S-wave velocity estimation system, program, and method for analysis

Abstract

To provide an S-wave velocity estimation system, a program, and a method for analysis which can rapidly and accurately estimate the average S-wave velocity from the ground surface to a depth of at least 30 meters.SOLUTION: An S-wave velocity estimation system includes: acquisition means for acquiring the horizontal-to-vertical spectrum ratio of a microtremor in a first position; estimation processing means for acquiring the average S-wave velocity from the ground surface to a depth of 30 meters in the first position, by providing the acquired horizontal-to-vertical spectral ratio of a microtremor in the first position and a microtopographic classification in the first position as inputs to a model trained by a machine learning to estimate the average S-wave velocity of the ground; and output means for outputting the aquired average S-wave velocity.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present disclosure relates to an S-wave velocity estimation system, a program, and an analysis method.

Background Art

[0002] As an index for easily evaluating ground characteristics, the average S-wave velocity (AVS30) from the ground surface to 30 m underground is often used. As described in Non-Patent Document 1, AVS30 is calculated from the S-wave velocity structure (layered S-wave velocity) of the ground. Non-Patent Document 1 also describes a technique for estimating AVS30 by machine learning from the horizontal / vertical spectrum ratio of natural earthquakes.

Prior Art Documents

Non-Patent Documents

[0003]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the technique of Non-Patent Document 1, when calculating the average S-wave velocity from the ground surface to 30 m underground from the horizontal / vertical spectrum ratio of natural earthquakes, it was difficult to speed up the preprocessing of input data. Also, when estimating AVS30 by machine learning from the horizontal / vertical spectrum ratio, it was difficult to improve the accuracy.

[0005] In view of such a point, an object of the present disclosure is to provide an S-wave velocity estimation system, a program, and an analysis method capable of estimating the average S-wave velocity from the ground surface to at least 30 m underground at high speed and with high accuracy.

Means for Solving the Problems

[0006] The S-wave velocity estimation system according to the first aspect of the present disclosure is (1) an acquisition means for acquiring the horizontal / vertical spectral ratio of ambient vibration at a first position; an estimation processing means for obtaining the average S-wave velocity from the ground surface to 30 m underground at the first position by inputting the acquired horizontal / vertical spectral ratio of ambient vibration at the first position and the microtopography classification at the first position to a learned model that has performed machine learning for estimating the average S-wave velocity of the ground; an output means for outputting the acquired average S-wave velocity; and includes.

[0007] The S-wave velocity estimation system according to one embodiment of the present disclosure is (2) coordinates including the elevation of the first position are further input to the learned model, The S-wave velocity estimation system according to (1) above.

[0008] The S-wave velocity estimation system according to one embodiment of the present disclosure is (3) the learned model is a neural network model, The S-wave velocity estimation system according to (1) or (2) above.

[0009] The program according to the second aspect of the present disclosure is (4) to a computer, A function of accumulating the obtained horizontal / vertical spectral ratio of the ambient vibration and the microtopography classification at the position where the horizontal / vertical spectral ratio was obtained. A function of accumulating the average S-wave velocity from the ground surface to 30 m underground in association with the horizontal / vertical spectral ratio and the microtopography classification. A function of performing machine learning on the relationship between the accumulated horizontal / vertical spectral ratio and the microtopography classification and the average S-wave velocity from the ground surface to 30 m underground. A function of estimating the average S-wave velocity from the ground surface to 30 m underground by using the result of the machine learning when the horizontal / vertical spectral ratio and the microtopography classification are input. A function of outputting the estimated average S-wave velocity from the ground surface to 30 m underground. A program for executing the above.

[0010] As a third aspect of the present disclosure, an analysis method (5) A step of storing a learned model obtained by performing machine learning for estimating the average S-wave velocity of the ground; A step of obtaining the horizontal / vertical spectral ratio of the ambient vibration at a first position; A step of obtaining the average S-wave velocity from the ground surface to 30 m underground at the first position by giving the obtained horizontal / vertical spectral ratio of the ambient vibration at the first position and the microtopography classification at the first position to the learned model; A step of outputting the average S-wave velocity from the ground surface to 30 m underground; including An analysis method executed by a computer.

