A method for HVSR inversion based on multi-peak forest optimization algorithm
Patent Information
- Application Number
- CN202311176559.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-13
- Publication Date
- 2026-11-03
- Estimated Expiration
- 2043-09-13
AI Technical Summary
Existing HVSR inversion methods are prone to getting trapped in local minima, leading to multiple solutions and making it difficult to obtain the global optimal solution.
The Multi-Mounted Forest Optimization Algorithm (MMFOA) is used for HVSR inversion. The main population is divided into sub-populations through population clustering. By utilizing its global optimization ability, local minima are avoided and the global optimal solution is obtained.
It effectively avoids local minima in the optimization process, obtains the global optimal solution, and improves the accuracy and stability of the inversion results, especially in accurately extracting formation parameters under noisy data conditions.
Smart Images
Figure CN117471528B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of urban micro-motion detection technology, specifically involving an HVSR inversion method based on the Multi-peak Forest Optimization Algorithm (MMFOA), which is applicable to the study of multi-parameter inversion problems in urban spatial detection. Background Technology
[0002] Shallow surface geological structures are closely related to human survival and development, making their study an important research topic. Geophysical exploration utilizes geophysical methods combined with signals acquired by detection equipment to infer the physical properties of underground structures, thereby addressing issues such as the stability of building foundations, the detection of underground caves, ground fissures, landslides, and urban spatial exploration. Single-station micromotion measurement methods have been widely used in recent years due to their practicality and low cost. The most commonly used method for single-station micromotion data is the HVSR (Horizontal-to-Vertical Spectral Ratio), which is the ratio of the Fourier amplitude spectra of the horizontal components (east-west component (E), north-south component (N)) and vertical components (Z) of the micromotion. This ratio allows for the inversion of shallow velocity profiles based on the dynamic characteristics of single-station micromotion signals. It offers advantages such as being noise-free, having flexible construction methods, high efficiency, and minimal impact on urban traffic and residents' lives, making it a green and environmentally friendly new detection technology. The HVSR method is currently widely used in various fields, such as seismology, exploration geophysics, and building monitoring.
[0003] Inversion of shallow velocity profiles based on HVSR (Hyperwave Velocity Spectroscopy) has been a popular research area in the past decade or so. Currently, methods for inverting HVSR curves mainly rely on nonlinear conventional optimization algorithms, including Monte Carlo algorithms, genetic algorithms, and particle swarm optimization. However, since HVSR inversion is a curve fitting process, different parameters (thickness, density, P-wave velocity, S-wave velocity, and damping) have different effects on the shape of the synthesized HVSR curve. This leads to multiple solutions in the HVSR inversion results, and conventional optimization algorithms are prone to getting trapped in local minima. Therefore, effectively avoiding getting trapped in local minima during the optimization process is a problem that urgently needs to be solved. Summary of the Invention
[0004] The purpose of this invention is to provide a method for shallow surface shear wave velocity inversion based on the multi-peak forest optimization algorithm. By utilizing its global optimization capability, it finds all local and global solutions, obtains the global optimal solution model, and achieves global optimization, thereby solving the problems of easily getting trapped in local minima and multi-peaks in the inversion process of traditional HVSR methods.
[0005] The objective of this invention is achieved through the following technical solution:
[0006] A method for HVSR inversion based on a multi-peak forest optimization algorithm includes the following steps:
[0007] Step 1: At the measurement point, use a three-component detector to collect micro-motion data and extract the east-west component, north-south component, and vertical component data of the micro-motion signal from the three-component detector.
[0008] Step 2: Ratio the average value of the characteristic function of the micro-motion signal within the short-time window to the average value of the characteristic function of the micro-motion signal within the long-time window at each measurement point to eliminate short-time interference. Then, perform Fourier transform on the processed three-component micro-motion data to obtain the Fourier amplitude spectrum, and smooth the amplitude spectrum using the Konno & Ohmachi window. Smoothing is performed by convolving the window function with the Fourier amplitude spectrum to obtain a reliable frequency domain HVSR curve and perform reliability judgment.
