Double-source strong earthquake signal identification method and device for high-rise building monitoring

Abstract

The invention relates to the technical field of vibration detection, in particular to a double-source strong earthquake signal identification method and device for high-rise building monitoring, and the method comprises the following steps: S1, preparing and installing a double-source strong earthquake instrument: enabling an MEMS accelerometer and a quartz accelerometer to be vertically arranged on a Z axis, and horizontally distributed on an X axis and a Y axis; s2, synchronously acquiring vibration signal components of the two accelerometers in three axial directions in real time and performing consistency judgment; s3, strong motion seismograph double-source signal feature extraction: starting from T0, recording data vectors of set time duration for three components of the MEMS accelerometer and the quartz accelerometer, and forming a recognition model input matrix; and S4, strong motion seismograph double-source signal type identification: forming a sample feature data set, optimizing the one-dimensional convolutional neural network, and constructing a double-source signal identification model. The method has the advantages that multiple types of sensors are integrated to recognize vibration signals, and recognition accuracy, reliability and anti-jamming capability are improved.

Description

Technical Field

[0001] The present invention relates to the field of vibration detection technology, in particular to a dual-source strong earthquake signal recognition method and a device thereof for high-rise building monitoring. Background Art

[0002] The frequency and intensity of global earthquake activities have put forward higher requirements for the accuracy and reliability of earthquake monitoring systems. At present, earthquake monitoring mainly relies on a single type of accelerometer, such as quartz accelerometer or MEMS accelerometer (also known as micro-electromechanical system accelerometer). These single sensors have some limitations, such as slow response time and sensitivity to noise, which makes it difficult to accurately distinguish between earthquake signals and external interference in complex environments, especially in the detection of strong earthquakes and micro earthquakes.

[0003] Quartz accelerometers are widely used in precision detection due to their high accuracy and low noise, but their long response time affects their performance in high-frequency dynamic detection. MEMS accelerometers are widely used in earthquake detection equipment due to their small size, economy and fast response. However, MEMS sensors are not as sensitive and anti-interference as quartz sensors, and are easily affected by environmental noise and artificial vibration, increasing the risk of false alarms.

[0004] Therefore, combining the high precision of quartz accelerometers and the fast response of MEMS accelerometers to develop more accurate earthquake monitoring equipment has become the key to solving the problems of existing earthquake monitoring systems. Especially in strong earthquake monitoring, distinguishing between real earthquake signals and interference from environmental noise or artificial vibrations and improving the reliability of data analysis are technical challenges that urgently need to be solved in the field of earthquake monitoring.

[0005] The shortcomings of existing earthquake monitoring systems include: relying on a single accelerometer, such as quartz or MEMS. The limitations of these sensors make it difficult to accurately distinguish between earthquake signals and external interference in complex environments, especially in strong earthquake and micro-earthquake detection. Quartz accelerometers have a long response time, which affects high-frequency dynamic detection; although MEMS accelerometers have a fast response, they lack sensitivity and anti-interference capabilities, are easily affected by environmental noise and man-made vibrations, and have a high false alarm rate. In strong earthquake monitoring, distinguishing between real earthquake signals and interference from environmental noise or man-made vibrations, and improving the reliability of data analysis are the core technical challenges facing the earthquake monitoring field.

[0006] Therefore, there is an urgent need to design a device that can well combine quartz acceleration sensors and MEMS acceleration sensors, integrate the detection characteristics and advantages of both, and improve the accuracy of vibration data detection and the anti-interference ability of the system. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a dual-source strong earthquake signal identification method and device for high-rise building monitoring, which is used to combine a high-precision quartz acceleration sensor and a high-response speed MEMS acceleration sensor to overcome the unnecessary interference data generated by the different working properties of the sensors, and to analyze the difference in vibration data between the two in real time through an intelligent algorithm, effectively distinguish between seismic signals and environmental noise or human interference, integrate the detection characteristics and advantages of the two, and improve the accuracy of vibration data detection and the anti-interference ability of the system.

