument records; at the same time, it filters out high-frequency components with frequencies above 25Hz to avoid high-frequency noise interference caused by environmental factors.
[0039] After filtering, the data preprocessing module 10 performs time-dimensional integration on the corrected acceleration time history dataset. Since the pulse characteristics of the near-fault region are most significant in the integrated velocity time history, the integration transformation operation provides the necessary basis for subsequent extraction of velocity pulse signals, thereby obtaining the original ground motion signal velocity time history that accurately represents ground motion.
[0040] S130, Establish a wavelet basis function database for continuous wavelet transform matching.
[0041] To accommodate the diverse velocity pulse waveform characteristics caused by rupture directionality and slippage effects in near-fault regions, the data preprocessing module 10 loads 25 specific wavelet basis functions into the system's underlying storage to construct a basic comparison sample library. This wavelet basis function database is divided into five waveform feature sets based on differences in mathematical morphology: The first set of wavelet classes includes db2, db3, db4, db5, db6, db7, db8, coif1, coif2, sym2, sym3, and sym4. The 12 wavelet basis functions within this set all possess orthogonality, bioorthogonality, compact support, and approximate symmetry.
[0042] The second set of wavelet classes includes gaus1, gaus2, gaus3, gaus4, gaus5, gaus6, gaus7, and gaus8. The eight wavelet basis functions in this set do not possess compactly supported orthogonality or compactly supported biorthogonality, but they do possess symmetry.
[0043] The third set of wavelet classes includes fk6, fk8, and fk14. The three wavelet basis functions in this set possess orthogonality, bioorthogonality, and compact support, but lack symmetry.
[0044] The fourth type of wavelet basis function is specified as rbio2.4. This wavelet basis does not possess compactly supported orthogonality, but it does possess compactly supported biorthogonality and symmetry.
[0045] The fifth type of wavelet basis function is designated as bior1.3. This wavelet basis does not possess orthogonality and symmetry, but it does possess bioorthogonality and compact support.
[0046] The data preprocessing module 10 encapsulates the aforementioned 25 wavelet basis functions in a wavelet basis function database for subsequent use during multi-angle spatial scanning and pulse component extraction. This database provides tight support reflecting the edge characteristics of ground motion time histories, facilitating the symmetry and orthogonality of ground motion reconstruction accuracy. By considering the characteristics of different wavelet bases, it can select the most suitable wavelet basis function to characterize the chosen ground motion time histories. The database design provides rich waveform matching templates, forming the underlying support conditions for achieving adaptive optimization.
[0047] This embodiment discloses the mathematical derivation process for extending traditional single-component recognition to omnidirectional recognition in two-dimensional and three-dimensional space. To overcome the drawback of single-component recognition, which easily ignores the seismic motion direction effect and thus leads to the omission of strong velocity pulses, the pulse direction determination module 20 performs multi-dimensional continuous wavelet transform processing. It performs multi-angle continuous wavelet transform on the preprocessed orthogonal seismic acceleration time history dataset, calculates the maximum value of the wavelet coefficients of the linear combination under different spatial directions to determine the direction of the strongest velocity pulse. The specific steps of step S200 are as follows: S210, based on the linear combination of bidirectional orthogonal ground motions at any angle in the horizontal plane, derives and calculates the corresponding wavelet coefficients and extreme directions.
[0048] For bidirectional orthogonal ground motion data in the horizontal plane, such as Figure 4 As shown, the horizontal component velocity pulse of this embodiment is illustrated. In the figure, the two horizontal orthogonal components are the first horizontal seismic motion component. Second horizontal ground motion component The pulse direction determination module 20 projects the bidirectional orthogonal ground motion combination onto an arbitrary horizontal rotation angle. The combined ground motion obtained above Represented as: ; (1) In formula 1, This represents the first horizontal seismic motion component. This represents the second horizontal seismic motion component. It is the angle between two orthogonal ground motions in the horizontal plane, i.e., the horizontal rotation angle; For any angle The combined ground motion is obtained by linear combination of horizontal bidirectional orthogonal components.
[0049] Pulse direction determination module 20 pairs of combined ground motion The continuous wavelet transform is calculated using the following formula: (2) In formula 2, Horizontal rotation angle Lower combined ground motion wavelet coefficients; For wavelet scaling parameters, corresponding to frequency or scaling; These are the wavelet shift parameters, corresponding to the time position; The complex conjugate of the mother wavelet function; For time.
