Elevator motor non-contact vibration detection device and method based on millimeter wave radar

Abstract

The invention discloses an elevator motor non-contact vibration detection device and method based on a millimeter-wave radar. The millimeter-wave radar comprises a transmitting antenna, a receiving antenna, a radio frequency assembly, a digital-to-analog converter and a digital signal processor. The millimeter wave radar is installed in the radial direction of the corresponding mine hoist. The method comprises the steps that millimeter wave radars are installed at the installation position of the mine hoist, the angle of each millimeter wave radar transmitting antenna is adjusted, and it is ensured that wireless signal energy transmitted by the millimeter wave radars is focused on a head sheave and a motor of the corresponding mine hoist; processing the intermediate frequency signal to obtain an enhanced target signal; determining whether a head sheave swing signal exists in the signals or not; judging whether the head sheave shaking amplitude is within a normal range or not, and deducing an abnormal reason; whether the motor vibrates or not is analyzed, and the reason of motor vibration is deduced; and when the elevator is abnormal, early warning is carried out, and the elevator stops working. According to the device and the method, the abnormal condition of the elevator can be quickly and efficiently detected.

Description

Technical Field

[0001] The present invention belongs to the technical field of coal mine production safety, and in particular relates to a non-contact vibration detection device and method for a hoist motor based on millimeter wave radar. Background Art

[0002] The basic national conditions of my country's energy are still dominated by traditional energy. Coal, as an important component of traditional energy, is still the current basic energy and important industrial raw material. In order to effectively ensure the output of coal resources, more attention needs to be paid to safety issues in the coal mining process. At present, the safety issues of coal are mainly reflected in the following two aspects: on the one hand, the underground geological environment of coal mines is complex and the mining process is complicated. It is easy to encounter dangerous working conditions such as roof crushing during the mining process. On the other hand, under the influence of harsh factors such as high temperature, dust, and gas in coal mines, it is easy to aggravate the aging of coal mining equipment. At the same time, the harsh environment is easy to interfere with the reliable monitoring of the working status of mining equipment, resulting in the inability to monitor the dangerous state in a timely and effective manner, which will greatly increase the risk of coal mine accidents. As an important coal mining equipment, the mine hoist plays an extremely important role in the transportation of personnel and materials in the mine. Because the underground conditions and working environment are extremely harsh compared to those on the ground, it will have an adverse effect on the operation and monitoring process of the hoist. In order to effectively ensure the safe and reliable transportation of personnel and materials, it is necessary to ensure the safe operation of the mine hoist.

[0003] At present, computer-based vision technology is mainly used to detect the operating status of mine hoists. The video stream of the sheave is collected at an ultra-high frame rate, and the collected images are subjected to image denoising based on the multi-scale Retinex algorithm with color gain weighting. This method requires the installation of multiple cameras in the underground tunnel of the coal mine along the length direction at a set interval. Therefore, a large number of cameras and video transmission equipment need to be installed in the early stage, with high initial investment and maintenance costs, large construction volume, and a long construction period. In addition, the camera cannot collect accurate image data under the conditions of obstruction and dust, and the accuracy of long-distance monitoring will drop sharply. At the same time, the image denoising process is cumbersome and has high requirements for detection equipment. Moreover, there is also the problem of personnel privacy protection.

[0004] In order to effectively monitor the mine hoist under the conditions of the mine, researchers in the industry have begun to explore detection technology based on infinite perception. Among the many radio technologies, new solutions based on millimeter waves have gradually attracted attention due to their many advantages such as easy installation, low cost, strong adaptability, strong anti-interference ability and strong penetration, and have become an ideal choice for wireless perception detection technology in underground environments. However, wireless perception detection technology based on millimeter-wave radar signals still faces many challenges in complex underground environments, such as poor wireless perception accuracy in the presence of dynamic interference; single features extracted from a single channel cannot accurately reflect the working status of the mine hoist; when the radar simultaneously monitors the characteristics of multiple working parts of the hoist, how to effectively separate the collected characteristic signals and how to fuse and analyze multiple characteristic signals are still technical problems that need to be solved in the industry. Summary of the invention

[0005] In view of the problems existing in the above-mentioned prior art, the present invention provides a non-contact vibration detection device and method for a hoist motor based on millimeter-wave radar. The device has a simple structure and low investment cost, and can quickly and efficiently detect abnormal conditions of the hoist in a non-contact manner; the method can accurately and efficiently determine whether the working status of the mine hoist is normal, and is particularly suitable for detection operations in complex mines with a lot of dust and obstructions.