Advantages of the Invention

[0011] According to the present disclosure, the average S-wave velocity from at least the ground surface to 30 m underground can be estimated at high speed and with high accuracy.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Embodiments for Carrying Out the Invention

[0013] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the components shown in the following drawings, the same components are denoted by the same reference numerals.

[0014] The S-wave velocity estimation system is used, for example, when evaluating ground characteristics. As shown in FIG. 1, a seismometer 11 of an acquisition means (described later) of the S-wave velocity estimation system is arranged on the surface of the ground 1000.

[0015] Here, FIG. 2 shows a schematic configuration of the S-wave velocity estimation system 100. The S-wave velocity estimation system 100 includes an acquisition unit 10, an estimation processing unit 20, and an output unit 30.

[0016] The acquisition unit 10 acquires the horizontal / vertical spectral ratio of ambient vibrations (hereinafter also referred to as "horizontal / vertical spectral ratio") at the installation position (the first position). Note that ambient vibrations are caused by wind, waves, passing vehicles, operating machinery, etc. By acquiring the waveform of ambient vibrations instead of the waveform of natural earthquakes, a large number of horizontal / vertical spectral ratios can be acquired in a short time.

[0017] The acquisition unit 10 may include a seismometer 11, a logger 12, and a computer. The seismometer 11 measures three components (vertical motion, horizontal motion) of ambient vibrations. The seismometer 11 is wired to the logger 12 and transmits the measured data. The logger 12 wirelessly transmits the data measured by the seismometer 11 to the computer at regular intervals. The computer may exist on the cloud. The computer may calculate a large number of horizontal / vertical spectral ratios from the waveform of ambient vibrations acquired by the seismometer 11. The computer may average the calculated large number of horizontal / vertical spectral ratios.

[0018] The horizontal / vertical spectral ratio of ambient vibrations may be acquired in a format corresponding to the input data for the learned model. The acquisition unit 10 may convert the measured information into a format corresponding to the input data for the learned model.

[0019] The acquisition unit 10 may acquire coordinates including the elevation of the first position. The acquisition unit 10 may include a Global Positioning System (GPS), an altimeter, etc. The acquisition unit 10 may include a keyboard or the like, and the coordinates including the elevation of the first position may be input by the user to the acquisition unit 10.

[0020] The estimation processing means 20 obtains the average S-wave velocity of the ground 1000 at the first position by providing, as inputs, the horizontal / vertical spectral ratio of the microtremor at the first position and the microtopography classification at the first position, which are acquired by the acquisition means 10, to a learned model that has performed machine learning for estimating the average S-wave velocity of the ground 1000. The average S-wave velocity to be obtained may be the average S-wave velocity from the ground surface to 30 m underground at the first position, or the ground may be divided into a plurality of layers in the depth direction, and the average S-wave velocity in each layer at each depth may be obtained. The estimation processing means 20 may be connected to the acquisition means 10 by wire or wirelessly.

[0021] By using the learned model, the estimation processing means 20 can obtain the average S-wave velocity at high speed. Therefore, the user can confirm on-site that the obtained average S-wave velocity is appropriate. The user may acquire the horizontal / vertical spectral ratio of the microtremor again as necessary.

[0022] The output means 30 outputs the average S-wave velocity acquired by the estimation processing means 20. When the average S-wave velocity includes values for a plurality of depths from the ground surface, the output means 30 may output the values of the average S-wave velocity for one or more depths. The output means 30 may include a display for displaying the average S-wave velocity. The output means 30 may be connected to the estimation processing means 20 by wire or wirelessly.

[0023] (Input data) FIG. 3 shows an example of input data for the learned model.

[0024] The input data includes the horizontal / vertical spectral ratio of the microtremor. The horizontal / vertical spectral ratio of the microtremor may include values for a plurality of frequencies. For example, the horizontal / vertical spectral ratio of the microtremor may be acquired for 20 frequencies. The 20 frequencies may include 0.2 Hz, 0.245 Hz, and 10 Hz.