[0050] Substituting the aforementioned formula (1) into formula (2) and expanding, we get: ; (3) To simplify the computational structure, the wavelet coefficients of the first horizontal component are defined separately. and the second horizontal component wavelet coefficients : (4) (5) Therefore, the analytical expression for the wavelet coefficients of the horizontal combined ground motion can be simplified to: (6) According to the horizontal rotation angle The continuous variation law, the maximum value of the horizontal combined seismic motion wavelet coefficient is: (7) The angle corresponding to this extreme value is the direction angle of the strongest velocity pulse. The calculation formula is: (8) In the actual numerical solution, the pulse direction determination module 20 traverses the wavelet translation parameters. and wavelet scaling parameters Extract the term with the largest absolute value of the combined wavelet coefficients in formula (7) and assign its corresponding angle. The direction of the strongest velocity pulse has been determined.
[0051] S220, the pulse direction determination module 20 extends the direction recognition dimension from a two-dimensional horizontal plane to a three-dimensional space, and calculates the wavelet coefficients of the linear combination of three-dimensional orthogonal ground motion under arbitrary spatial azimuth and pitch angles.
[0052] Introducing vertical ground motion components ,like Figure 5 As shown, this embodiment illustrates the three-component velocity pulse and the combined ground motion of the three-dimensional orthogonal ground motion in any spatial direction. Characterized as: (9) in, The angle between two orthogonal ground motions in a horizontal plane in three-dimensional space; The pitch angle is the angle between the spatial direction of the earthquake and the horizontal plane. This represents the first horizontal seismic motion component. This represents the second horizontal seismic motion component. This represents the vertical seismic motion component. Pulse direction determination module 20 for this spatial combined ground motion Perform continuous wavelet transform: (10) in, These are the wavelet coefficients of the three-component combined ground motion.
[0053] Substituting formula (9) into formula (10), we get: (11) in, The first wavelet coefficient in the horizontal direction, The wavelet coefficients are the second component in the horizontal direction. These are the wavelet coefficients for the vertical ground motion component. In specific low-level signal processing, a discretization approximation method can be used to achieve numerical integration.
[0054] S230 extracts the extreme values of the combined wavelet coefficients and locks the direction of the strongest velocity pulse in three-dimensional space.
[0055] The maximum value of the wavelet coefficient of the three-component combined ground motion is: (12) The formulas for calculating the coordinates of the direction of the strongest velocity pulse in space corresponding to this maximum value are as follows: (13) (14) The pulse direction determination module 20 determines the spatial angle obtained by solving the above formulas (13) and (14) as the direction of the strongest velocity pulse in the current ground motion record. This multidimensional omnidirectional scanning mechanism effectively avoids the omissions caused by extracting features only on a single physical orthogonal axis.
[0056] Based on this, the present invention provides specific technical implementation details for the pulse separation process of unidirectional waveform data under different wavelet transform scales. S300, using the ground motion velocity time history under the direction of the strongest velocity pulse as the initial record, the wavelet basis function database is traversed to extract potential pulse components under different wavelet basis functions. This step is executed by the potential pulse extraction module 30, and specifically includes the following implementation process: S310, extract the seismic velocity time history in the direction of the strongest velocity pulse calculated by formula (13) and formula (14), and construct it as the initial input record.
[0057] The latent pulse extraction module 30 receives the coordinates of the strongest velocity pulse direction locked in the previous step, and performs spatial vector projection of the three-dimensional orthogonal or horizontal two-dimensional orthogonal ground motion data after filtering and preprocessing in step S120 along this direction to obtain one-dimensional waveform data, i.e., the velocity time history of the original ground motion signal. Because the strongest seismic energy is concentrated in this projection direction, the original seismic signal velocity time history... The low-frequency high-energy wave characteristics induced by the near-fault rupture directionality effect were effectively characterized and set as the initial input record for pulse separation to effectively eliminate interference from non-dominant direction clutter signals.
[0058] S320: Traverse the wavelet basis function database constructed in step S100, and perform continuous wavelet transform and pulse separation on the initial input records.
[0059] The latent pulse extraction module 30 sequentially extracts 25 wavelet basis functions from five categories in the database. In each iteration, it locks a specific wavelet basis function as the kernel function to process the velocity time history of the original ground motion signal. Perform continuous wavelet transform throughout the entire time domain. By expanding the one-dimensional time-domain signal into a two-dimensional domain of time and scale through wavelet transform, the local pulse energy hidden in the complex background waveform is transformed into a significant response distribution.