[0006] In order to achieve the above-mentioned invention object, the present invention provides a non-contact vibration detection device for a hoist motor based on a millimeter wave radar, comprising a mine hoist, a millimeter wave radar, a data acquisition module and a data processing terminal;

[0007] The millimeter wave radar includes a transmitting antenna, a receiving antenna, a radio frequency component, a digital-to-analog converter and a digital signal processor;

[0008] The transmitting antenna is used to convert the received outgoing electrical signal into an electromagnetic wave signal and transmit it in a specific direction;

[0009] The receiving antenna is used to convert the received reflected electromagnetic waves into reflected electrical signals and send them to the radio frequency component;

[0010] The radio frequency component includes a synthesizer and a mixer. The mixer is used to synthesize the received reflected electrical signal and the transmitted electrical signal into an intermediate frequency signal and send it to the digital-to-analog converter; the synthesizer is used to generate a linear frequency modulated transmitted electrical signal from the received single frequency transmitted electrical signal and then send it to the transmitting antenna and the mixer;

[0011] The digital-to-analog converter is used to perform analog-to-digital conversion on the received intermediate frequency signal and then send it to the digital signal processor, which is used to perform analog-to-digital conversion on the received transmission electrical signal and then send it to the radio frequency component;

[0012] The digital signal processor is used to send a transmission signal to the digital-to-analog converter according to a set frequency, and is used to process the received intermediate frequency signal;

[0013] The number of the millimeter wave radars is multiple and the number is the same as the number of mine hoists. The millimeter wave radars are installed in the radial direction of the corresponding mine hoists, and the wireless signal energy emitted by the millimeter wave radars is focused on the sheave and motor of the corresponding mine hoist.

[0014] The data processing terminal is connected to a plurality of millimeter radars respectively through data acquisition modules.

[0015] The present invention has a simple structure and low investment cost. The millimeter wave radar is installed in the radial direction of the mine hoist, and the swing displacement of the sheave and the vibration displacement of the motor can be obtained by calculating the phase change of the millimeter wave radar signal. Therefore, the swing of the sheave and the vibration of the motor can be sensed by processing the transmission signal and the reflection signal, and the abnormality of the hoist can be detected quickly and efficiently in a non-contact manner. The data processing terminal is connected to multiple millimeter wave radars through a data acquisition module, and the data processing terminal can be used to receive the signals collected by multiple millimeter wave radars and the alarm signals sent, which is conducive to the unified recording and monitoring of the detection conditions of multiple mine hoists, so that effective treatment measures can be taken for abnormal hoists in a timely manner.

[0016] The present invention also provides a non-contact vibration detection method for a hoist motor based on a millimeter wave radar, which uses a non-contact vibration detection device for a hoist motor based on a millimeter wave radar, and includes the following steps:

[0017] Step 1: Conduct on-site surveys of each mine hoist in the coal mine, and determine the position with good signal detection effect in the radial direction of the mine hoist as the installation position of the corresponding millimeter wave radar; mark the installation position, and record the specific coordinates and environmental description of the installation position;

[0018] Step 2: Install millimeter-wave radars at the installation positions of each mine hoist in the coal mine, and adjust the angle of each millimeter-wave radar transmitting antenna to ensure that the wireless signal energy emitted by the millimeter-wave radar is focused on the sheave and motor of the corresponding mine hoist. At the same time, adjust the angle of each millimeter-wave radar receiving antenna to ensure that the reflected wireless signal energy is focused on the location of the receiving antenna;

[0019] Step 3: Set the working parameters and sampling parameters of each millimeter-wave radar, and start the millimeter-wave radar. During the operation of the mine hoist, use the synthesizer in the millimeter-wave radar to generate a linear frequency modulated continuous wave, and transmit it to the target part through the transmitting antenna. After being reflected by the target part and processed by the mixer, an intermediate frequency signal is generated, and the intermediate frequency signal is received and recorded in real time by the digital signal processor.