[0008] In order to achieve the above purpose, a dual-source strong earthquake signal identification method for high-rise building monitoring is designed, including the following steps: S1. Preparation and installation of dual-source strong earthquake instrument: install the MEMS accelerometer box on a flat surface, ensure that the Z axis of the MEMS accelerometer is vertically upward, and the X and Y axes are horizontally distributed; install the quartz accelerometer box on the flat outer wall of the building in the same area, ensure that the Z axis of the quartz accelerometer is vertically upward, and the X and Y axes are horizontally distributed; S2. Dual-source signal acquisition and consistency analysis: real-time synchronous acquisition of the vibration signal components of the MEMS accelerometer and the quartz accelerometer in three axes. When the amplitude of the vertical component of the Z axis of one of the accelerometers exceeds the set threshold, the moment is marked as the trigger moment T0, and the dual-source signal consistency judgment step is entered. If the consistency passes the judgment, it enters step S3. If it is inconsistent, it is determined to be artificial interference or instrument failure. The dual-source signal consistency judgment step is as follows: S21. Starting from the trigger threshold moment T0, record the data of the MEMS accelerometer and the quartz accelerometer setting time; S22. Calculate the difference eigenvalues ​​between the sub-components of the X, Y and Z axis components of the above two accelerometers; S23. Establish a dual-source signal consistency evaluation model, input The difference feature value is used to obtain the feature judgment value; S24. After substituting the feature judgment value into the decision tree model, the consistency of the dual-source signal is judged according to the specified output label value; S3. Feature extraction of the strong earthquake instrument dual-source signal: starting from T0, the data vectors of the set time length are recorded for the three components of the MEMS accelerometer and the quartz accelerometer to form the recognition model input matrix. The signal feature extraction method is as follows: S31. Positive envelope area of ​​X, Y, and Z axes: Normalize the original vertical signals of the X, Y, and Z axes, calculate the positive envelope area, and use the serial number as the horizontal axis of the coordinate axis and the X, Y, and Z acceleration values ​​as the vertical axis. Axis, draw a plane scatter plot, connect adjacent positive points in sequence, and the entire range enclosed by these curves projected onto the coordinate axis is the area that needs to be calculated; S32. Calculate the negative envelope area of ​​the X, Y, and Z axes; S4. Strong motion detector dual-source signal type identification, as follows: S41. According to the eigenvalues ​​extracted in step S3, obtain the known seismic wave type label value to form a sample feature data set; S42. Use a decomposition-based multi-objective evolutionary algorithm to optimize the one-dimensional convolutional neural network, adjust the convolution kernel size, pooling method, and regularization strategy to build a dual-source signal recognition model.

[0009] Preferably, the method of the present invention also includes: step S1 also includes a MEMS accelerometer box installation verification method: observe whether the free ball on one side of the MEMS accelerometer box shell straightens the vertical rope and completely covers the spherical pattern on the back of the cavity where the ball is located. If it is not completely covered, it means that the MEMS accelerometer is not placed correctly, and a leveling system is used to place it on its base and adjust it until the free ball completely overlaps with the spherical pattern behind it.

[0010] Preferably, the method of the present invention also includes: in step S1, the quartz accelerometer box is installed on a flat outer wall of a building more than 1 meter from the ground, and the quartz accelerometer box installation verification method is as follows: observe whether the free ball on one side of the quartz accelerometer box shell straightens the vertical rope and completely covers the spherical pattern on the back of the cavity where the ball is located. If it is not completely covered, it means that the quartz accelerometer is not placed correctly and needs to be adjusted until the free ball and the vertical line and the spherical pattern and long scale line behind it completely coincide with each other.

[0011] Preferably, the method of the present invention also includes: the step S1 also includes: without destroying the condition that the Z axis of the aforementioned dual-source accelerometer is in an absolutely vertical direction, judging whether the shell surfaces of the quartz accelerometer box and the MEMS accelerometer box are parallel by using laser ranging units at the four corners of the bottom of the shell, thereby judging whether the X, Y directions of the three-axis quartz accelerometer and the X, Y component directions of the three-axis MEMS accelerometer are parallel in a one-to-one correspondence.

[0012] Preferably, the method of the present invention further comprises: the step S2 is specifically as follows: the difference feature value comprises: rank correlation coefficient, cosine similarity, KL divergence and JS divergence, and the calculation method of the difference feature is as follows;

[0013]

[0014] A represents any one of the three axes X, Y, and Z. i It represents the point value of the i-th data point on the A-axis component within 2s from the trigger threshold time T0 of the MEMS accelerometer. It represents the average value of the point value on the A-axis component of the MEMS accelerometer within 2s from the trigger threshold time T0, y i It represents the point value of the ith point on the A-axis component of the quartz accelerometer within 2s from the trigger threshold time T0. It represents the average value of the point value on the A-axis component of the quartz accelerometer within 2s from the trigger threshold time T0, CC A represents the rank correlation coefficient of the two accelerometers on the A-axis component, CS A represents the cosine similarity of the two accelerometers on the A-axis component, KL A represents the KL divergence of the two accelerometers on the A-axis component, JS A represents the JS divergence of the two accelerometers on the A-axis component, i represents the data point sequence number of each accelerometer signal involved in the calculation, i=1,2,3,...,N, where N is the accelerometer data length; the calculation method of the dual-source signal consistency evaluation model is as follows:

[0015]

[0016] Where E(Q,M,A) represents the average characteristic value of the two accelerometer signals in the A-axis direction, Q represents the quartz accelerometer, M represents the MEMS accelerometer, A represents any one of the three axes X, Y, and Z, CC A It represents the rank correlation coefficient of the two accelerometers on the A-axis component, CS A represents the cosine similarity of the two accelerometers on the A-axis component, KL A represents the KL divergence of the two accelerometers on the A-axis component, JS A Represent the JS divergence of the two accelerometers on the A-axis component; obtain the average eigenvalues ​​of the three groups of two accelerometers on the X, Y, and Z axes, respectively, and form the corresponding feature matrix; form a one-dimensional vector with the obtained calculation results of the dual-source signal consistency evaluation model, and group and label the formed one-dimensional vector to form a label vector; substitute the feature matrix and label vector into the CART decision tree algorithm to establish a dual-source signal consistency evaluation model.