[0060] S330, reconstruct and extract the latent pulse signal under the current wavelet basis function.
[0061] The latent pulse extraction module 30 calculates the wavelet coefficient distribution of the current wavelet basis function across the entire frequency band and time period, locating the maxima in the wavelet coefficient matrix. The dominant frequency band and time window corresponding to this maxima characterize the location of the velocity pulse component with the highest energy concentration in the current signal. The maximum wavelet coefficient is then obtained. and its corresponding wavelet scaling parameters With wavelet translation parameters Based on this, the velocity time history of the candidate pulse signal is reconstructed. .
[0062] Using the current mother wavelet function By combining the extreme point parameters, inversion reconstruction is performed to extract the pulse signal velocity time history. The calculation expression is: (15) in, To reconstruct the velocity time history of the extracted candidate pulse signal; This is the maximum value in the wavelet coefficient matrix, i.e., the term with the largest absolute value; For the wavelet scaling parameters corresponding to the maxima, in this embodiment, the energy normalization factor is used for inverse reconstruction. ; The wavelet shift parameter corresponding to the maximum value; For the specific mother wavelet function currently being called, if the mother wavelet function is a complex-valued function, then its real part is taken as the velocity time history of the candidate pulse signal during reconstruction; For time. This inverse reconstruction process is equivalent to stripping away extreme energy to separate candidate pulse signals from a complex seismic background. Near-fault velocity pulses typically exhibit a single dominant energy cluster in the time-frequency domain. Therefore, single-point reconstruction using the maximum wavelet coefficients can effectively extract the main pulse component while ignoring sidelobe details with extremely low energy proportions.
[0063] The latent pulse extraction module 30 repeats the above projection and reconstruction process for each wavelet basis function in the database, and finally outputs 25 independently extracted candidate pulse signal velocity time histories. Subsequent signal recovery and reverse reconstruction calculations can be implemented using conventional discrete inverse wavelet transform filter banks or integral reconstruction techniques.
[0064] After pulse extraction in this step, the process proceeds to step S400, where the energy ratio index of the potential pulse components is calculated, and the cumulative energy arrival time parameter is extracted. A dual-criteria is used to determine the validity of the potential pulse components under different wavelet basis functions. The quantification criteria for pulse features after removing fixed threshold limitations are then disclosed. The detailed steps are as follows: S410, the pulse feature discrimination module 40 extracts the feature parameters of the original ground motion and the candidate pulse signal obtained in step S330, and calculates the combined parameters composed of the peak ground velocity ratio and the energy ratio.
[0065] The actual near-fault velocity pulses account for a high proportion of energy and peak amplitude in the overall ground motion. The pulse feature discrimination module 40 extracts the velocity time history of the original ground motion signal. and the candidate pulse signal velocity time history extracted in step S300 By obtaining the global maximum term of the absolute value of the velocity time history, the peak ground velocity of the original ground motion and the peak ground velocity of the candidate pulse signal are calculated separately, thus obtaining the peak ground velocity ratio. Similarly, global square integration is performed on the two velocity time histories to calculate the original ground motion energy and the candidate pulse signal energy, respectively. The ratio of the candidate pulse signal energy to the original ground motion energy is used as the energy ratio. .
[0066] Based on this, the pulse feature discrimination module 40 constructs a combination of parameters that comprehensively reflect the velocity pulse features. The calculation formula is as follows: (16) in, These are combined parameters representing the characteristics of the velocity pulse. This is the ratio of peak ground speed; This refers to the energy ratio. In constructing the aforementioned combined parameter PC, this invention fully considers the complementarity and correlation between the peak ground velocity ratio and the energy ratio in reflecting pulse intensity. Statistical analysis of a large number of strong earthquake records (covering different magnitudes, distances, and site conditions) reveals that... and Although both indicators can effectively distinguish between pulsed and non-pulsed ground motions, they exhibit a significant positive correlation and high information redundancy. To eliminate the impact of multicollinearity on subsequent discrimination accuracy, this invention employs principal component analysis to linearly reduce the dimensionality of the two indicators and fuse them. Specifically, based on a sample set containing thousands of real ground motion records, two ratio parameters for each record are calculated, and the first principal component direction of its covariance matrix is extracted. The loading vector of this principal component is the weight coefficient of each indicator. After optimization calculation, the optimal weighting coefficients of PGVratio and Energyratio are determined to be 0.63 and 0.777, respectively, thereby constructing the combined parameter PC shown in formula (16). This combined parameter can compress the two-dimensional feature space to one dimension while retaining most of the discrimination information of the original two indicators (cumulative variance contribution rate exceeds 90%), effectively reducing the process complexity of the subsequent classification model and improving the robustness of the discrimination boundary.