[0025] The input data includes micro-topographic divisions. The micro-topographic divisions may be those provided by the National Research Institute for Earth Science and Disaster Resilience. The micro-topographic divisions may also be divided by other methods. Specifically, as shown in FIG. 3, the micro-topographic divisions may be divided into mountains, foothills, natural levees, backswamps, lakes, etc. For example, the Holocene alluvium and Pleistocene terraces can also be distinguished by the micro-topographic divisions. Site classes C, D, and E can also be distinguished by the micro-topographic divisions.

[0026] Since the input data includes micro-topographic divisions, the estimation processing means 20 of the S-wave velocity estimation system 100 can estimate the average S-wave velocity from the ground surface to at least 30 m underground with higher accuracy.

[0027] The input data may further include coordinates including the elevation of the first position. More specifically, the coordinates may include latitude, longitude, and elevation.

[0028] The input data may include the bedrock depth at which the S-wave velocity exceeds a predetermined value. The predetermined value of the S-wave velocity may be 760 m / s and 1500 m / s. The bedrock depth may utilize known data.

[0029] Since the input data includes coordinates including the elevation of the first position or the bedrock depth at which the S-wave velocity exceeds a predetermined value, the estimation processing means 20 of the S-wave velocity estimation system 100 can estimate the average S-wave velocity from the ground surface to 30 m underground with even higher accuracy.

[0030] (Output data) FIG. 4 shows an example of output data for the learned model.

[0031] The output data includes the average S-wave velocity from the ground surface to 30 m underground. Further, the output data may include the respective average S-wave velocities for a plurality (e.g., 12) of depth layers from the ground surface. The depths may be 0 - 1 m, 1 - 2 m, 2 - 3.3 m, 3.3 - 4.8 m, 4.8 - 6.9 m, 6.9 - 10 m, 10 - 14 m ··· 43 - 62 m, 62 - 90 m.

[0032] (Hardware Configuration of Computer) The S-wave velocity estimation system 100 may be implemented by a computer 200 shown in FIG. 5. Hereinafter, the hardware configuration of the computer 200 will be described in detail. The computer 200 includes a display unit 210, an operation unit 220, a storage unit 230, a control unit 240, and a communication unit 250. Note that the computer 200 may be the same as the computer included in the acquisition means 10.

[0033] The display unit 210 corresponds to the output means 30 of the S-wave velocity estimation system 100. The display unit 210 may be a display device. The display device may be, for example, a liquid crystal display or an OEL (Organic Electroluminescence) display, etc., but is not limited thereto and can be any display device.

[0034] The operation unit 220 is an input interface that receives user operations. The input interface may be, for example, a pointing device such as a mouse, a physical key, or a touch panel provided integrally with the display unit 210, etc., but is not limited thereto and can be any input interface.

[0035] The storage unit 230 is a storage device including one or more memories. The memory may be, for example, a semiconductor memory, a magnetic memory, or an optical memory, etc., but is not limited thereto and can be any memory. The storage unit 230 functions as, for example, a primary storage device or a secondary storage device. The storage unit 230 is, for example, built into the computer 200, but a configuration in which it is externally connected to the computer 200 via an arbitrary interface is also possible.

[0036] The storage unit 230 stores a learned model obtained by performing machine learning.

[0037] In the present embodiment, a model learned by teacher data is stored in the storage unit 230, but a pre-learning model may be stored in the storage unit 230. In this case, learning of the prediction model is performed by the machine learning function described later.

[0038] The control unit 240 is one or more processors. The processor can be, for example, a general-purpose processor or a dedicated processor specialized for specific processing, but is not limited thereto and can be any processor. The control unit 240 controls the operation of the entire S-wave velocity estimation system 100.

[0039] The communication unit 250 is one or more interfaces that communicate with an external device via wireless or wired means. In the present embodiment, the communication unit 250 can communicate with the logger 12 shown in FIG. 1. The communication between the communication unit 250 and the logger 12 is executed via an arbitrary device such as a PLC, for example, but a configuration in which the communication is executed without passing through the device is also possible.

[0040] (Software configuration of the computer) Next, the software configuration of the computer 200 will be described. A program used to control the operation of the computer 200 is stored in the storage unit 230. When the program is read by the control unit 240, the control unit 240 is caused to execute an input acquisition function, an association function, a machine learning function, an estimation function, and an output function.