[0017] Preferably, the method of the present invention further comprises: the step S3 is specifically as follows: the forward envelope area calculation method is as follows:

[0018]

[0019] in, It represents the closed irregular envelope area formed by the envelope composed of the positive data points of the A axis of the quartz accelerometer starting from the trigger threshold time T0 and lasting for 10s projected onto the horizontal axis of the two-dimensional plane, where the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; It represents the closed irregular envelope area formed by the envelope composed of the positive data points of the A axis of the MEMS accelerometer projected onto the horizontal axis of the two-dimensional plane starting from the trigger threshold time T0 and lasting for 10 seconds, wherein the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; It represents the arithmetic mean of the areas of the two aforementioned envelope regions; It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i+1th positive data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i-th positive data point on the A axis of the MEMS accelerometer. represents the absolute value of the amplitude of the i+1th positive data point on the A axis of the MEMS accelerometer, N1 represents the length of the new data vector consisting of the positive value of the quartz accelerometer signal starting from the trigger threshold time T0 and lasting for 10s, N2 represents the length of the new data vector consisting of the positive value of the MEMS accelerometer signal starting from the trigger threshold time T0 and lasting for 10s, i represents the data point sequence number; the negative envelope area calculation method is as follows:

[0020]

[0021] in, It indicates the closed irregular envelope area formed by the envelope composed of the negative data points of the A axis of the quartz accelerometer starting from the trigger threshold time T0 and lasting for 10s projected onto the horizontal axis of the two-dimensional plane, where the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; It represents the closed irregular envelope area formed by the envelope composed of the negative data points of the A axis of the MEMS accelerometer projected onto the horizontal axis of the two-dimensional plane starting from the trigger threshold time T0 and lasting for 10 seconds, wherein the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; represents the arithmetic mean of the areas of the two aforementioned envelope regions; It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i+1th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the MEMS accelerometer. It represents the absolute value of the amplitude of the i+1th negative data point on the A-axis of the MEMS accelerometer. N3 represents the length of the new data vector consisting of the negative value of the quartz accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. N4 represents the length of the new data vector consisting of the negative value of the MEMS accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. i represents the serial number.

[0022] The present invention also provides a dual-source strong earthquake signal identification device for high-rise building monitoring, comprising: a plurality of accelerometer boxes, one side of the accelerometer box is provided with a groove; a free ball is provided in the groove; a vertical groove and a ball groove are provided on one side surface of the cavity of the groove; a MEMS accelerometer or a quartz accelerometer is provided in the accelerometer box to form a MEMS accelerometer box or a quartz accelerometer box, the MEMS accelerometer box and the quartz accelerometer box are placed in the same area, the MEMS accelerometer box and the quartz accelerometer box are signal-connected to a data processing module, and the data processing module includes any one of the above-mentioned dual-source strong earthquake signal identification methods for high-rise building monitoring.

[0023] Preferably, the device of the present invention also includes: the free ball includes a vertical rope and a swinging ball, the upper end of the vertical rope is connected to the upper top surface of the groove cavity, and the lower end is connected to the swinging ball, and the swinging ball is suspended in the groove cavity by the vertical rope to swing freely; the vertical groove is arranged in the shape of the vertical rope, and the ball groove is arranged in the shape of the swinging ball; so that when the accelerometer box is in a horizontal state, the area projection of the free ball in the horizontal direction fully overlaps with the ball groove and the vertical groove.

[0024] Preferably, the device of the present invention further comprises: a plurality of laser distance measuring units are symmetrically arranged on the bottom of the accelerometer box, for determining whether the axes of the accelerometers in the plurality of accelerometer boxes are parallel.

[0025] Compared with the prior art, the present invention has the following advantages:

[0026] Compared with the vibration recognition model that only uses a single type of sensor, the present invention improves the recognition accuracy by integrating multiple types of sensors to identify vibration signals, and at the same time specifically solves the problems of high false trigger rate and inaccurate seismic wave recognition in the multi-sensor and multi-signal source vibration recognition model. First, the shell structure of the accelerometer sensor is optimized to ensure the accuracy of multiple types of sensors during the signal acquisition process. Then, a signal consistency judgment method is proposed to ensure that the strong earthquake monitoring system can still work normally under environmental interference and internal faults. In addition, a data set containing multiple positive and negative half-envelope areas is constructed, and the parameters of the one-dimensional convolutional neural network are optimized using a decomposition-based multi-objective evolutionary algorithm. Finally, we combine the optimized one-dimensional convolutional neural network with the data set to form an accurate and reliable dual-source signal recognition model, thereby improving the accuracy, reliability and anti-interference ability of vibration signal recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 , is a flow chart of dual-source signal identification in the method of the present invention;

[0028] Figure 2 , is a leveling flow chart of the dual-source strong seismograph system in the method of the present invention;

[0029] Figure 3 , is a schematic diagram of the structure of the device of the present invention;

[0030] Figure 4 , is a logical schematic diagram of a CART decision tree algorithm for predicting new characteristic values ​​of dual-source signals in the method of the present invention;

[0031] In the figure: 1 accelerometer box, 2 groove, 3 vertical rope, 4 swing ball, 5 vertical groove, 6 ball groove, 7 laser ranging unit. DETAILED DESCRIPTION

[0032] In order to make the purpose, principle and structure of the present invention more clear, it is further described below in conjunction with the drawings and specific embodiments.