[0020] Step 4: The received intermediate frequency signal is processed in chronological order by a digital signal processor. The processing process is as follows:

[0021] S1: De-noise the intermediate frequency signal and enhance the target signal strength to obtain an enhanced target signal;

[0022] S2: After receiving the enhanced signal, the digital signal processor first determines whether there is a sheave swing signal. If there is no sheave swing signal, S1 is executed again. If there is a sheave swing signal, the reflection source of the signal is first determined. If the signal is a reflection signal from other strong stationary targets in the underground tunnel of the coal mine, the average cancellation algorithm is used to filter out the stationary clutter. If the signal comes from other dynamic interference signals, the differential judgment algorithm is used to filter the signal.

[0023] S3: First, obtain the sky wheel swing signal through filtering, and then preliminarily process the sky wheel swing signal to obtain the distance from the sky wheel to the millimeter wave radar and the phase change of the sky wheel. Then, combine the range Doppler algorithm to analyze whether the sky wheel swing amplitude is within the normal range. When the sky wheel swing amplitude is within the normal range, it is determined that the sky wheel is in a normal working state, and S1 is re-executed. When the sky wheel swing amplitude exceeds the normal range, it is determined that the sky wheel is in an abnormal state, and S4 is executed;

[0024] S4: Process the signal to obtain a frequency spectrum, observe the speed change of the overhead sheave swing signal over time through the frequency spectrum, and then determine the intensity of the overhead sheave swing, and preliminarily infer the cause of the mine hoist abnormality;

[0025] S5: Use a high-pass filter to eliminate static signals and low-frequency signals in the signal, separate high-frequency vibration signals, and obtain motor vibration signals. Use the range Doppler algorithm to process the motor vibration signals in fast and slow time dimensions to calculate the motor position. Then observe the spectrum at the corresponding position, and intuitively analyze whether the motor is vibrating through the spectrum. If there is motor vibration, further calculate the vibration intensity and speed of the motor, and infer the cause of the motor vibration.

[0026] S6: Analyze the characteristics of the overhead sheave swing signal and the motor vibration signal based on the spectrum diagram, and build a multi-feature fusion framework based on the attention mechanism to obtain a mine hoist safety detection model;

[0027] S7: Input the enhanced signal obtained in real time into the mine hoist safety detection model, and use the mine hoist safety detection model to monitor the operating status of the mine hoist in real time. When it is determined that the sheave swing exceeds the sheave swing threshold, the abnormally operating mine hoist is first marked with a risk logo, and it is noted that there is a risk of aging and breakage of the sheave traction rope. Then, the high-frequency signal, that is, the motor vibration signal, is further extracted from the millimeter-wave radar single channel. If it is confirmed that there is a continuous motor vibration signal, an early warning signal is sent to the data processing terminal. After receiving the early warning signal, the data processing terminal controls the abnormally marked mine hoist to stop working, and sends early warning information to relevant personnel, so that relevant personnel can carry out subsequent inspection and maintenance work in time.

[0028] As a preferred embodiment, in step 4 S1, the process of denoising the signal and enhancing the target signal is as follows:

[0029] S11: Obtain the output amplitude of the n-order Butterworth filter according to formula (1);

[0030]

[0031] Where j is the imaginary unit, ω is the angular frequency of the signal; ω c is the cut-off frequency;

[0032] S12: Based on formula (2), the square amplitude function of the n-order Butterworth filter is obtained;

[0033]

[0034] S13: Due to According to formula (3), the transfer function H with phase frequency shift characteristics is n (s)H n Derivation of (-s);

[0035]

[0036] In the formula, H n (s) There are n extreme points s1, s2, ...s n , H n The n poles of (-s) are H n (s) corresponds to the negative symmetric point, that is, -s1 = s1e jπ , -s2=-s2e jπ ,…,-s n =s n e jπ ,

[0037]

[0038] S14: Transfer function H n (s)H n The extreme value of (-s) is solved to obtain the gain s of each order, as shown in formula (4);

[0039]

[0040] S15: The signal after clutter removal can be obtained by subtracting the increment from the signal. After filtering by the Butterworth filter, the signal gradually decreases to zero in the stop band.

