A method for automatically picking multi-modal dispersion curves of seismic surface waves

Abstract

The present invention discloses a method for automatically picking up multi-modal dispersion curves of seismic surface waves, which belongs to the field of earth exploration and information technology, including non-normalized dispersion energy imaging of seismic surface wave data; segmenting the dispersion energy map and obtaining the maximum value points of each part; estimating the next dispersion curve point by using a time series prediction algorithm; using a penalty function to search up and down with the predicted point as the center; searching forward and backward to extract all dispersion curve points; using an evaluation function to select the optimal dispersion curve; re-searching the next modal dispersion curve after the current mode is set to zero. The present invention adopts the above-mentioned method for automatically picking up multi-modal dispersion curves of seismic surface waves, which can provide accurate and reliable dispersion curve data for subsequent surface wave inversion, is helpful for geological disaster detection, engineering geological survey and other aspects, solves the problem of low efficiency of manual picking of dispersion curves, and has strong practicality and innovation in practice.

Description

Technical Field

[0001] The invention relates to the field of earth detection and information technology, and in particular to a method for automatically picking up multi-modal dispersion curves of seismic surface waves. Background Art

[0002] Dispersion curve picking is similar to first arrival and velocity picking in seismic exploration, which is a very tedious and time-consuming task. In the past, dispersion curve picking was mostly done manually, and the results were greatly affected by the experience of the picker and consumed a lot of manpower and time. In particular, the amount of data in three-dimensional seismic exploration is large, and the work of dispersion curve picking will greatly increase the production cycle. Currently, major commercial software, such as the more famous Parkseis, Georgia, Surfseis, Geopsy or EasyMASW, only supports users to manually pick dispersion curves based on dispersion imaging images. This manual picking method may often be affected by the experience and subjective understanding of the processing personnel, resulting in inconsistent picking standards or even errors. Studies have shown that a slight error in dispersion curve picking may automatically select an erroneous dispersion point with lower energy in the software, which will cause the imaging error of the shear wave velocity profile to further expand for subsequent inversion. Therefore, it is extremely necessary to develop an automatic dispersion curve picking method that is not affected by user subjective factors.

[0003] In addition, the dispersion curve picking method based on artificial intelligence has defects such as weak generalization or migration ability and high training cost. Therefore, exploring a fast and accurate surface wave dispersion curve automatic picking method is a very practical content for surface wave inversion in 3D seismic exploration data. Summary of the invention

[0004] The purpose of the present invention is to provide a method for automatically picking up multi-modal dispersion curves of seismic surface waves to solve the problems mentioned in the above background technology.

[0005] To achieve the above object, the present invention provides a method for automatically picking up multi-modal dispersion curves of seismic surface waves, comprising the following steps:

[0006] S1. Using multiple signal classification methods to perform dispersion energy imaging on seismic surface wave data to obtain a non-normalized multi-modal surface wave dispersion energy map;

[0007] S2, dividing the multi-modal surface wave dispersion energy map into g regions, and calculating the maximum value of each region;

[0008] S3, taking the maximum value obtained in S2 as the initial point, and using the existing tracked dispersion curve points to predict the next dispersion curve point using a time series prediction algorithm;

[0009] S4, according to the variation law of the dispersion curve, a penalty function with a velocity jump penalty factor α and a space jump penalty factor β is introduced, and the dispersion curve point predicted by S3 is used as the center to search up and down to determine the optimal dispersion curve point position;

[0010] S5, taking the g initial points in S2 as the center, search the dispersion curve points forward and backward according to the methods of S3 and S4, and obtain g dispersion curves;

[0011] S6, using the evaluation function to select the optimal dispersion curve for the g dispersion curves obtained in S5;

[0012] S7, setting a certain width area centered on the optimal dispersion curve obtained in S6 to 0, repeating steps S2 to S6 to calculate the dispersion curve point of the next mode, and so on, until the optimal dispersion curves of all modes are calculated.

[0013] Preferably, the time series prediction algorithm in S3 predicts the dispersion curve points, comprising the following steps:

[0014] S31, when the predicted point is adjacent to the initial point, a neighboring point interpolation algorithm is used to predict the next dispersion curve point forward and backward respectively;

[0015] S32, when the interval between the prediction point and the initial point is 1 to 2 points, linear interpolation is used to predict the dispersion curve point forward and backward;

[0016] S33. When the interval between the predicted point and the initial point is greater than 3 points, the dispersion curve points are predicted forward and backward using exponential smoothing, spline interpolation, Lagrangian or quadratic polynomial linear regression analysis according to the variation law of the dispersion curve.

[0017] Preferably, the penalty function in S4 is:

[0018]

[0019] Where Δ is the maximum allowed search interval, DC p (h) is the predicted value of the hth frequency sample point, DC α (h) is the dispersion curve point corresponding to the frequency h of the reference point dispersion energy diagram, α is the velocity jump penalty factor, and β is the spatial jump penalty factor of the reference point dispersion diagram. The reference point dispersion curve can be calculated using known drilling information or the dispersion curve of the adjacent point that has been picked up.