[0009] Step 3: Use the MMFOA algorithm to invert the HVSR curve, fit the actual HVSR curve and the theoretical HVSR curve to the greatest extent, and obtain the shear wave velocity structure under each region.
[0010] Step 4: Infer and interpret the anomalous geological bodies based on the abnormal characteristics of the shear wave velocity profile.
[0011] Furthermore, in step 1, when collecting micro-motion data, the acquisition frequency range is approximately 0.2–25 Hz, while ensuring the coupling between the sensor and the site.
[0012] Furthermore, in step 2, the average value of the characteristic function of the micro-motion signal within the short-time window at each measuring point is compared with the average value of the characteristic function of the micro-motion signal within the long-time window. The specific formula is as follows:
[0013]
[0014] Where i is the sampling time, ns is the length of the short time window, nl is the length of the long time window, and f(j) is the characteristic function value of the microseismic signal at time i.
[0015] Furthermore, in step 2, the reliability criteria for the HVSR curve are as follows:
[0016] 21. Peak frequency f0 and time window length l w Satisfy: f0 > 10 / l w ;
[0017] 22. Number of time windows N c Satisfy: n c (f0) > 200;
[0018] 23. If f0 > 0.5 Hz, then for 0.5f0 < f < 2f0, the standard deviation of HVSR satisfies σ a (f) < 2;
[0019] If f0 < 0.5 Hz, then for 0.5f0 < f < 2f0, the standard deviation of HVSR satisfies σ. a (f) < 3;
[0020] Furthermore, in step 3, the MMFOA algorithm inverts the HVSR curve, specifically including the following steps:
[0021] 31. Load the observed HVSR data; use random population initialization to obtain the initial model, and calculate the reciprocal of the objective function value of the initial population;
[0022] 32. Perform five key iterative steps of MMFOA: forest initialization, local seeding, age-based population constraint, population clustering, and global seeding;
[0023] 33. By fitting the actual HVSR curves and the theoretical HVSR curves to the greatest extent possible, an optimal set of formation model solutions can be found.
[0024] Furthermore, in step 3, the value of the objective function is a function of the underground parameters, and the formula is:
[0025]
[0026] Where I is the number of samples, (HVSR) m For actual measurement of HVSR, (HVSR) s The synthesized HVSR curve is shown at frequency f. The synthesized HVSR is approximated by the equivalent linear method based on the viscoelastic Kelvin-Voigt model. The core of the optimal solution model is to find the global minimum of the objective function.
[0027] Furthermore, the five key iterative steps of MMFOA in step 3 are as follows:
[0028] a. Forest initialization: Initialize using a random population to obtain an array of length 1*(N+1), including N variables and tree age;
[0029] b. Local seeding: Applied to target trees with an age of 0. During this stage, some trees with an age of 0 are added near the target tree, and the target tree will become 1 year older.
[0030] c. Age-based population restriction: In this stage, old trees that have exceeded their maximum lifespan will be removed from the forest and added to the candidate population;
[0031] d. Population clustering: The Basic Sequential Algorithm Scheme (BSAS) is used to search the clustering tree in MMFOA. BSAS will access the data only once, uses a niche mechanism to handle this problem, and creates an adaptive system.
[0032] e. Global seeding operation: Select a percentage of the candidate population.
[0033] Compared with the prior art, the beneficial effects of the present invention are:
[0034] This invention presents an HVSR inversion method based on a multi-peak forest optimization algorithm. It introduces population clustering, that is, it uses an efficient clustering algorithm to divide the main population into sub-populations. Compared with traditional optimization inversion methods, it has a strong global search capability, which can effectively avoid getting trapped in local minima during the optimization process, and thus obtain the global optimal solution. It can also effectively extract formation parameters for anomalous layer velocity models and noisy data. It has the advantages of small population size and fast convergence speed, and has great application potential for multi-parameter inversion. Attached Figure Description
[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0036] Figure 1 This is a schematic diagram of BSAS.
[0037] Figure 2 This is a flowchart illustrating the steps of the HVSR inversion method based on the multi-peak forest optimization algorithm of this invention.