[0041] As a preferred embodiment, in step 4 S2, the process of using the average cancellation algorithm to filter out stationary clutter is as follows:

[0042] S21: Calculate the mean of all received pulses according to formula (5);

[0043]

[0044] In the formula, m and i represent the sampling points of the fast and slow time dimensions respectively;

[0045] S22: Obtain the target echo signal using the average cancellation algorithm according to formula (6);

[0046] R[m,n]=R[m,n]-C[m] (6).

[0047] In step 4 S2, the process of filtering the signal using the differential judgment algorithm is as follows:

[0048] S23: Use the signal vector to intuitively represent the superposition of multiple signals in the received echo signal, as shown in formula (7);

[0049] S=S v +S l +S d +S b (7);

[0050] Where S represents all reflected signals at the receiver, S b Represents the background static interference signal, S d Represents other dynamic interference signals around the mine hoist, S l Represents the swing signal of the sky wheel, S v Represents the motor vibration signal;

[0051] S24: Use a low-pass filter and resample the phase signal in the range box to eliminate motor vibration interference, and obtain all reflected signals S through formula (8) l+d ;

[0052] S l+d=S l +S d (8).

[0053] As a preferred embodiment, in step 4 S3, the process of preliminarily processing the sky wheel swing signal is as follows:

[0054] A31: According to formula (9) and formula (10), define the FMCW signal emitted by the millimeter wave radar to sense the target motion and the received reflected signal;

[0055]

[0056] S rx (t) = Aexp[j2πf c (t-τ)+jπB(t-τ) 2 / T (10);

[0057] In the formula, f c represents the scanning start frequency, B represents the bandwidth of the chirp, T represents the period of the chirp, and τ represents the delay of the received signal;

[0058] A 32: Use mixer and filter to remove carrier S from received signal rx (t), and obtain the intermediate frequency signal, and obtain S according to formula (11) IF (t);

[0059]

[0060] Where A is the path loss, c is the speed of light, and λ is the wavelength;

[0061] A 33: Use Fourier transform to transform the intermediate frequency signal, and obtain the distance d from the wheel to the millimeter wave radar according to formula (12);

[0062]

[0063] Where τ is the delay between the transmitted signal and the received signal;

[0064] A 34: The phase change of the sky wheel is calculated according to formula (13):

[0065]

[0066] Where λ is the wavelength and Δd is the change in the distance from the wheel to the millimeter-wave radar.

[0067] As a preferred embodiment, in step 4 S3, the process of analyzing whether the shaking amplitude of the sky wheel is within the normal range by combining the range Doppler algorithm is as follows:

[0068] B31: Convert the sinusoidal wave signal into an N*M complex matrix for pulse compression, and straighten the distance pulse into a vector, where N is the chirp number and M is the number of sample points;

[0069] B32: First, autocorrelate the pulse compressed signal with the transmitted signal, then extract the horizontal axis frequency corresponding to the spectrum peak, transform the processed echo signal in the fast time dimension, and obtain the target distance R according to formula (14);

[0070]

[0071] In the formula, f staticBeat is the frequency of the beat signal when the target is stationary, f movingBeat is the frequency of the beat signal when the target is in motion, f c is the initial frequency of the sweep, C is the speed of light, t c is the frequency sweep period, f is the center frequency of the Chirp signal;

[0072] B33: Process the azimuth frequency domain, i.e., the slow time dimension, and obtain the Doppler frequency f according to formula (15) d ;

[0073]

[0074] Where, f is the center frequency of the Chirp signal;

[0075] B34: Transform formula (15) to obtain formula (16), and obtain the target speed v through formula (16);

[0076]

[0077] B35: Based on the distance d from the sky wheel to the millimeter wave radar and the phase change of the sky wheel Doppler frequency f d The target speed v is used to determine whether the swing amplitude of the sky wheel exceeds the set sky wheel swing threshold. When it exceeds the set sky wheel swing threshold, it is determined that the sky wheel is in an abnormal state.