[0067] S420, the pulse feature discrimination module 40 introduces the peak ground velocity variable, calculates a new pulse discrimination index, and performs extreme value verification.
[0068] To overcome the omission defect caused by using a single fixed velocity threshold, the pulse feature discrimination module 40 uses the peak ground velocity of the original ground motion. Combined parameters with velocity pulse characteristics Perform joint mapping and calculate the new pulse discrimination index. The calculation formula is as follows: (17) in, For new pulse discrimination indicators; This represents the peak ground velocity of the original ground motion.
[0069] In the process of constructing the above formula (17), this invention fully considers the constraint of the multi-scale nature of ground motion intensity on the fixed threshold discrimination. Specifically, this invention is based on a large number of ground motion records with labeled pulse attributes in the NGA-West2 strong earthquake database (covering a wide range of magnitudes from 3.0 to 7.9, fault distances from 0.05 to 1533 km, and site shear wave velocities from 94 to 2100 m / s). It uses a support vector machine (SVM) classifier, with peak ground velocity ratio, energy ratio, and original ground motion peak ground velocity as input features, and performs nonlinear classification boundary fitting through a second-order polynomial kernel function. After iterative optimization, the new pulse discrimination index PI_new, as shown in formula (17), is finally obtained. This index internalizes the influence of PGV into the discrimination formula, avoids the drawback of artificially setting absolute velocity thresholds, and realizes adaptive discrimination of weak and strong earthquake records. Classification boundary ( The determination of (= 0) is based on the principle of maximizing the consistency between manual and automatic discrimination in the training set. Cross-validation shows that its overall classification accuracy is better than the traditional single fixed threshold method.
[0070] After obtaining this metric, Compare with the zero-point threshold. When If the current candidate signal meets the basic pulse properties in terms of energy concentration and amplitude intensity, it is allowed to proceed to the next level of verification; otherwise, it is discarded. Because The formula has already taken this into account. This step directly eliminates the impact of traditional screening pulses. The absolute threshold setting effectively improves the adaptive recognition capability for multi-scale pulses.
[0071] S430 performs dual verification of the velocity time history of the original ground motion signal and the cumulative energy arrival time of the velocity time history of the candidate pulse signal.
[0072] Due to the directional effects and slippage effects near the fault front, the energy of destructive velocity pulses typically concentrates rapidly in the early stages of seismic wave propagation. To eliminate false low-frequency signals extracted from the wake, the pulse feature discrimination module 40 sets arrival time discrimination conditions in the time dimension. The threshold for this condition is selected based on statistical analysis of a large number of actual ground motion records: for non-pulse ground motions, the moment when the cumulative square velocity reaches 17% of the total energy corresponds precisely to the moment when the ground motion intensity (usually indicated by a sudden increase in acceleration) fully erupts, meaning that the main energy envelope of the ground motion has begun to be released at this time. If a true pulse exists, the energy of the pulse signal should have been significantly concentrated before this moment. Conversely, for extracted candidate pulse signals, if they are valid pulses, their cumulative energy should rapidly reach 5% of the total energy, meaning that the main energy of the pulse is concentrated in the very early stages. Therefore, this invention defines the moment when the original ground motion reaches 17% of the total energy as... The moment when the candidate pulse signal reaches 5% of the total energy is The execution time comparison logic, when the condition is met... If the pulse signal appears earlier than or simultaneously with the energy burst of the background ground motion, it conforms to the physical characteristics of a true near-fault pulse, confirming that the time has met the discrimination criteria. Conversely, if the pulse energy lags behind the energy burst of the background ground motion, it is highly likely to be a false pulse caused by a wake or clutter. This should be considered in conjunction with the preceding steps. A result greater than 0 will result in the output of candidate signals that simultaneously meet both verification conditions as valid velocity pulse components. This principle can be summarized as follows: Velocity time history of the original ground motion signal By integrating the velocity time history v(t) of the candidate pulse signal with the square, a sequence of its cumulative energy functions is constructed. For non-pulse-type ground motions, the moment when the energy reaches 17% corresponds precisely to the moment when the ground motion intensity (usually indicated by a sudden increase in acceleration) fully erupts. The moment when the cumulative energy function from the original ground motion signal reaches 17% of the total energy is extracted and denoted as... Simultaneously, the moment when the cumulative energy function of the candidate pulse signal reaches 5% of the total energy is extracted and denoted as... .