[0041] The input acquisition function is a function of accumulating the acquired horizontal / vertical spectrum ratio of the ambient vibration and the microtopography classification at the position where the horizontal / vertical spectrum ratio is acquired. The computer 200 may acquire the horizontal / vertical spectrum ratio of the ambient vibration using the measurement data of the seismometer 11 (FIG. 1) described above. The user may input the microtopography classification using the operation unit 220. The horizontal / vertical spectrum ratio and the microtopography classification may be accumulated in the storage unit 230 of the computer 200.

[0042] The association function is a function that accumulates the average S-wave velocity from the ground surface to at least 30 m underground in association with the horizontal / vertical spectrum ratio and the microtopography classification. The average S-wave velocity may be calculated in advance from the horizontal / vertical spectrum ratio of the ambient vibration. The average S-wave velocity may be measured in advance by multi-channel analysis of surface waves (MASW) or microtremor array survey (MAM). The average S-wave velocity may be accumulated in the storage unit 230.

[0043] The machine learning function is a function that machine-learns the relationship between the horizontal / vertical spectrum ratio and the microtopography classification, and the average S-wave velocity from the ground surface to at least 30 m underground, which are accumulated by the input acquisition function and the association function. Any method can be applied to the machine learning.

[0044] The estimation function is a function that estimates the average S-wave velocity from the ground surface to at least 30 m underground by using the result of machine learning by the machine learning function when the horizontal / vertical spectrum ratio and the microtopography classification are input.

[0045] The output function is a function that outputs the average S-wave velocity from the ground surface to at least 30 m underground estimated by the estimation function. The computer 200 may output the average S-wave velocity from the ground surface to at least 30 m underground estimated by the output means 30 of the S-wave velocity estimation system 100 described above.

[0046] (Neural network) The learned model may be a neural network model. As shown in FIG. 6, the neural network model 300 is composed of an input layer 310, hidden layers 321, 322, and an output layer 330. Input data is input to the input layer 310. The output layer 330 outputs output data. The hidden layers 321, 322 exist between the input layer 310 and the output layer 330. The input layer 310, the hidden layers 321, 322, and the output layer 330 may be fully connected layers.

[0047] The hidden layer may be composed of a first hidden layer 321 and a second hidden layer 322. The first hidden layer 321 may include 60 neurons. The second hidden layer 322 may include 60 neurons.

[0048] By using the learned model as the neural network model 300, the estimation processing means 20 of the S-wave velocity estimation system 100 can estimate the average S-wave velocity from the ground surface to at least 30 m underground with higher accuracy.

[0049] (Analysis method executed by a computer) With reference to FIG. 7, an analysis method executed by the computer 200 will be described.

[0050] Step S100: The control unit 240 of the computer 200 stores a learned model obtained by performing machine learning for estimating the average S-wave velocity of the ground 1000. The learned model may be stored in the storage unit 230 of the computer 200.

[0051] Step S101: The control unit 240 of the computer 200 acquires the horizontal / vertical spectral ratio of the ambient vibration at the first position (observation point position). By acquiring the waveform of the ambient vibration instead of the waveform of the natural earthquake, data can be acquired at any location and time, and a large number of horizontal / vertical spectral ratios can be acquired in a short time. The acquired large number of horizontal / vertical spectral ratios may be averaged.

[0052] Step S102: The control unit 240 of the computer 200 provides the learned model with the horizontal / vertical spectral ratio of the ambient vibration at the first position acquired in step S101 and the microtopography classification at the first position, thereby acquiring the average S-wave velocity from the ground surface to at least 30 m underground at the first position (actually, all the observation point positions).

[0053] Step S103: The control unit 240 of the computer 200 outputs the average S-wave velocity from the ground surface to at least 30 m underground at the first position (all the observation point positions).

Example

[0054] <Abstract> In Example 1, supervised machine learning was applied to estimate AVS30 from the horizontal / vertical spectral ratio together with other information. This machine learning consists of a neural network with two hidden layers. A pair of H / V spectrum (input layer) and Vs profile (output layer) is used as training data. The machine learning was applied to several different locations, and it was confirmed that the predicted AVS30 generally agrees with the observed AVS30. Here, data in the southern Kanto Plain of Japan was used.