[0033] Embodiment 1:

[0034] This embodiment provides a dual-source strong earthquake signal recognition method for high-rise building monitoring, the specific contents of which are as follows:

[0035] Step 1: Preparation and installation of dual-source strong motion seismometer.

[0036] The dual-source strong earthquake instrument includes a MEMS accelerometer box, a quartz accelerometer box and a data processing module.

[0037] See also Figure 2 First, place the MEMS accelerometer box on the flat ground in the same area to ensure that its Z component is the true vertical direction, and X and Y are distributed in the horizontal plane. The inspection method is: observe whether the vertical rope 3 of the free ball on one side of the MEMS accelerometer box shell is straightened, and observe whether the vertical rope 3 of the free ball, the projection of the swinging ball 4, and the vertical groove 5 and the ball groove 6 of the groove 2 of the accelerometer box 1 are completely covered. If not completely covered, it means that the MEMS accelerometer is not placed correctly, and it needs to be placed on its base with the help of a third-party leveling system for fine-tuning until the free ball and the vertical groove 5 and the ball groove 6 behind it completely overlap.

[0038] Then, the quartz accelerometer box in the dual-source strong seismometer is fitted and fixed on the flat wall of the building in the same area (the box should be more than 1 meter away from the ground), ensuring that its Z component is the true vertical direction, and X and Y are horizontal plane distributions. The inspection and adjustment method is: observe whether the free ball on one side of the quartz accelerometer box shell straightens the vertical rope, and observe whether the vertical rope 3 of the free ball, the projection of the swinging ball 4, and the pattern of the vertical groove 5 and the ball groove 6 of the groove 2 of the accelerometer box 1 are completely covered. If not completely covered, it means that the quartz accelerometer is not placed correctly, and fine-tuning is required until the vertical rope 3 of the free ball, the projection of the swinging ball 4, and the pattern of the vertical groove 5 and the ball groove 6 of the groove 2 of the accelerometer box 1 are completely covered.

[0039] Finally, without destroying the Z-axis of the dual-source accelerometer composed of the aforementioned MEMS accelerometer and quartz accelerometer in an absolutely vertical direction, the laser ranging units on the four corners of the bottom of the accelerometer case are used to determine whether the X, Y directions of the three-axis quartz accelerometer and the X, Y component directions of the three-axis MEMS accelerometer are parallel one by one (that is, the bottom edges of the shell remain parallel one by one), that is, the X-axis is parallel to the X-axis, and the Y-axis is parallel to the Y-axis. The specific operation method is as follows: (1) keep the accelerometer box arranged on the wall of the building still, and slightly rotate the accelerometer box set on the flat ground so that the side with the XYZ mark is basically flush with the outside of the accelerometer box with the XYZ mark set on the wall of the building; (2) operate the computer to make the laser measuring units 7 at the two corners of the bottom of the accelerometer box set on the flat ground simultaneously emit laser signals to the laser measuring units 7 at the two corners of the bottom of the accelerometer box set on the wall of the building, so as to measure whether the multi-point distances of the bottom edge lines of the two accelerometer box outer shell surfaces are equal (or whether the distance measurement difference is 0). If they are equal, it means that the two outer shell surfaces are flush. Otherwise, the two surfaces need to be manually adjusted according to the measurement value.

[0040] Step 2: Dual-source signal acquisition and consistency analysis

[0041] See also Figure 1 .

[0042] (1) Real-time synchronous acquisition of a total of 6 ground vibration signal components of the above two accelerometers ( q represents a quartz accelerometer, m represents a MEMS accelerometer, x, y, and z represent three axes, and a represents a signal component). When the vertical component amplitude (also known as the Z axis) of one of the accelerometers is or When the threshold a is exceeded (the threshold triggering moment is recorded as T0), the dual-source signal consistency judgment step is entered. If they are consistent, the dual-source signal type identification step is entered. Otherwise, it is considered that there is a dual-source signal inconsistency problem caused by artificial interference or instrument failure.

[0043] (2) First, taking the vertical Z component as an example, data of 2s length is recorded from the trigger threshold time T0 (the sampling rate is set to 2000Hz, or other values ​​between 200-4000Hz); since the accelerometer boxes of the two accelerometers are not installed in the same place, the instrument output signals will be different, but the signal difference will show a similar waveform fluctuation development trend, and the similar waveform fluctuation development trend is different only in terms of amplitude and local minimum value. The difference eigenvalues ​​between the three subcomponents of the vertical Z component of the two accelerometers are calculated, and the difference between the three-component outputs of the two accelerometers is described by four eigenvalues, namely, Spearman Coefficient of Rank Correlation (hereinafter referred to as CC), Cosine Similarity (hereinafter referred to as CS), KL Divergence (hereinafter referred to as KL), and JS Divergence (hereinafter referred to as JS). They respectively represent the waveform shape similarity, angle, statistical error and distribution difference between the two groups of vectors, and their value ranges are [-1, 1], [-1, 1], [0, + ∞], and [0, 1]. The eigenvalue calculation formula is as follows:

[0044]