[0038] Figure 3 shows the inversion results of the horizontally layered increasing velocity model. Figure 3a The shear wave velocity profiles obtained from GA and MMFOA inversions and the composite shear wave velocity profile are shown. Figure 3b The HVSR curves obtained from GA and MMFOA inversion and the composite HVSR curve;
[0039] Figure 4 shows the inversion results of the low-velocity layer model. Figure 4a The shear wave velocity profiles obtained from GA and MMFOA inversions and the composite shear wave velocity profile are shown. Figure 4b The HVSR curves obtained from GA and MMFOA inversion and the composite HVSR curve;
[0040] Figure 5 shows the model inversion results under noisy conditions. Figure 5a The shear wave velocity profiles obtained from GA and MMFOA inversions and the composite shear wave velocity profile are shown. Figure 5b The HVSR curves obtained from GA and MMFOA inversion and the composite HVSR curve are shown. Detailed Implementation
[0041] The present invention will be further described below with reference to embodiments:
[0042] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.
[0043] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0044] Due to the pluralistic nature of the HVSR curve inversion process, conventional optimization algorithms are prone to getting trapped in local minima. Therefore, effectively avoiding local minima during the optimization process is a problem that urgently needs to be solved. To address this issue, this invention provides an HVSR inversion method based on a multi-peaked forest optimization algorithm. This method determines the globally optimal population by dividing the forest population into subpopulations and utilizes the algorithm's global optimization capability to obtain the globally optimal solution model, such as... Figure 2 As shown, the specific steps include:
[0045] Step 1: Use a three-component detector at the measurement point to collect micro-motion data. The acquisition frequency range is 0.2 to 25 Hz. At the same time, ensure the coupling between the sensor and the site to extract the east-west component (E), north-south component (N), and vertical component (Z) of the micro-motion signal from the three-component detector.
[0046] Step 2: The three-component micro-motion data at each measuring point are processed using the inverse STA / LTA algorithm to remove short-time interference. STA is the average value of the micro-motion signal characteristic function within a short-time window for each measuring point, and LTA is the average value of the micro-motion signal characteristic function within a long-time window. The specific formula is as follows:
[0047]
[0048] Where i is the sampling time, ns is the length of the short time window, nl is the length of the long time window, and f(j) is the characteristic function value of the microseismic signal at time i.
[0049] Then, the processed three-component micro-motion data is subjected to Fourier transform to obtain the Fourier amplitude spectrum. The amplitude spectrum is then smoothed using a Konno & Ohmachi window. Smoothing is performed by convolving the window function with the Fourier amplitude spectrum to obtain the frequency domain HVSR curve. The obtained HVSR curve needs to be reliably judged based on the following criteria:
[0050] 21. Peak frequency f0 and time window length l w Satisfy: f0 > 10 / l w ;
[0051] 22. Number of time windows N c Satisfy: n c (f0) > 200;
[0052] 23. If f0 > 0.5 Hz, then for 0.5f0 < f < 2f0, the standard deviation of HVSR satisfies σ a (f) < 2;
[0053] If f0 < 0.5 Hz, then for 0.5f0 < f < 2f0, the standard deviation of HVSR satisfies σ. a (f) < 3;
[0054] Step 3: Use the MMFOA algorithm to invert the HVSR curve, fit the actual HVSR curve and the theoretical HVSR curve to the greatest extent, and obtain the shear wave velocity structure under each region.
[0055] First, an initial model is obtained using a random population initialization, and the reciprocal of the objective function value of the initial population is calculated. The objective function is a function of the underground parameters, and the formula is:
[0056]
[0057] Where I is the number of samples, (HVSR) m For actual measurement of HVSR, (HVSR) s The synthesized HVSR curve is shown at frequency f. The synthesized HVSR is approximately calculated using the equivalent linear method based on the viscoelastic Kelvin-Voigt model. The core of the optimal solution model is to find the global minimum of the objective function.
[0058] Then, the five key iterative steps of MMFOA are performed: forest initialization, local seeding, age-based population constraint, population clustering, and global seeding, which specifically include the following steps;
[0059] Forest initialization is performed using a random population, resulting in an array of length 1*(N+1), which includes N variables and the tree age;
[0060] Local seeding is applied to target trees with an age of 0. During this stage, some trees with an age of 0 are added near the target tree, and the target tree will become 1 year older.