[0073] The execution time comparison logic, when the following conditions are met: When the time elapsed reaches the discrimination condition, it is confirmed that the pulse has reached the discrimination condition. If the condition is not met, it is judged as a false pulse (such as low-frequency clutter in the wake) and is discarded. Combined with the new pulse discrimination index in the previous formula (17) If the time-filtered result is passed, and both of these conditions are met, then the candidate signal output is an effective velocity pulse component. The calculation of the cumulative energy integral curve of the underlying signal can be achieved using a conventional discrete-time series summation algorithm.
[0074] After valid pulses are selected through dual verification, step S500 calculates the improved dimensionless matching evaluation parameters between the potential pulse component velocity time history and the original ground motion velocity time history. The optimal wavelet basis function is determined based on the similarity of the signal waveforms, and the corresponding optimal strong velocity pulse signal and pulse period are output. Specific technical implementation details are provided for the adaptive optimization algorithm mechanism that addresses the problem of poor matching degree of fixed wavelet basis functions, specifically implemented through the following steps: S510, derive and calculate the improved dimensionless matching evaluation parameters for evaluating candidate pulse signals.
[0075] To quantitatively evaluate the degree of agreement between the pulse waveforms extracted by different wavelet basis functions and the actual near-fault pulse morphology, the optimal wavelet basis function selection module 50 is based on the original ground motion velocity time history. Evaluation parameters are constructed using the extracted pulse signal velocity time history v(t). This improves the dimensionless matching evaluation parameters. The calculation formula is: ;(Formula 18) in, To improve the dimensionless matching evaluation parameters; The velocity-time history of the original ground motion signal; The velocity-time history of the extracted pulse signal; For time.
[0076] This formula, in the form of dimensionless physical quantities, eliminates the absolute numerical interference caused by different earthquake magnitudes and can objectively reflect the waveform matching degree of pulses extracted by different wavelet basis functions.
[0077] S520, to improve dimensionless matching evaluation parameters The minimum value is the objective function, and the optimal wavelet basis function for the current record is locked.
[0078] For all valid candidate pulse signals selected in the previous steps, if no valid candidate pulse signal exists, the current ground motion record is determined to be a non-pulse ground motion, and the identification process ends; if valid candidate pulse signals exist, the corresponding improved dimensionless matching evaluation parameters are calculated one by one. Subsequently, in all the calculated results The optimal velocity pulse signal is the specific candidate pulse signal that corresponds to the minimum value in the numerical set.
[0079] because The smaller the value, the better the shape of the corresponding wavelet basis function matches the actual velocity pulse waveform. Therefore, the optimal wavelet basis function selection module 50 locks the mother wavelet function corresponding to the minimum value as the optimal wavelet basis function of the current ground motion record, and confirms the waveform reconstructed by this function as the final optimal velocity pulse signal. This adaptive mechanism effectively solves the technical problem of poor waveform matching when a single fixed wavelet basis is used to deal with complex and diverse fault rupture mechanisms.
[0080] S530 uses the zero-point method to calculate and output the pulse period of the target ground motion record.
[0081] After acquiring the optimal velocity pulse signal, the optimal wavelet basis function selection module 50 identifies the zero-crossing point distribution of the optimal velocity pulse signal waveform on the time axis. This is achieved by locating the initial zero-crossing moment of the main pulse wave packet. and the time of termination of zero crossing The time difference is directly converted and defined as the pulse period. The calculation formula is: ;(Formula 19) in, The pulse period recorded for the target ground motion; The zero-crossing point corresponding to the end position of the main pulse packet; It is the initial zero-crossing time corresponding to the starting position of the main pulse wave packet.
[0082] After obtaining this parameter, the optimal wavelet basis function selection module 50 will select the optimal wavelet basis function name, the optimal velocity pulse signal time history, the strongest velocity pulse direction coordinates, and the pulse period. Output the data. Zero-crossing detection of the underlying waveform can be achieved using conventional sign change detection or linear interpolation approximation methods.
[0083] To further clarify the collaborative working process of the technical solution described in this invention, a specific working scenario example will be used below. The data used in this embodiment for verification comes from the NGA-West2 database of the Pacific Earthquake Engineering Research Center (PEER).