[0045] A represents any one of the three axes X, Y, and Z. i It represents the point value of the i-th data point on the A-axis component within 2s from the trigger threshold time T0 of the MEMS accelerometer. It represents the average value of the point value on the A-axis component of the MEMS accelerometer within 2s from the trigger threshold time T0, y i It represents the point value of the ith point on the A-axis component of the quartz accelerometer within 2s from the trigger threshold time T0. It represents the average value of the point value on the A-axis component of the quartz accelerometer within 2s from the trigger threshold time T0, CC A represents the rank correlation coefficient of the two accelerometers on the A-axis component, CS A represents the cosine similarity of the two accelerometers on the A-axis component, KL A represents the KL divergence of the two accelerometers on the A-axis component, JS A represents the JS divergence of the two accelerometers on the A-axis component, i represents the data point sequence number of each accelerometer signal involved in the calculation, i = 1, 2, 3, ..., N, where N is the accelerometer data length;

[0046] Substituting Z into A, we can obtain the characteristic values ​​of the two accelerometers on the Z-axis component: CC Z , CS Z , KL Z JS Z Similarly, the characteristic values ​​of the two accelerometers of the X and Y axis components can be obtained: CC X , CS X , KL X ,JS X 、CC Y , CS Y , KL Y ,JS Y .

[0047] (3) The dual-source signal consistency evaluation model is established based on the eigenvalues ​​of the above components:

[0048]

[0049] Where E(Q,M,A) represents the average characteristic value of the two accelerometer signals in the A-axis direction, Q represents the quartz accelerometer, M represents the MEMS accelerometer, A represents any one of the three axes X, Y, and Z, CC A It represents the rank correlation coefficient of the two accelerometers on the A-axis component, CS A represents the cosine similarity of the two accelerometers on the A-axis component, KL A represents the KL divergence of the two accelerometers on the A-axis component, JS A It represents the JS divergence of the two accelerometers on the A-axis component. After substituting it into the calculation, we can get E(Q,M,X), E(Q,M,Y), and E(Q,M,Z).

[0050] After substituting E(Q,M,X), E(Q,M,Y) and E(Q,M,Z) into the decision tree model, the consistency of the dual source signals is determined according to the specified output label values.

[0051] See also Figure 4The specific process is: collect 50 natural earthquake event signals, 50 artificial blasting event signals, 50 accelerometer installation abnormality signals or sensor failure signals (the signal can be obtained by installation orientation errors and hammering buildings), calculate a total of 150 sets of E(Q,M,X), E(Q,M,Y), E(Q,M,Z) eigenvalues ​​to form a 150×3 feature matrix; design a one-dimensional cell vector consisting of 150 rows of label values, set the label values ​​of the 1st to 100th rows to "Y", and the label values ​​of the 101st to 150th rows to "N"; substitute the set of eigenvalues ​​and the one-dimensional vector with label values ​​into the CART decision tree algorithm to establish a trained model. When predicting the new eigenvalue of the dual-source signal, if it is "Y", it is determined that the signal output consistency of the current dual-source system is high. If it is "N", it is determined that the current system layout is abnormal or the sensor is faulty, and it will not enter the subsequent strong vibration seismic wave identification link. The CART decision tree algorithm, taking the eigenvalue set judgment method on the Z-axis component as an example, can be set as follows: in the first comparison link, when E(Q,M,Z)<4.75, output "Y", otherwise enter the next comparison link; when E(Q,M,Z)≥5.05, output "N", otherwise enter the next comparison link; when E(Q,M,X) is greater than or equal to 6.5, output "Y", otherwise enter the next comparison link; when E(Q,M,Y)<3.1, output "N", otherwise output "Y".

[0052] Step 3: Strong motion detector dual-source signal feature extraction process

[0053] When step 2 determines that the signal output consistency of the dual-source system is high, the strong vibration signal identification link is entered to identify strong vibration signals, artificial blasting signals, and other event signals (mechanical construction, human activities, etc.).

[0054] Specifically, the process of extracting signal features is as follows:

[0055] (1) Positive envelope area of ​​the X, Y, and Z axes: Normalize the original signal of a certain axis and calculate the positive envelope area. That is, use the serial number as the horizontal axis and the axial acceleration value of the axis as the vertical axis to draw a plane scatter plot. Connect adjacent positive points in sequence. The entire range enclosed by these curves projected onto the coordinate axis is the area to be calculated.

[0056] The forward envelope area is calculated as follows:

[0057]

[0058] in, It represents the closed irregular envelope area formed by the envelope line composed of the positive data points of the quartz accelerometer A axis projected onto the horizontal axis of the two-dimensional plane starting from the trigger threshold time T0 and lasting for 10s, wherein the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis.

[0059] It represents the closed irregular envelope area formed by the envelope composed of the positive data points of the A-axis of the MEMS accelerometer which starts from the trigger threshold time T0 and lasts for 10s and is projected onto the horizontal axis of the two-dimensional plane, where the A-axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis.

[0060] It represents the arithmetic mean of the areas of the two envelope regions mentioned above.

[0061] It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i+1th positive data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i-th positive data point on the A axis of the MEMS accelerometer. It represents the absolute value of the amplitude of the i+1th positive data point on the A-axis of the MEMS accelerometer. N1 represents the length of the new data vector composed of the positive value of the quartz accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. N2 represents the length of the new data vector composed of the positive value of the MEMS accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. i represents the data point sequence number.