[0061] In the age-based population restriction phase, older trees that have exceeded their maximum lifespan will be removed from the forest and added to the candidate population;
[0062] Population clustering employs the Basic Sequential Algorithm Scheme (BSAS) to search the clustering tree in MMFOA. BSAS accesses the data only once, utilizes a niche mechanism to handle this problem, and creates an adaptive system. This adaptive system selects cluster heads based on Euclidean distance, which is the neighborhood radius between the cluster head and its neighboring sequence. Populations with distances exceeding the Euclidean distance are set as new cluster heads, while populations with distances below the Euclidean distance are sorted according to the reciprocal of the objective function value, with the maximum value becoming the new cluster head. The principle of BSAS is as follows: Figure 1 As shown;
[0063] Global seeding is a process of selecting a percentage of candidate populations.
[0064] Finally, by maximally fitting the actual HVSR curves and the theoretical HVSR curves to minimize the objective function value, an optimal set of formation model solutions is found. Then, the shear wave velocity structure beneath each region is obtained. The method flow is as follows: Figure 2 As shown.
[0065] Step 4: Infer and interpret the anomalous geological body based on the abnormal characteristics of the shear wave velocity structure.
[0066] To better illustrate the effects of the above-described specific implementation methods, a specific embodiment is given below.
[0067] Example 1
[0068] To study different geological conditions, this embodiment sets up two different four-layer synthetic models for inversion to test the performance of the multi-peak forest optimization algorithm. The two models are a horizontally layered increasing velocity model and a model with interspersed low-velocity layers. To make the experimental results more obvious, this embodiment uses a genetic algorithm (GA) for comparison. The inversion parameters used by the genetic algorithm include the number of iterations, population size, and number of genes, which are 100, 1000, and 100, respectively. The minimum and maximum frequency values used for inversion are 0.1Hz and 10Hz, respectively.
[0069] Under the horizontally layered increasing velocity model, using MMFOA inversion parameters with a maximum iteration count of 100, a forest limit of 1000, a maximum allowable tree age of 30, a candidate population percentage of 10, and a threshold of 80, the shear wave velocity profiles obtained from GA and MMFOA inversions are shown below. Figure 3a As shown, the HVSR curves obtained by GA and MMFOA inversion are similar to the synthesized HVSR curves. Figure 3b As shown, the MMFOA algorithm can obtain inversion results with high accuracy.
[0070] Under the low-velocity layer model, using MMFOA inversion parameters with a maximum iteration count of 100, a forest limit of 1000, a maximum allowed tree age of 30, a candidate population percentage of 10, and a threshold of 80, the shear wave velocity profiles obtained from GA and MMFOA inversions are shown below. Figure 4a As shown, the HVSR curves obtained by GA and MMFOA inversion are similar to the synthesized HVSR curves. Figure 4b As shown, the results of the multi-peak forest optimization algorithm fit the synthetic model better for near-surface anomalous geological structures.
[0071] Noise in the micro-motion data can affect the selection of the optimal HVSR curve, leading to instability in the inversion results. Therefore, it is necessary to test the noise resistance of this invention. 5% noise was added to the HVSR curve data of the horizontal layered increasing velocity model to perturb the HVSR curve, and then the inversion effect was tested. Using MMFOA inversion parameters with a maximum iteration count of 100, a forest limit of 1000, a maximum allowed tree age of 30, a candidate population percentage of 20%, and a threshold of 200, the shear wave velocity profiles obtained from GA and MMFOA inversions are compared with the synthetic shear wave velocity profile as shown below. Figure 5a As shown, the HVSR curves obtained by GA and MMFOA inversion are similar to the synthesized HVSR curves. Figure 5b As shown, the MMFOA algorithm has good stability for inversion of noisy data, verifying the effectiveness of the proposed invention.
[0072] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.