[0084] For the validation dataset, set the following filtering criteria: the range of earthquake magnitudes includes... The magnitude of the fault was within the vicinity, and the closest distance between the site and the fault rupture surface was no more than 60 km. After screening, a total of 2652 sets of triaxial orthogonal ground motion records were obtained as the analysis dataset.
[0085] Here are two verification examples: I. Verification of bidirectional orthogonal ground motion pulse identification in the horizontal plane: The ground motion RSN77 recorded at the Pacoima Dam station during the 1971 San Fernando earthquake was selected as the subject of analysis. The magnitude of this earthquake event was... Distance from the epicenter The peak values of the horizontal bidirectional orthogonal accelerations were 1.22g and 1.24g, respectively, at km.
[0086] Figure 6 The horizontal bidirectional orthogonal acceleration records of the seismic ground motion RSN77 are shown, where (a) is the acceleration record in the NS direction and (b) is the acceleration record in the EW direction. The time histories of these horizontal bidirectional orthogonal acceleration records are input into the system and preprocessed using a 4th-order Butterworth bandpass filter with a passband range of 0.05–25 Hz to obtain the corrected velocity time histories. The system performs a two-dimensional directional scan, calculating the wavelet coefficients of the linear combination at different angles. The direction corresponding to the maximum value of the wavelet coefficients is taken as the direction of the strongest velocity pulse of RSN77.
[0087] Figure 7 To obtain the optimal wavelet basis results for the seismic ground motion RSN77, the system systematically extracted the pulse signal by traversing 25 wavelet basis functions along the direction of the strongest pulse, and calculated the improved dimensionless parameters for each. The calculation results show that, The minimum value is 0.53, and the corresponding wavelet basis function is coif2. Based on the parameters... Based on the evaluation rules, the system selects coif2 as the optimal wavelet basis for this record.
[0088] Figure 8 The results of velocity pulse identification for ground motion RSN77 based on the optimal wavelet basis are shown in (a) the velocity time history comparison between the original ground motion and the pulse signal, (b) the velocity time history and energy arrival time of the pulse signal, and (c) the cumulative squared velocity time history comparison. After extracting the velocity pulse signal using coif2, the system calculates the new pulse discrimination index. Meanwhile, the moment when the accumulated energy of the pulse signal reaches 5% of the total energy is later than the moment when the accumulated energy of the original ground motion reaches 17% of the total energy, thus satisfying the arrival time verification condition. The system determines that RSN77 is a near-fault pulse-type ground motion and outputs its pulse direction and pulse period using the zero-point method. .
[0089] Reference Figure 9 The image shows a comparison of velocity pulse identification for ground motion RSN77 under different wavelet bases in this embodiment. (a) shows the pulse signal extracted by the optimal wavelet base coif2, (b) shows the pulse signal extracted by db4, (c) shows the pulse signal extracted by sym4, and (d) shows a comparison of the residual signals after pulse extraction by different wavelet bases. The extraction results of the optimal wavelet base coif2 are compared with those of db4 and sym4. Waveform data shows that the peak value of the pulse extracted by db4 is lower. In contrast, the waveform extracted by coif2 has a higher degree of agreement with the original signal, and the residual signal is more stable. This result indicates that using a specific wavelet base will cause identification bias. The parameter optimization mechanism is reasonable.
[0090] Verification of two- and three-component seismic pulse identification: The ground motion RSN161 recorded at the Brawley Airport station during the 1979 Imperial Valley earthquake was selected as the subject of analysis. The magnitude of this earthquake event was... Distance from the epicenter km. The peak values of the three orthogonal accelerations were 0.16g, 0.19g, and 0.15g, respectively.
[0091] Reference Figure 10 The data consists of three orthogonal acceleration records for the ground motion RSN161, where (a) is the acceleration record in the NS direction, (b) is the acceleration record in the EW direction, and (c) is the acceleration record in the UD direction. The system inputs the three orthogonal acceleration time histories and performs data preprocessing using a 4th-order Butterworth bandpass filter. Subsequently, the system performs a spatial combined scan of the three-component ground motion and calculates the wavelet coefficients at different azimuth and elevation angles. The system extracts the maximum value of the wavelet coefficients and determines the corresponding spatial direction as the direction of the strongest velocity pulse.
[0092] Reference Figure 11 The image shows the optimal wavelet basis results for the ground motion RSN161. In the locked pulse direction, the system traverses the wavelet basis function library and calculates the improved dimensionless parameters for each wavelet basis function. The calculation results show that the minimum The value is 0.44, corresponding to the optimal wavelet basis function rbio2.4. This result differs from the wavelet basis function in Example 1, indicating that the wavelet basis functions for different ground motion records need to be adaptively matched according to the actual waveform.