[0078] As a preferred embodiment, in step 4 S4, the process of determining the intensity of the sheave swing and preliminarily inferring the cause of the mine hoist abnormality is as follows:

[0079] S41: define a window function ω(t) to intercept the signal and divide the signal into small signal segments;

[0080] S42: Perform short-time Fourier transform on the original signal x(t), place the window function ω(t) at the beginning of the signal, and make the center position of the window function ω(t) correspond to t=τ0, and obtain y(t) after processing, as shown in formula (17);

[0081] y(t)=x(t)·w(t-τ0) (17);

[0082] S43: further performing Fourier transform on the windowed signal to obtain the spectrum distribution X(ω) according to formula (18);

[0083]

[0084] S44: The signal is a discretely distributed point sequence, so the obtained X(ω) is finally arranged into a spectrum sequence X[N], and the defined function S(ω,τ) is used to represent the result when the center of the window function is at τ, as shown in formula (19);

[0085]

[0086] S45: eliminating the dynamic interference caused by the swing of the sheave and other parts of the hoist, and obtaining a spectrum diagram of the sheave swing signal without dynamic interference;

[0087] S46: Continuously move the window function ω(t) to repeat the operations from S41 to S45 for different segmented signals, and finally obtain the spectrum result of the complete signal. The intensity of the sheave swing can be determined through the temporal speed change reflected in the spectrum diagram, and the cause of the mine hoist abnormality can be further analyzed.

[0088] As a preferred embodiment, in step 4 S45, the process of eliminating the dynamic interference caused by the swing of the hoist to other parts of the hoist is as follows:

[0089] S451: Obtain an estimated signal of dynamic interference according to formula (20)

[0090]

[0091] Where n, N∈Z, n≠0, S l+d (R) is the signal of the sky wheel swing;

[0092] S452: Use short-time Fourier transform to obtain the Doppler spectrum of the sky wheel motion related signal STFT (S l+d ) and the Doppler spectrum of the dynamic interference signal According to formula (21), the time-frequency diagram STFT (S L );

[0093]

[0094] As a preferred embodiment, in S6 of step 4, a multi-feature fusion framework is constructed based on the attention mechanism to obtain a mine hoist safety detection model as follows:

[0095] S61: Establish motor vibration encoder;

[0096] Use embedding in machine learning to transform the input segment x=(X1,...,X n ) is converted to a continuous representation output segment y = (Y1,..., Y n ), where x and y are both real numbers. For the embedded segmentation vector, the attention function is calculated using formula (22);

[0097]

[0098] In the formula, Q represents a set of query matrices, and the keys and values ​​are packed together into matrices K and V. represents the scale factor;

[0099] S62: Establishing a head wheel swing encoder;

[0100] According to the time dimension, the spectrum graph L∈R H×W×C Perform segmentation to form a series of flattened 2D surface molecular formulas L p ∈R M×(H×W×C) , where (H, W) is the resolution of the original spectrogram, (H, P) is the resolution of each patch, C is the number of channels, and M = W / P is the number of patches obtained;

[0101] S63: Multimodal fusion;

[0102] Through the attention-based multi-feature fusion method, the motor vibration features and the wheel swing features are effectively fused; s =att((K l ,V l ),Q s ) and A l =att((K s ,V s ),Q l ), realize the feature association between the sky wheel swing and the motor vibration through cross attention information; let AF = att((K F ,V F ),Q s )+att((K F ,V F ),Q l ), by combining attention information, feature complementation between the sky wheel swing and the motor vibration is achieved.