[0055] <Machine Learning Method> The supervised machine learning consists of a fully connected neural network with two hidden layers. A pair of the observed H / V spectrum (input layer) and Vs profile (output layer) is used as training data. The input layer consists of the observed H / V spectrum resampled to 20 samples in the frequency range of 0.2 to 10 Hz, the site coordinates (latitude, longitude, altitude), and the topographical classification of the site. The output layer is the velocity profile obtained from MASW, MAM, or the inversion of H / V. The Vs profile from the ground surface to a depth of 90 m was represented by 12 layers.

[0056] <Results> In this example, this method was applied to several locations in the United States and Japan. Hereinafter, the experimental results in the southern Kanto Plain of Japan are shown. At an investigation site of 50 km × 50 km, MASW, MAM, and / or H / V were measured at about 1500 locations. Pairs of H / V spectra, together with their coordinates, elevations, and microtopographical classifications (J-SHIS, 2023: Non-Patent Document 2), and the Vs profiles obtained from the inversion of MASW, MAM, and H / V spectra, constitute the training data. Figure 8 shows the AVS30 and the peak frequencies of the H / V spectra used for training. The AVS30 and peak frequencies of H / V are in the ranges of 100 to 500 m / s and 0.6 to 10 Hz, respectively.

[0057] Both AVS30 and H / V have distinct differences between the Holocene alluvial plain and the Pleistocene terrace. AVS30 ranges from 100 to 180 m / s (site class E) and 180 to 500 m / s (site classes C and D) in the Holocene alluvial plain and Pleistocene terrace, respectively. The H / V peak frequency ranges from 0.7 to 2.0 Hz and 2.0 to 10 Hz in the Holocene alluvial plain and Pleistocene terrace, respectively. Figure 9 shows the comparison between the observed AVS30 and the AVS30 predicted based on the trained network. In this example, half of the pairs were used for training. The remaining pairs were used to evaluate the prediction accuracy. The trained network predicted nearly correct values for the data used in training. Also, not very different values were predicted for the data not used in training.

[0058] <Conclusion> The results suggested that supervised machine learning can roughly predict the Vs profile or AVS30 from the H / V spectrum, along with other available information. Different from the conventional inversion of H / V, machine learning is computationally fast and has no non-uniqueness. The measurement of H / V is much easier and faster compared to conventional direct Vs measurements such as MASW, MAM, or velocity records. H / V measurement and machine learning make it possible to interpolate the Vs profile obtained from direct Vs measurements and depict the 2D or 3D Vs model of the region.

Example

[0059] As Example 2, a prototype S-wave velocity estimation system according to the present invention was fabricated and its performance was evaluated.

[0060] In the fabricated S-wave velocity estimation system, the learned model is a neural network model. The learned model was given, as inputs, the horizontal / vertical spectrum ratio, microtopography classification, and coordinates including the elevation of the learning data acquisition position, and was machine-learned. The machine-learned model was stored in the storage unit of the S-wave velocity estimation system.

[0061] Using the acquisition means 10 of the prototype S-wave velocity estimation system, the horizontal / vertical spectral ratios of ambient vibrations were acquired at a large number of randomly selected test data acquisition positions at all times. From the obtained test data, example input data and comparative example input data were created. The example input data and the comparative example input data are different from the data used for machine learning.

[0062] In the prototype S-wave velocity estimation system, for the learned model, the example input data was given as input, and the example output data was acquired. The example input data is the coordinates including the horizontal / vertical spectral ratio of ambient vibrations at the first position, the microtopography classification at the first position, and the elevation of the first position. Fig. 10 shows the results with the vertical axis being the example output data (predicted AVS) and the horizontal axis being the observed AVS.

[0063] Furthermore, in the prototype S-wave velocity estimation system, for the learned model, the comparative example input data was given as input, and the comparative example output data was acquired. The comparative example input data is only the horizontal / vertical spectral ratio of ambient vibrations at the first position. Fig. 11 shows the results with the vertical axis being the comparative example output data (predicted AVS) and the horizontal axis being the observed AVS.

