Method for predicting an on-site earthquakes using artificial intelligence and seismic P-wave parameters

1 . A method for obtaining a seismic transverse wave prediction model, comprising the following steps of:

using at least one sensor to obtain a plurality of seismic historical data for a local position, wherein each of the seismic historical data includes seismic longitudinal wave information and a corresponding seismic transverse wave feature value, the seismic longitudinal wave information includes a plurality of data related to a seismic transverse wave, and the plurality of data consist of at least one of an absolute value of maximum acceleration, an absolute value of maximum velocity, an absolute value of maximum displacement, an equivalent period, an integration value of absolute acceleration and an integration value of velocity square; and

obtaining a seismic transverse wave prediction model for the local position via an artificial intelligence calculation by:

inputting the seismic historical data into the artificial intelligence calculation module, wherein the artificial intelligence calculation module comprises:

an input layer comprising six to eight first-order neurons and configured to receive the plurality of data;

a second-order hidden layer comprising 10 to 30 second-order neurons respectively connected to the six to eight first-order neurons;

a third-order hidden layer comprising 10 to 30 third-order neurons respectively connected to the 10 to 30 second-order neurons; and

an output layer comprising a fourth-order neuron connected to the 10 to 30 third-order neurons and configured to output a predicted feature value of real-time seismic transverse wave; and

for each of the second-order hidden layer, the third-order hidden layer and the output layer, calculating a signal summation value U i by each neuron i in a current layer from a previous layer according to:

U

i

=

∑

j

=

1

M

W

ij

⁢

V

j

+

I

i

,

wherein:

i is a serial number of a neuron in the current layer;

j is a serial number of a neuron in the previous layer;

Wij is a connection weighting between neuron i in the previous layer and neuron i in the current layer;

Ui is a sum of signals of neurons in the previous layer connected to Wij;

Vj is an activation output of neurons;

Ii is a random noise or an error correction constant; and

M is a total number of neurons in the previous layer,

calculating the activation output of each neuron according to:

V

i

=

f

⁡

(

U

i

)

=

1

1

+

exp

⁡

(

U

i

)

,

wherein the seismic transverse wave prediction model is configured to output the predicted feature value, and the predicted feature value is one of a maximum transverse wave surface acceleration (PGA), a maximum shear wave velocity value (PGV) and a maximum transverse wave displacement value, and is configured to be converted into an on-site seismic intensity according to a predetermined seismic intensity conversion relationship.

2 . The method according to claim 1 , wherein the artificial intelligence calculation module is a neural network algorithm module.

3 . The method according to claim 1 , further comprising:

obtaining a plurality of reference seismic historical data, each of which includes the seismic longitudinal wave information and the corresponding seismic transverse wave feature value; and

based on the plurality of seismic historical data and the plurality of reference seismic historical data, obtaining the seismic transverse wave prediction model for the local position via the artificial intelligence calculation module.

4 . The method according to claim 3 , wherein the seismic historical data are collected by a seismic measurement unit near the local position, and the reference seismic historical data are collected by a Taiwanese seismic measurement unit.

5 . The method according to claim 1 , further comprising:

using at least one further sensor to obtain a real-time seismic longitudinal wave information including the plurality of data related to the seismic transverse wave from a measurement position, wherein sources of the plurality of seismic historical data include or exclude a source of the real-time seismic longitudinal wave information.

6 . An apparatus including at least one sensor for predicting an on-site seismic intensity, comprising:

a signal pre-processing module, configured to obtain a real-time seismic longitudinal wave vertical acceleration data by using the at least one sensor, and to transform the real-time seismic longitudinal wave vertical acceleration data into a velocity value, a period and a displacement amount; and

a real-time prediction module, coupled to the signal pre-processing module, and configured to:

receive a plurality of data;

obtain a seismic transverse wave prediction model via a neural network algorithm module based on a plurality of seismic historical data obtained by using at least one further sensor, wherein the neural network algorithm module comprises:

an input layer comprising six to eight first-order neurons;

a second-order hidden layer comprising 10 to 30 second-order neurons respectively connected to the six to eight first-order neurons;

a third-order hidden layer comprising 10 to 30 third-order neurons respectively connected to the 10 to 30 second-order neurons; and

an output layer comprising a fourth-order neuron connected to the 10 to 30 third-order neurons and configured to output a predicted feature value of real-time seismic transverse wave;

only input six to eight seismic longitudinal wave related parameters selected from the plurality of data into the input layer of the seismic transverse wave prediction model;

obtain the predicted feature value of a real-time seismic transverse wave corresponding to the real-time seismic longitudinal wave vertical acceleration data at a measurement position by using the seismic transverse wave prediction model based on the real-time seismic longitudinal wave vertical acceleration data, the velocity value, the period and the displacement amount; and

convert the predicted feature value into the on-site seismic intensity according to a predetermined seismic intensity conversion relationship, wherein:

the plurality of data includes an absolute value of maximum acceleration, an absolute value of maximum velocity, an absolute value of maximum displacement, an equivalent period, an integration value of absolute acceleration and an integration value of velocity square; and

the predicted feature value includes one of a maximum transverse wave surface acceleration (PGA) value, a maximum shear wave velocity value (PGV) and a maximum transverse wave displacement value.

7 . The apparatus according to claim 6 , wherein the plurality of seismic historical data are collected by a local seismic measurement unit.

8 . A method for predicting an on-site seismic intensity, comprising the following steps of:

obtaining a real-time seismic longitudinal wave information from a measurement position by using at least one sensor, wherein the real-time seismic longitudinal wave information includes a plurality of data;

obtaining a seismic transverse wave prediction model via a neural network algorithm module based on a plurality of seismic historical data obtained by using at least one further sensor, wherein the neural network algorithm module comprises:

an input layer comprising six to eight first-order neurons;

a second-order hidden layer comprising 10 to 30 second-order neurons respectively connected to the six to eight first-order neurons;

a third-order hidden layer comprising 10 to 30 third-order neurons respectively connected to the 10 to 30 second-order neurons; and

an output layer comprising a fourth-order neuron connected to the 10 to 30 third-order neurons and configured to output a predicted feature value of real-time seismic transverse wave;

only inputting six to eight seismic longitudinal wave related parameters of the plurality of data into the input layer of the seismic transverse wave prediction model;

using the seismic transverse wave prediction model to obtain the predicted feature value of real-time seismic transverse wave corresponding to the real-time seismic longitudinal wave information at the measurement position based on the real-time seismic longitudinal wave information, wherein the predicted feature value includes one of a maximum transverse wave surface acceleration (PGA) value, a maximum shear wave velocity value (PGV) and a maximum transverse wave displacement value; and

converting the predicted feature value into the on-site seismic intensity according to a predetermined seismic intensity conversion relationship, wherein:

the plurality of data include an absolute value of maximum acceleration, an absolute value of maximum velocity, an absolute value of maximum displacement, an equivalent period, an integration value of absolute acceleration and an integration value of velocity square.

9 . The method according to claim 8 , wherein the seismic transverse wave prediction model is obtained further based on a plurality of reference seismic historical data in combination with the plurality of seismic historical data.

10 . The method according to claim 8 , wherein the predicted feature value includes a surface maximum velocity value.