[0103] In the present invention, the millimeter wave radar uses beamforming to focus energy on the mine hoist sheave and motor part, which can ensure the maximization of the energy of the transmitted signal and the reflected signal, thereby facilitating the accuracy of the detection result. The radar is used to transmit a linear frequency modulation continuous wave, which is reflected by the target part and processed by the mixer to generate an intermediate frequency signal, and then the intermediate frequency signal is subjected to radar signal analysis and processing, so as to determine the position of the mine hoist, whether the sheave swings, and the slight vibration of the motor. Further combined with the RDA algorithm and the STFT algorithm to convert the signal frequency domain to the time domain and segment the processed signal, the distance from the hoist sheave swinging to the radar in the time domain, the phase change, and the speed change of the motor vibration signal can be determined. Therefore, the abnormal swing of the sheave and the abnormality of the motor can be accurately judged. In the process of processing the signal, for the reflected signal from other strong stationary targets in the underground tunnel of the coal mine, the average phase cancellation algorithm is used to filter out the stationary clutter, and for the signal from other dynamic interference signals, the differential judgment algorithm is used to filter the signal, thereby effectively removing clutter and interference signals, ensuring the accuracy of the subsequent analysis results.

[0104] Abnormal operation of the hoist will cause a series of minor equipment activities such as the swing of the sheave and the vibration of the motor. A single feature is not enough to prove that the hoist has failed, so combining multiple features for judgment and analysis can improve the detection accuracy, avoid waste of resources, and save maintenance costs. This application innovatively combines the sheave shaking characteristics with the motor vibration characteristics by building a fusion module, and obtains a mine hoist safety detection model. The mine hoist safety detection model effectively integrates the corresponding characteristics of the sheave and the motor, effectively improving the accuracy of mine hoist working status recognition. At the same time, it fully combines the inherent correlation and complementarity between the two modes, and can accurately and efficiently determine whether the mine hoist is working normally. It is particularly suitable for detection operations in complex mines with a lot of dust and obstructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 is a flow chart of the present invention;

[0106] Figure 2 It is a schematic diagram of the detection device in the present invention;

[0107] Figure 3 It is a schematic diagram of the millimeter wave radar signal transmission and reception process in the present invention;

[0108] Figure 4 It is a linear pulse, distance-amplitude schematic diagram of the millimeter wave radar transmission signal and the receiving signal in the present invention;

[0109] Figure 5 It is a schematic diagram of multi-feature fusion verification in the present invention. DETAILED DESCRIPTION

[0110] The present invention will be further described below in conjunction with the accompanying drawings.

[0111] like Figures 1 to 5 As shown, the present invention provides a non-contact vibration detection device for a hoist motor based on a millimeter wave radar, comprising a mine hoist, a millimeter wave radar, a data acquisition module and a data processing terminal;

[0112] The millimeter wave radar includes a transmitting antenna, a receiving antenna, a radio frequency component, a digital-to-analog converter and a digital signal processor;

[0113] The transmitting antenna is used to convert the received outgoing electrical signal into an electromagnetic wave signal and transmit it in a specific direction;

[0114] The receiving antenna is used to convert the received reflected electromagnetic waves into reflected electrical signals and send them to the radio frequency component;

[0115] The radio frequency component includes a synthesizer and a mixer. The mixer is used to synthesize the received reflected electrical signal and the transmitted electrical signal into an intermediate frequency signal and send it to the digital-to-analog converter; the synthesizer is used to generate a linear frequency modulated transmitted electrical signal from the received single frequency transmitted electrical signal and then send it to the transmitting antenna and the mixer;

[0116] The digital-to-analog converter is used to perform analog-to-digital conversion on the received intermediate frequency signal and then send it to the digital signal processor, which is used to perform analog-to-digital conversion on the received transmission electrical signal and then send it to the radio frequency component;

[0117] The digital signal processor is used to send a transmission signal to the digital-to-analog converter according to a set frequency, and is used to process the received intermediate frequency signal;

[0118] The number of the millimeter wave radars is multiple and the number is the same as the number of mine hoists. The millimeter wave radars are installed in the radial direction of the corresponding mine hoists, and the wireless signal energy emitted by the millimeter wave radars is focused on the sheave and motor of the corresponding mine hoist.