[0062] (2) Negative envelope area of ​​X, Y, and Z axes:

[0063] The negative envelope area is calculated as follows:

[0064]

[0065] in, It indicates the closed irregular envelope area formed by the envelope composed of the negative data points of the A-axis of the quartz accelerometer starting from the trigger threshold time T0 and lasting for 10s, projected onto the horizontal axis of the two-dimensional plane, where the A-axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis.

[0066] It represents the closed irregular envelope area formed by the envelope composed of the negative data points of the A-axis of the MEMS accelerometer which starts from the trigger threshold time T0 and lasts for 10s and is projected onto the horizontal axis of the two-dimensional plane, where the A-axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis.

[0067] It represents the arithmetic mean of the areas of the two envelope regions mentioned above.

[0068] It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i+1th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the MEMS accelerometer. It represents the absolute value of the amplitude of the i+1th negative data point on the A-axis of the MEMS accelerometer. N3 represents the length of the new data vector consisting of the negative value of the quartz accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. N4 represents the length of the new data vector consisting of the negative value of the MEMS accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. i represents the serial number.

[0069] Similarly, calculate the positive envelope area of ​​the X-axis, the negative envelope area of ​​the X-axis, the positive envelope area of ​​the Y-axis, and the negative envelope area of ​​the Y-axis.

[0070] Step 4: Strong motion detector dual-source signal type identification process

[0071] (1) Calculate the above-mentioned characteristic parameters to form a 6-dimensional × 1000-row sample feature data set with 1000 known seismic wave type label values ​​(including 600 rows of natural earthquake event signals, 300 artificial blasting event signals, and 100 environmental noise signals or artificial construction signals), and randomly split it into a training set, a validation set, and a test set in a ratio of 7:1:2;

[0072] (2) Then, based on the decomposition multi-objective evolutionary algorithm (MOEA / D), the hyperparameter optimization of the neural network parameters such as the convolution kernel size, pooling operation method (maximum pooling or average pooling), and regularization method (L1 or L2) of the existing one-dimensional convolutional neural network (1D-CNN) is performed in order to obtain the best parameters to replace the original default parameters or default methods in the model, thereby forming a new prediction model - MOEA / D-ID-CNN. The new model then identifies the type of unknown signals.

[0073] (3) The core of multi-objective optimization is to construct multiple objective functions. Here we construct an objective function with the training set as the sample, the above three hyperparameters (convolution kernel size, pooling operation method, and regularization method) as independent variables, and the prediction effect as the dependent variable. The prediction effect is evaluated by AUC value (area under the ROC curve), mean absolute error (MAE), mean bias error (MBE), and determination coefficient (R2).

[0074] Since optimization is to find the minimum value of the objective function, the above evaluation indicators need to be transformed. The final form of the solution model of MOEA / D multi-objective optimization problem is:

[0075] minF = {f1,f2,f3,f4};

[0076]

[0077] In the above formula, f1 is the mean bias error (MBE), which can determine whether the prediction model has positive or negative bias; f2 is the mean absolute error (MAE); f3 is the calculation formula of 1 minus the coefficient of determination (R2 score), which is used to evaluate the degree of explanation of the independent variable on the dependent variable in the regression analysis; f4 is the calculation formula of 1 minus the AUC value; y i The predicted values ​​identified for the model; is the actual value; is the average value of the predicted value set; n is the total number of samples involved in the regression evaluation calculation; M is the number of points on the ROC curve.

[0078] Embodiment 2:

[0079] This embodiment provides a dual-source strong earthquake signal recognition device for high-rise building monitoring. Figure 3 . It includes: a plurality of accelerometer boxes 1, one side of the accelerometer box 1 is provided with a groove 2; a free ball is arranged in the groove 2; a vertical groove 5 and a ball groove 6 are arranged on one side of the cavity of the groove 2; a MEMS accelerometer or a quartz accelerometer is arranged in the accelerometer box 1 to form a MEMS accelerometer box or a quartz accelerometer box, the MEMS accelerometer box and the quartz accelerometer box are placed in the same area, the MEMS accelerometer box and the quartz accelerometer box are connected to the data processing module signal, and the data processing module is the dual-source strong earthquake signal recognition method for high-rise building monitoring described in Example 1.

[0080] The free ball includes a vertical rope 3 and a swinging ball 4. The upper end of the vertical rope 3 is connected to the upper top surface of the cavity of the groove 2, and the lower end is connected to the swinging ball 4. The swinging ball 4 is suspended in the cavity of the groove 2 by the vertical rope 3 and swings freely. The vertical groove 5 is arranged in the same shape as the vertical rope 3, and the ball groove 6 is arranged in the same shape as the swinging ball 4. When the accelerometer box 1 is in a horizontal state, the area projection of the free ball in the horizontal direction fully overlaps with the ball groove 6 and the vertical groove 5. A plurality of laser ranging units 7 are symmetrically arranged on the bottom of the accelerometer box 1 to determine whether the axes of the accelerometers in the plurality of accelerometer boxes 1 are parallel.