[0064] Regarding the example output data, for the average S-wave velocity (AVS10) from the ground surface to 10 m underground, the adjusted coefficient of determination R 2 was 0.7073. For the average S-wave velocity (AVS30) from the ground surface to 30 m underground, the adjusted coefficient of determination R 2 was 0.8552.

[0065] Regarding the comparative example output data, for the average S-wave velocity from the ground surface to 10 m underground, the adjusted coefficient of determination R 2 was 0.53. For the average S-wave velocity from the ground surface to 30 m underground, the adjusted coefficient of determination R 2 was 0.7684.

[0066] As described above, as a result of comparing the output data of the example and the output data of the comparative example, it became clear that the average S-wave velocity can be estimated with high accuracy.

[0067] Note that the prototyped S-wave velocity estimation system was able to instantaneously (within 0.1 seconds) acquire the output data of the example and the output data of the comparative example.

[0068] Although the present invention has been described based on the drawings and examples, it should be noted that those skilled in the art can easily make various modifications or corrections based on this disclosure. Therefore, it should be noted that these modifications and corrections are included in the scope of the present invention. For example, the functions included in each means, each step, etc. can be rearranged so as not to be logically contradictory, and a plurality of means or steps, etc. can be combined into one or divided.

[0069] Also, in the above-described embodiment, an example of the operation of the S-wave velocity estimation system 100 has been described with reference to FIG. 2. However, a configuration in which some steps included in the above-described operations, or some operations included in one step, are omitted within a range where there is no logical contradiction is also possible. Also, a configuration in which the order of a plurality of steps included in the above-described operations is changed within a range where there is no logical contradiction is also possible.

[0070] Also, in the above-described embodiment, although various means realized by the control unit 240 of the computer 200 have been described as a software configuration, at least some of the means may be a concept including software resources and / or hardware resources.

Explanation of Reference Numerals

[0071] 100: S-wave velocity estimation system 10: Acquisition means 20: Estimation processing means 30: Output means 200: Computer 210: Display unit 220: Operation unit 230: Storage unit 240: Control Unit 300: Neural Network Model 310: Input Layer 321: First Hidden Layer 322: Second Hidden Layer 330: Output Layer 1000: Ground

Claims

1. An acquisition means for acquiring the horizontal / vertical spectral ratio of the ambient vibration constantly occurring at the first position; An estimation processing means for obtaining the average S-wave velocity from the ground surface to 30 m underground at the first position by providing, as inputs, the horizontal / vertical spectral ratio of the ambient vibration constantly occurring at the first position and the microtopography classification at the first position, to a learned model obtained by performing machine learning for estimating the average S-wave velocity of the ground; An output means for outputting the obtained average S-wave velocity; Comprising: An S-wave velocity estimation system.

2. Coordinates including the elevation of the first position are further provided as inputs to the learned model, The S-wave velocity estimation system according to Claim 1.

3. The learned model is a neural network model, The S-wave velocity estimation system according to Claim 1 or 2.

4. On a computer, An acquisition function for accumulating the obtained horizontal / vertical spectral ratio of the ambient vibration and the microtopography classification at the position where the horizontal / vertical spectral ratio is obtained; An association function for accumulating the average S-wave velocity from the ground surface to 30 m underground in association with the horizontal / vertical spectral ratio and the microtopography classification; A machine learning function for performing machine learning on the relationship between the accumulated horizontal / vertical spectral ratio and microtopography classification and the average S-wave velocity from the ground surface to 30 m underground; An estimation function for estimating the average S-wave velocity from the ground surface to 30 m underground using the result of the machine learning when the horizontal / vertical spectral ratio and the microtopography classification are input; An output function for outputting the estimated average S-wave velocity from the ground surface to 30 m underground; A program for causing the above to be executed.

5. A step of storing a learned model obtained by performing machine learning for estimating the average S-wave velocity of the ground; A step of acquiring the horizontal / vertical spectral ratio of the ambient vibration constantly occurring at the first position; A step of obtaining the average S-wave velocity from the ground surface to 30 m underground at the first position by providing the horizontal / vertical spectral ratio of the ambient vibration constantly occurring at the first position and the microtopography classification at the first position to the learned model; A step of outputting the average S-wave velocity from the ground surface to 30 m underground; Including: An analysis method executed by a computer.

Images