Claims
1. A method for HVSR inversion based on a multi-peak forest optimization algorithm, characterized in that: Includes the following steps: Step 1: At the measurement point, use a three-component detector to collect micro-motion data and extract the east-west component, north-south component, and vertical component data of the micro-motion signal from the three-component detector. Step 2: Ratio the average value of the characteristic function of the micro-motion signal within the short-time window to the average value of the characteristic function of the micro-motion signal within the long-time window at each measurement point to eliminate short-time interference. Then, perform Fourier transform on the processed three-component micro-motion data to obtain the Fourier amplitude spectrum, and smooth the amplitude spectrum using the Konno & Ohmachi window. Smoothing is performed by convolving the window function with the Fourier amplitude spectrum to obtain a reliable frequency domain HVSR curve and perform reliability judgment. Step 3: Use the MMFOA algorithm to invert the HVSR curve, fit the actual HVSR curve and the theoretical HVSR curve to the greatest extent, and obtain the shear wave velocity structure under each region. Step 4: Infer and interpret the anomalous geological bodies based on the abnormal characteristics of the shear wave velocity profile.
2. The HVSR inversion method based on the multi-peak forest optimization algorithm according to claim 1, characterized in that: Step 1: When collecting micro-motion data, the acquisition frequency range is approximately 0.2 to 25 Hz, while ensuring the coupling between the sensor and the site.
3. The HVSR inversion method based on the multi-peak forest optimization algorithm according to claim 1, characterized in that, In step 2, the average value of the characteristic function of the micro-motion signal within the short-time window at each measuring point is compared with the average value of the characteristic function of the micro-motion signal within the long-time window. The specific formula is as follows: in, i Let be the sampling time, ns be the length of the short time window, nl be the length of the long time window, and f(j) be the characteristic function value of the microseismic signal at time i.
4. The HVSR inversion method based on the multi-peak forest optimization algorithm according to claim 1, characterized in that, In step 2, the reliability criteria for the HVSR curve are as follows:
21. Peak frequency f0 and time window length l w Satisfy: f0 > 10 / l w ; 22. Number of time windows N c Satisfy: n c (f0) > 200; 23. If f0 > 0.5 Hz, then for 0.5f0 < f < 2f0, the standard deviation of HVSR satisfies σ a (f) < 2; If f0 < 0.5 Hz, then for 0.5f0 < f < 2f0, the standard deviation of HVSR satisfies σ. a (f) < 3.
5. The HVSR inversion method based on the multi-peak forest optimization algorithm according to claim 1, characterized in that, In step 3, the MMFOA algorithm inverts the HVSR curve, specifically including the following steps:
31. Load the observed HVSR data; use random population initialization to obtain the initial model, and calculate the reciprocal of the objective function value of the initial population; 32. Perform five key iterative steps of MMFOA: forest initialization, local seeding, age-based population constraint, population clustering, and global seeding; 33. By fitting the actual HVSR curves and the theoretical HVSR curves to the greatest extent possible, an optimal set of formation model solutions can be found.
6. The HVSR inversion method based on the multi-peak forest optimization algorithm according to claim 5, characterized in that, In step 3, the value of the objective function is a function of the underground parameters, and the formula is: Where I is the number of samples, (HVSR) m For actual measurement of HVSR, (HVSR) s The synthesized HVSR curve is shown at frequency f. The synthesized HVSR is approximately calculated using the equivalent linear method based on the viscoelastic Kelvin-Voigt model. The core of the optimal solution model is to find the global minimum of the objective function.
7. The HVSR inversion method based on the multi-peak forest optimization algorithm according to claim 5, characterized in that, The five key iterative steps of MMFOA in step 3 are as follows: a. Forest initialization: Initialize using a random population to obtain an array of length 1*(N+1), including N variables and tree age; b. Local seeding: Applied to target trees with an age of 0. During this stage, some trees with an age of 0 are added near the target tree, and the target tree will become 1 year older. c. Age-based population restriction: In this stage, old trees that have exceeded their maximum lifespan will be removed from the forest and added to the candidate population; d. Population clustering: The Basic Sequential Algorithm Scheme (BSAS) is used to search the clustering tree in MMFOA. BSAS will access the data only once, uses a niche mechanism to handle this problem, and creates an adaptive system. e. Global seeding operation: Select a percentage of the candidate population.