[0093] Reference Figure 12 The results of RSN161 velocity pulse identification based on the optimal wavelet basis are shown in (a) for the velocity time history comparison between the original ground motion and the pulse signal, (b) for the velocity time history and energy arrival time of the pulse signal, and (c) for the cumulative squared velocity time history comparison. After extracting the velocity pulse signal based on rbio2.4, the system calculates the new pulse discrimination index. Simultaneously, the arrival time of the signal's pulse energy meets the discrimination requirements. The system determines that a velocity pulse exists in the record and calculates the pulse period using the zero-point method. .
[0094] Reference Figure 13 The figures show a comparison of RSN161 velocity pulse identification based on different wavelet bases. (a) shows the pulse signal extracted by the optimal wavelet base rbio2.4, (b) by db4, (c) by sym4, and (d) shows a comparison of the residual signals after pulse extraction using different wavelet bases. Comparison of the extraction results from rbio2.4, db4, and sym4 shows that the peak value of the pulse extracted by db4 is lower, with a peak value difference of approximately 20 cm / s. The pulse waveform extracted by rbio2.4 shows good agreement with the original ground motion, and the residual value is smaller. This step verifies the effectiveness of this method in identifying three-component ground motion pulses.
[0095] Overall recognition results and statistical analysis: The algorithm of this invention was applied to the aforementioned dataset of 2652 sets of ground motion records and compared with the existing single-component fixed wavelet basis method (Shahi method).
[0096] Reference Figure 14 In the horizontal direction, the method of this invention identified 253 horizontal pulse-type ground motions, while the Shahi method identified 212. This invention identified 41 more records than the Shahi method.
[0097] Reference Figure 15 In the three orthogonal components, the method of this invention identified 269 three-component pulse-type ground motions, while the extended Shahi method identified 224. This invention identifies 45 more records than the extended Shahi method. Furthermore, compared to the horizontal identification result of this invention (253 records), the three-component identification process further identifies 16 more pulse-type ground motion records.
[0098] Statistical data show that the method provided by this invention can identify velocity pulses in three-component ground motions and supplement existing pulse-type ground motion databases.
[0099] Although embodiments of the invention have been shown and described, 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. A near-fault pulse identification method based on wavelet basis optimization, characterized in that, Includes the following steps: S100: Collect orthogonal component data of ground motion to form orthogonal ground motion acceleration time history dataset, perform filtering preprocessing on orthogonal ground motion acceleration time history dataset, and establish a continuous wavelet transform wavelet basis function database. S200 performs multi-angle continuous wavelet transform on the preprocessed orthogonal ground motion acceleration time history dataset, calculates the maximum value of the wavelet coefficients of the linear combination under different spatial directions to determine the direction of the strongest velocity pulse; S300 uses the ground motion velocity time history under the direction of the strongest velocity pulse as the initial record, and traverses the wavelet basis function database to extract the potential pulse components under different wavelet basis functions. S400 calculates the energy ratio index of potential pulse components and extracts the cumulative energy arrival time parameter. It then uses a dual discrimination criterion to determine the effectiveness of potential pulse components under different wavelet basis functions. S500 calculates the improved dimensionless matching evaluation parameters between the potential pulse component velocity time history and the original ground motion velocity time history, determines the optimal wavelet basis function based on the similarity of the signal waveform, and outputs the corresponding optimal strong velocity pulse signal and pulse period.
2. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 1, characterized in that, The steps for implementing step S100 are as follows: S110: Collect orthogonal component data of the target ground motion record and reconstruct it to generate a three-dimensional orthogonal ground motion acceleration time history dataset; S120 performs filtering on the three-dimensional orthogonal ground motion acceleration time history dataset to achieve baseline correction and high-frequency noise elimination; S130, Establish a wavelet basis function database for continuous wavelet transform matching.
3. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 2, characterized in that, In step S130, the wavelet basis function database is divided into the following five categories: The first set of wavelets includes db2, db3, db4, db5, db6, db7, db8, coif1, coif2, sym2, sym3, and sym4; The second set of wavelets includes gaus1, gaus2, gaus3, gaus4, gaus5, gaus6, gaus7 and gaus8; The third set of wavelets includes fk6, fk8 and fk14; The fourth type of wavelet basis function is specified as rbio2.4; The fifth type of wavelet basis function is specified as bior1.