[0119] The data processing terminal is connected to the plurality of millimeter wave radars respectively through the data acquisition module. As a preferred embodiment, the data acquisition module is connected to the millimeter wave radar through a USB interface;

[0120] As a preference, the number of the transmitting antenna is one, and the number of the receiving antennas is four.

[0121] As a preferred embodiment, the data processing terminal is a computer. Further preferably, the computer is a Dell gaming box equipped with an Intel(R) Core(TM) i7-10750H CPU, 16G RAM, and a system Windows 11. Further preferably, the computer is equipped with mmWave Studio 02.01.01.00.

[0122] As a preferred embodiment, the model of the millimeter wave radar is TI AWR1843 Boost EVM ES2.0; and the data acquisition module is DCA1000 EVM.

[0123] Before testing, debugging is required. First, make sure that all the ports connected to the device manager are normal. After debugging, open mmWave Studio 02.01.01.00 and configure the COM PORT port to the port number corresponding to Application / User UART to load the firmware BSS and MSS. Then set the number of transmitting antennas and receiving antennas to one for four, and then set the number of ADC sample points, sampling frequency, number of frames, number of chirps, etc. Finally, set the data acquisition module (acquisition card). The data collected by DCA1000 is transmitted to the computer through the UDP protocol, that is, the PC needs to read the network cable interface data. The IP address in the data acquisition module DCA1000 is 192.168.33.30, so the Ethernet static IP on the computer needs to be configured as 192.168.33.30. After all configurations are completed, use mmWave Studio 02.01.01.00 software to receive the FMCW reflected echo signal emitted by the millimeter wave radar.

[0124] The present invention has a simple structure and low investment cost. The millimeter wave radar is installed in the radial direction of the mine hoist, and the swing displacement of the sheave and the vibration displacement of the motor can be obtained by calculating the phase change of the millimeter wave radar signal. Therefore, the swing of the sheave and the vibration of the motor can be sensed by processing the transmission signal and the reflection signal, and the abnormality of the hoist can be detected quickly and efficiently in a non-contact manner. The data processing terminal is connected to multiple millimeter wave radars through a data acquisition module, and the data processing terminal can be used to receive the signals collected by multiple millimeter wave radars and the alarm signals sent, which is conducive to the unified recording and monitoring of the detection conditions of multiple mine hoists, so that effective treatment measures can be taken for abnormal hoists in a timely manner.

[0125] The present invention also provides a non-contact vibration detection method for a hoist motor based on a millimeter wave radar, which uses a non-contact vibration detection device for a hoist motor based on a millimeter wave radar, and includes the following steps:

[0126] Step 1: Conduct on-site surveys of each mine hoist in the coal mine, and determine the position with good signal detection effect in the radial direction of the mine hoist as the installation position of the corresponding millimeter wave radar; mark the installation position, and record the specific coordinates and environmental description of the installation position;

[0127] Step 2: Install millimeter-wave radars at the installation positions of each mine hoist in the coal mine, and adjust the angle of each millimeter-wave radar transmitting antenna to ensure that the wireless signal energy emitted by the millimeter-wave radar is focused on the sheave and motor of the corresponding mine hoist. At the same time, adjust the angle of each millimeter-wave radar receiving antenna to ensure that the reflected wireless signal energy is focused on the location of the receiving antenna;

[0128] Step 3: Set the working parameters and sampling parameters of each millimeter-wave radar, and start the millimeter-wave radar. During the operation of the mine hoist, use the synthesizer in the millimeter-wave radar to generate a linear frequency modulated continuous wave, and transmit it to the target part through the transmitting antenna. After being reflected by the target part and processed by the mixer, an intermediate frequency signal is generated, and the intermediate frequency signal is received and recorded in real time by the digital signal processor.

[0129] As a preferred embodiment, the sampling parameters of the millimeter wave radar are set as follows: the sampling rate is set to 2000kps, 200 frames are collected and the chirp number is set to 128, and the number of sample points is set to 128.