[0081] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes according to the technical solutions and novel concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A dual-source strong earthquake signal identification method for high-rise building monitoring, characterized in that: The steps include: S1. Preparation and installation of dual-source strong seismometer: Install the MEMS accelerometer box on a flat surface, ensuring that the Z axis of the MEMS accelerometer is vertically upward and the X and Y axes are horizontally distributed; install the quartz accelerometer box on the flat exterior wall of the building in the same area, ensuring that the Z axis of the quartz accelerometer is vertically upward and the X and Y axes are horizontally distributed; S2. Dual-source signal acquisition and consistency analysis: Real-time synchronous collection of vibration signal components of the three axes of the MEMS accelerometer and the quartz accelerometer. When the amplitude of the vertical component of the Z axis of one of the accelerometers exceeds the set threshold, the moment is marked as the trigger moment T0, and the dual-source signal consistency judgment step is entered. If the consistency passes the judgment, the step S3 is entered. If not, it is determined to be artificial interference or instrument failure. The steps of determining the consistency of the dual-source signals are as follows: S21. Starting from the trigger threshold time T0, record the MEMS accelerometer and quartz accelerometer set time data; S22. Calculate the difference eigenvalues ​​between the subcomponents of the X, Y and Z axis components of the two accelerometers; S23. Establish a dual-source signal consistency evaluation model, input the difference feature value to obtain the feature judgment value; S24. Substituting the feature criterion value into the decision tree model, and judging the consistency of the dual-source signal according to the specified output label value; S3. Extraction of dual-source signal features of strong earthquake instrument: Starting from T0, the data vectors of the three components of MEMS accelerometer and quartz accelerometer are recorded for a set time length to form the input matrix of the recognition model. The signal feature extraction method is as follows: S31. Positive envelope area of ​​X, Y, and Z axes: Normalize the original vertical signals of X, Y, and Z axes, calculate the positive envelope area, use the serial number as the horizontal axis of the coordinate axis, and the acceleration value of X, Y, and Z as the vertical axis, draw a plane scatter plot, connect adjacent positive points in sequence, and the entire range enclosed by these curves projected onto the coordinate axis is the area to be calculated; S32. Calculate the negative envelope area of ​​the X, Y, and Z axes; S4. Identification of the type of strong seismograph dual-source signal, as follows: S41. Obtaining known seismic wave type label values ​​according to the characteristic values ​​extracted in step S3 to form a sample characteristic data set; S42. A decomposition-based multi-objective evolutionary algorithm is used to optimize the one-dimensional convolutional neural network, adjust the convolution kernel size, pooling method and regularization strategy to build a dual-source signal recognition model.

2. A dual-source strong earthquake signal identification method for high-rise building monitoring as claimed in claim 1, characterized in that: The step S1 also includes a method for verifying the installation of the MEMS accelerometer box: observing whether the free ball on one side of the MEMS accelerometer box shell straightens the vertical rope and completely covers the spherical pattern on the back of the cavity where the ball is located. If it is not completely covered, it means that the MEMS accelerometer is not placed correctly. A leveling system is used to place it on its base and then adjust it until the free ball completely overlaps with the spherical pattern behind it.

3. A dual-source strong earthquake signal identification method for high-rise building monitoring as claimed in claim 1, characterized in that: In step S1, the quartz accelerometer box is installed on a flat outer wall of a building more than 1 meter from the ground. The installation verification method of the quartz accelerometer box is as follows: observe whether the free ball on one side of the quartz accelerometer box shell straightens the vertical rope and completely covers the spherical pattern on the back of the cavity where the ball is located. If it is not completely covered, it means that the quartz accelerometer is not placed correctly and needs to be adjusted until the free ball and the vertical line and the spherical pattern and long scale line behind it completely coincide with each other.

4. A dual-source strong earthquake signal identification method for high-rise building monitoring as claimed in claim 1, characterized in that: The step S1 also includes: without destroying the condition that the Z axis of the dual-source accelerometer is in an absolutely vertical direction, judging whether the shell surfaces of the quartz accelerometer box and the MEMS accelerometer box are parallel by using the laser ranging units at the four corners of the bottom of the shell, thereby judging whether the X, Y directions of the three-axis quartz accelerometer and the X, Y component directions of the three-axis MEMS accelerometer are parallel one by one.