[0020] Preferably, the evaluation functions in step S6 are maximum absolute value energy, maximum square energy and minimum mean square error.

[0021] Therefore, the present invention adopts the above-mentioned method for automatically picking up multi-modal dispersion curves of seismic surface waves, which has the following beneficial effects:

[0022] (1) By referring to the multi-ridge tracking principle in time-frequency analysis, combined with the greedy algorithm and time series prediction method, the automatic picking and classification of multi-modal dispersion curves can be achieved quickly and accurately in a strong noise environment;

[0023] (2) It has the advantages of strong noise resistance, high computational efficiency, simple method and easy programming implementation;

[0024] (3) It has the ability to automatically pick up and classify multiple modes, which is not available in many conventional dispersion curves.

[0025] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of an embodiment of the present invention;

[0027] Figure 2 The seismic surface wave data of the embodiment of the present invention;

[0028] Figure 3 This is a dispersion energy imaging diagram of an embodiment of the present invention;

[0029] Figure 4 This is a dispersion curve automatically picked up according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] Example

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] Reference Figure 1 The present invention discloses a method for automatically picking up multi-modal dispersion curves of seismic surface waves, comprising the following steps:

[0033] S1. The multiple signal classification method is used to perform dispersion energy imaging on seismic surface wave data to obtain a non-normalized multi-modal surface wave dispersion energy map.

[0034] S2. Divide the multimodal surface wave dispersion energy map into g parts and calculate the maximum value of each part.

[0035] S3, taking the maximum value obtained in S2 as the initial point, using the time series prediction algorithm, and using the tracked dispersion curve point to predict the next dispersion curve point.

[0036] The time series prediction algorithm predicts the dispersion curve points, including the following steps:

[0037] S31, when the predicted point is adjacent to the initial point, a neighboring point interpolation algorithm is used to predict the next dispersion curve point forward and backward respectively;

[0038] S32, when the interval between the prediction point and the initial point is 1 to 2 points, linear interpolation is used to predict the dispersion curve point forward and backward;

[0039] S33. When the interval between the predicted point and the initial point is greater than 3 points, the dispersion curve points are predicted forward and backward using exponential smoothing, spline interpolation, Lagrangian or quadratic polynomial linear regression analysis according to the variation law of the dispersion curve.

[0040] S4, according to the variation law of the dispersion curve, introduce a penalty function with a velocity jump penalty factor α and a space jump penalty factor β, and search up and down to determine the best dispersion curve point position with the dispersion curve point predicted by S3 as the center, to ensure the local continuity of the dispersion curve in the frequency direction and the gradual change of adjacent points in space. The penalty function is:

[0041]

[0042] Where Δ is the maximum allowed search interval, DC p (h) is the predicted value of the hth frequency sample point, DC α (h) is the dispersion curve point corresponding to the frequency h of the reference point dispersion energy diagram, α is the velocity jump penalty factor, and β is the spatial jump penalty factor of the reference point dispersion diagram. The reference point dispersion curve can be calculated using known drilling information or the dispersion curve of the adjacent point that has been picked up.

[0043] S5. Using the g initial points in S2 as the center, search for dispersion curve points forward and backward according to the methods of S3 and S4 to obtain g dispersion curves.

[0044] S6. Select the optimal dispersion curve for the g dispersion curves obtained in S5 using an evaluation function. The evaluation function is the maximum absolute value energy, the maximum square energy and the minimum mean square error.

[0045] S7, setting a certain width area centered on the optimal dispersion curve of S6 to 0, repeating steps S2 to S6 to calculate the dispersion curve point of the next mode, and so on, until the dispersion curves of all modes are calculated. Specific embodiments

[0047] The above method flow is explained by taking some synthetic surface wave dispersion energy data as an example.

[0048] 1: For the forward modeling seismic surface wave records (such as Figure 2) The multi-signal classification method is used to perform dispersion energy imaging to obtain a non-normalized multi-modal surface wave dispersion energy map, where the phase velocity range is 100-1000 m / s and the frequency range is 5-10 Hz (e.g. Figure 3 );

[0049] 2: Divide the entire dispersion energy diagram into two parts according to the frequency axis, where the frequency range of the left half is 5 to 27 Hz, and the frequency range of the right half is 28 to 50 Hz. Calculate the maximum value points of the two dispersion energy diagrams respectively.

[0050] 3: Take the above two maximum points as initial points respectively, and use the quadratic polynomial linear regression analysis method in the time prediction algorithm to predict the next dispersion curve point based on the dispersion curve point between the initial point and the current point.

[0051] 4: According to the variation law of the dispersion curve, a penalty function with a velocity jump penalty factor α and a space jump penalty factor β is introduced. The dispersion curve point predicted in step 3 is used as the center for upward and downward searches to determine the optimal dispersion curve point position, ensuring the local continuity of the dispersion curve in the frequency direction and the gradual change of adjacent points in space. The penalty function is as follows:

[0052]

[0053] Where Δ is the maximum allowed search interval, DC p (h) is the predicted value of the hth frequency sample point, DC a (h) is the dispersion curve point corresponding to the frequency h of the reference point dispersion energy diagram (the reference point dispersion curve can be calculated using known drilling information or the dispersion curve of the adjacent point that has been picked up), α is the velocity jump penalty factor, and β is the spatial jump penalty factor of the reference point dispersion diagram. In this study example, Δ is taken as 20, α is 0.06, and β is 0.