3.
4. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 1, characterized in that, The steps for implementing step S200 are as follows: S210, based on the linear combination of bidirectional orthogonal ground motions at any angle in the horizontal plane, derive and calculate the corresponding wavelet coefficients and extreme directions; The analytical expression for the wavelet coefficients of horizontal combined ground motion is: ;in, The first wavelet coefficient in the horizontal direction, The wavelet coefficients of the second horizontal component ; For wavelet scaling parameters, corresponding to frequency or scaling; These are the wavelet shift parameters, corresponding to the time position; For time; It is the angle between two orthogonal ground motions in the horizontal plane, i.e., the horizontal rotation angle; The maximum value of the horizontal combined ground motion wavelet coefficients is: ； The direction angle of the strongest velocity pulse corresponding to this extreme value The calculation formula is: ； S220 extends the direction recognition dimension from a two-dimensional horizontal plane to a three-dimensional space, and calculates the wavelet coefficients of the linear combination of three-dimensional orthogonal ground motion under arbitrary spatial azimuth and pitch angles. ； in, The first wavelet coefficient in the horizontal direction, The wavelet coefficients are the second component in the horizontal direction. These are the wavelet coefficients for the vertical ground motion component. In specific low-level signal processing, a discretization approximation method can be used to achieve numerical integration. The angle between two orthogonal ground motions in the horizontal plane; The pitch angle; S230 extracts the extreme values of the combined wavelet coefficients and locks the direction of the strongest velocity pulse in three-dimensional space; The maximum value of the wavelet coefficient of the three-component combined ground motion is: ； The formulas for calculating the coordinates of the direction of the strongest velocity pulse in space corresponding to this maximum value are as follows: ； 。 5. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 1, characterized in that, The steps for implementing step S300 are as follows: S310, Extract the seismic velocity time history in the direction of the strongest velocity pulse calculated in step S200, and construct it as the initial input record; S320, Traverse the wavelet basis function database established in step S100, and perform continuous wavelet transform and pulse separation on the initial input record; S330, reconstruct and extract the latent pulse signal under the current wavelet basis function.
6. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 1, characterized in that, The steps for implementing step S400 are as follows: S410 extracts characteristic parameters from the original ground motion and candidate pulse signals, calculates combined parameters consisting of the peak ground velocity ratio and energy ratio, and constructs a combined parameter that comprehensively reflects the characteristics of the velocity pulse. The calculation formula is as follows: ； in, These are combined parameters representing the characteristics of the velocity pulse. This is the ratio of peak ground speed; This is the energy ratio; S420 introduces a peak ground velocity variable, calculates a new pulse discrimination index, and performs extreme value verification. The peak ground velocity of the original seismic motion Combined parameters with velocity pulse characteristics Perform joint mapping and calculate the new pulse discrimination index. The calculation formula is as follows: ； in, For new pulse discrimination indicators; The peak ground velocity of the original seismic motion; S430 performs dual verification of the velocity time history of the original ground motion signal and the cumulative energy arrival time of the velocity time history of the candidate pulse signal.
7. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 6, characterized in that, In step S430, the dual verification step is as follows: The cumulative energy function sequence is constructed by integrating the velocity time history ug(t) of the original ground motion signal and the velocity time history v(t) of the candidate pulse signal with squares respectively. The moment when the cumulative energy function of the extracted raw ground motion signal reaches 17% of the total energy is denoted as [missing information]. Simultaneously, the moment when the cumulative energy function of the candidate pulse signal reaches 5% of the total energy is extracted and denoted as... ; The execution time comparison logic, when the following conditions are met: At that time, confirm that the time has reached the discrimination condition; and simultaneously execute the above. The determination result will output the candidate signal that simultaneously meets both verification conditions as the effective velocity pulse component.
8. The near-fault multi-component ground motion pulse identification method based on wavelet basis optimization according to claim 1, characterized in that, The steps for implementing step S500 are as follows: S510, derive and calculate the improved dimensionless matching evaluation parameters for evaluating candidate pulse signals, the calculation formula is as follows: ； in, To improve the dimensionless matching evaluation parameters; The velocity-time history of the original ground motion signal; The velocity-time history of the extracted pulse signal; For time; S520, to improve dimensionless matching evaluation parameters The minimum value of the objective function is used to lock the optimal wavelet basis function of the current record; S530 uses the zero-point method to calculate and output the pulse period of the target ground motion record.