5. A dual-source strong earthquake signal identification method for high-rise building monitoring as claimed in claim 1, characterized in that: The step S2 is specifically as follows: The difference feature values ​​include: rank correlation coefficient, cosine similarity, KL divergence and JS divergence. The calculation method of the difference feature is as follows: A represents any one of the three axes X, Y, and Z. i It represents the point value of the i-th data point on the A-axis component within 2s from the trigger threshold time T0 of the MEMS accelerometer. It represents the average value of the point value on the A-axis component of the MEMS accelerometer within 2s from the trigger threshold time T0, y i It represents the point value of the ith point on the A-axis component of the quartz accelerometer within 2s from the trigger threshold time T0. It represents the average value of the point value on the A-axis component of the quartz accelerometer within 2s from the trigger threshold time T0, CC A represents the rank correlation coefficient of the two accelerometers on the A-axis component, CS A represents the cosine similarity of the two accelerometers on the A-axis component, KL A represents the KL divergence of the two accelerometers on the A-axis component, JS A represents the JS divergence of the two accelerometers on the A-axis component, i represents the data point sequence number of each accelerometer signal involved in the calculation, i = 1, 2, 3, ..., N, where N is the accelerometer data length; The calculation method of the dual-source signal consistency evaluation model is as follows: Where E(Q,M,A) represents the average characteristic value of the two accelerometer signals in the A-axis direction, Q represents the quartz accelerometer, M represents the MEMS accelerometer, A represents any one of the three axes X, Y, and Z, CC A It represents the rank correlation coefficient of the two accelerometers on the A-axis component, CS A represents the cosine similarity of the two accelerometers on the A-axis component, KL A represents the KL divergence of the two accelerometers on the A-axis component, JS A It represents the JS divergence of the two accelerometers on the A-axis component; The average eigenvalues ​​of the three groups of two accelerometers on the X, Y, and Z axes are obtained, and the corresponding characteristic matrices are formed; the calculation results of the dual-source signal consistency evaluation model are formed into a one-dimensional vector, and the formed one-dimensional vectors are grouped and labeled to form a label vector; Substitute the feature matrix and label vector into the CART decision tree algorithm to establish a dual-source signal consistency evaluation model.

6. A dual-source strong earthquake signal identification method for high-rise building monitoring as claimed in claim 1, characterized in that: The step S3 is specifically as follows: The forward envelope area is calculated as follows: in, It represents the closed irregular envelope area formed by the envelope composed of the positive data points of the A axis of the quartz accelerometer starting from the trigger threshold time T0 and lasting for 10s projected onto the horizontal axis of the two-dimensional plane, where the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; It represents the closed irregular envelope area formed by the envelope composed of the positive data points of the A axis of the MEMS accelerometer projected onto the horizontal axis of the two-dimensional plane starting from the trigger threshold time T0 and lasting for 10 seconds, wherein the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; represents the arithmetic mean of the areas of the two aforementioned envelope regions; It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i+1th positive data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i-th positive data point on the A axis of the MEMS accelerometer. represents the absolute value of the amplitude of the i+1th positive data point on the A axis of the MEMS accelerometer, N1 represents the length of the new data vector consisting of the positive value of the quartz accelerometer signal starting from the trigger threshold time T0 and lasting for 10s, N2 represents the length of the new data vector consisting of the positive value of the MEMS accelerometer signal starting from the trigger threshold time T0 and lasting for 10s, and i represents the data point sequence number; The negative envelope area is calculated as follows: in, It indicates the closed irregular envelope area formed by the envelope composed of the negative data points of the A axis of the quartz accelerometer starting from the trigger threshold time T0 and lasting for 10s projected onto the horizontal axis of the two-dimensional plane, where the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; It represents the closed irregular envelope area formed by the envelope composed of the negative data points of the A axis of the MEMS accelerometer projected onto the horizontal axis of the two-dimensional plane starting from the trigger threshold time T0 and lasting for 10 seconds, wherein the A axis is any one of the X, Y, and Z axes, and the two-dimensional plane is composed of the amplitude of the point as the vertical axis and the serial number of the point as the horizontal axis; represents the arithmetic mean of the areas of the two aforementioned envelope regions; It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i+1th negative data point on the A axis of the quartz accelerometer. It represents the absolute value of the amplitude of the i-th negative data point on the A axis of the MEMS accelerometer. It represents the absolute value of the amplitude of the i+1th negative data point on the A-axis of the MEMS accelerometer. N3 represents the length of the new data vector consisting of the negative value of the quartz accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. N4 represents the length of the new data vector consisting of the negative value of the MEMS accelerometer signal starting from the trigger threshold time T0 and lasting for 10s. i represents the serial number.

7. A dual-source strong earthquake signal identification device for high-rise building monitoring, characterized in that: include: A plurality of accelerometer boxes, each having a groove formed on one side thereof; A free ball is arranged in the groove; a vertical groove and a ball groove are arranged on one side of the cavity of the groove; A MEMS accelerometer or a quartz accelerometer is arranged in the accelerometer box to form a MEMS accelerometer box or a quartz accelerometer box. The MEMS accelerometer box and the quartz accelerometer box are placed in the same area. The MEMS accelerometer box and the quartz accelerometer box are signal-connected to a data processing module. The data processing module includes a dual-source strong earthquake signal identification method for high-rise building monitoring as described in any one of claims 1-6.

8. A dual-source strong earthquake signal identification device for high-rise building monitoring as claimed in claim 7, characterized in that: Also includes: The free ball comprises a vertical rope and a swinging ball, the upper end of the vertical rope is connected to the upper top surface of the groove cavity, and the lower end is connected to the swinging ball, and the swinging ball is suspended in the groove cavity by the vertical rope and swings freely; The vertical groove is arranged in the same shape as the vertical rope, and the small ball groove is arranged in the same shape as the swinging small ball; When the accelerometer box is in a horizontal state, the area projection of the free ball in the horizontal direction fully overlaps with the ball groove and the vertical groove.

9. A dual-source strong earthquake signal identification device for high-rise building monitoring as claimed in claim 7, characterized in that: A plurality of laser distance measuring units are symmetrically arranged on the bottom of the accelerometer box, and are used to determine whether the axes of the accelerometers in the accelerometer boxes are parallel.

Images