[0054] 5: Calculate all dispersion curve points according to step 3 and step 4. In this embodiment, there are 2 initial points, so 2 different dispersion curves are obtained.

[0055] 6: For the above two dispersion curves, use the evaluation function to select the optimal dispersion curve. In this embodiment, the evaluation function uses the maximum energy method, that is, the maximum value of the sum of the total energy of the two dispersion curves is taken as the optimal fundamental modal dispersion curve. The maximum energy method function used is:

[0056]

[0057] Where DC is the dispersion curve, i is the initial point, h is the frequency point, and n is the number of frequency points.

[0058] 7: Set the dispersion energy diagram area with the optimal dispersion curve in step 6 as the center and a width of 200 to 0. Follow steps 1 to 6 again to obtain the dispersion curve of the dispersion energy diagram at this time as the dispersion curve of the first high-order mode.

[0059] Figure 4 The method of this embodiment is compared with the dispersion curve picked by the latest local maximum dispersion curve picking method. Figure 4 It can be seen that the local maximum method has obvious jitter and the dispersion curve picking error is large, while the method proposed in this embodiment effectively reduces the influence of random strong noise, making the dispersion curve picking more stable and continuous, and has great practical value.

[0060] Therefore, the present invention adopts the above-mentioned method for automatically picking multi-modal dispersion curves of seismic surface waves, and adopts the forward and backward greedy algorithm to track the dispersion curve. In order to ensure the stability of tracking, the concept of time series prediction is introduced into the dispersion curve tracking. At the same time, the dispersion imaging diagram is divided into multiple segments to obtain local maximum points, which reduces the interference of abnormal strong noise on the initial value of dispersion curve tracking. It can be applied to geotechnical engineering investigation, non-destructive testing and geological disaster detection, and has high practical value and innovation. It can provide accurate and reliable dispersion curve data for subsequent surface wave inversion, which is helpful for geological disaster detection, engineering geological survey and other aspects, solves the problem of low efficiency of manual picking of dispersion curves, and has strong practicality and innovation in practice.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for automatically picking up multi-modal dispersion curves of seismic surface waves, characterized in that: The following steps are involved: S1. Using multiple signal classification methods to perform dispersion energy imaging on seismic surface wave data, and obtaining non-normalized surface wave dispersion energy maps of several modes; S2, dividing the surface wave dispersion energy map of each mode into g regions, and calculating the maximum value of each region; S3, taking the maximum value obtained in S2 as the initial point, using the time series prediction algorithm, and using the tracked dispersion curve points to predict the next dispersion curve point; S4, according to the variation law of the dispersion curve, a penalty function with a velocity jump penalty factor α and a space jump penalty factor β is introduced, and the dispersion curve point predicted by S3 is used as the center to search up and down to determine the optimal dispersion curve point position; S5, taking the g initial points in S2 as the center, search the dispersion curve points forward and backward according to the methods of S3 and S4, and obtain g dispersion curves; S6, selecting the optimal dispersion curve using the evaluation function for the g dispersion curves obtained in S5; S7, setting the width region centered on the optimal dispersion curve obtained in S6 to 0, repeating steps S2 to S6 to calculate the dispersion curve point of the next mode, and calculating the optimal dispersion curves of all modes.

2. The method for automatically picking up multi-modal dispersion curves of seismic surface waves according to claim 1, characterized in that: The time series prediction algorithm in S3 predicts dispersion curve points, including the following steps: S31, when the predicted point is adjacent to the initial point, a neighboring point interpolation algorithm is used to predict the next dispersion curve point forward and backward respectively; S32, when the interval between the prediction point and the initial point is 1 to 2 points, linear interpolation is used to predict the dispersion curve point forward and backward; S33. When the interval between the predicted point and the initial point is greater than 3 points, the dispersion curve points are predicted forward and backward using exponential smoothing, spline interpolation, Lagrangian or quadratic polynomial linear regression analysis according to the variation law of the dispersion curve.

3. The method for automatically picking up multi-modal dispersion curves of seismic surface waves according to claim 1, characterized in that: The penalty function in S4 is: Where Δ is the maximum allowed search interval, DC p (h) is the predicted value of the hth frequency sample point, DC α (h) is the dispersion curve point corresponding to the frequency h of the reference point dispersion energy diagram, α is the velocity jump penalty factor, and β is the spatial jump penalty factor of the reference point dispersion diagram.

4. The method for automatically picking up multi-modal dispersion curves of seismic surface waves according to claim 1, characterized in that: The evaluation functions in step S6 are maximum absolute value energy, maximum square energy and minimum mean square error.

Images