[0130] Step 4: The received intermediate frequency signal is processed in chronological order by a digital signal processor. The processing process is as follows:

[0131] S1: The radar calculates the distance from the target to the millimeter wave radar through delay calculation. Figure 2 As shown in the figure, the delay is the time interval between the transmitting and receiving antennas, and the distance difference between two consecutive chirps is not obvious. Usually, the distance is used to continue to calculate the target signal angle and speed for analysis. The millimeter wave radar simultaneously senses and collects the sky wheel swing signal and the motor vibration signal. mmWave Studio 02.01.01.00 is used to receive the reflected signal and analyze the characteristics of the sky wheel swing signal and the motor vibration signal.

[0132] The collected echo signal is in the form of a sine wave and is converted into a binary raw data file by mmWave Studio. The echo signal collected by mmWave Studio is read and processed by MATLAB. Usually, the radar signal processing process includes signal generation, signal extraction, signal conversion and other aspects.

[0133] ① Signal transmission and reception, Tx transmits linear frequency modulated continuous wave, and Rx receives the signal and outputs analog signal through the mixer.

[0134] ②Convert the echo signal into data, converting the sine wave into a binary signal that can be processed by the machine

[0135] ③Process the obtained digital signal file to remove noise, enhance the target signal, and perform Fourier transform to determine the target position, signal frequency and other information.

[0136] De-noising the intermediate frequency signal and enhancing the target signal strength to obtain an enhanced target signal;

[0137] S2: After receiving the enhanced signal, the digital signal processor first determines whether there is a sheave swing signal. If there is no sheave swing signal, S1 is executed again. If there is a sheave swing signal, the reflection source of the signal is first determined. If the signal is a reflection signal from other strong stationary targets in the underground tunnel of the coal mine, the average cancellation algorithm is used to filter out the stationary clutter. If the signal comes from other dynamic interference signals, the differential judgment algorithm is used to filter the signal.

[0138] S3: Range-Doppler Arithmetic (RDA) is a method for linear frequency modulation millimeter wave sensing equipment to effectively extract target range and velocity information characteristics. Range-Doppler Arithmetic extracts target feature information, usually by operating on image information (range Doppler heat map). The amplitude-range image after Fourier transformation is processed to change the horizontal and vertical coordinates to the Doppler domain, that is, the Chirp sequence of multiple cycles emitted by the FMCW radar and the echo signal are processed in the fast and slow time dimensions.

[0139] First, the sky wheel swing signal is obtained through filtering, which can facilitate the analysis of the sky wheel swing amplitude, frequency, speed, phase information, etc.; then, the sky wheel swing signal is preliminarily processed to obtain the distance from the sky wheel to the millimeter wave radar and the phase change of the sky wheel, and then, the range Doppler Arithmetic (RDA) algorithm is combined to analyze whether the sky wheel swing amplitude is within the normal range. When the sky wheel swing amplitude is within the normal range, it is determined that the sky wheel is in a normal working state, and S1 is re-executed. When the sky wheel swing amplitude exceeds the normal range, it is determined that the sky wheel is in an abnormal state, and S4 is executed;

[0140] S4: The main reason for the sheave shaking depends on the wire rope used for sheave pulling. If the pulling rope is damaged to varying degrees, such as broken wire, wear, deformation, or broken strands, the sheave will become unstable. In addition to the sheave structure itself, the sheave shaking may also be caused by improper placement of cargo, resulting in uneven force on the sheave pulling rope. The above factors will lead to abnormal working conditions of the mine hoist. However, if the sheave shakes due to external forces during stable operation, thus triggering an early warning, this will cause unnecessary panic and waste of resources in the mine. Therefore, it is necessary to determine the actual cause of the sheave shaking.

[0141] The signal is processed to obtain a spectrum diagram, through which the speed change of the sheave swing signal over time is observed, and then the intensity of the sheave swing is determined, and the cause of the mine hoist abnormality is preliminarily inferred;

[0142] S5: The reasons for motor vibration are complex, and any error may cause immeasurable losses to underground mine operations. Motor vibration factors such as encoder feedback pulse error, transmission speed, and uneven force on the drum can cause abnormal operation of the hoist. However, the slight vibration of the motor may be due to the structural problem of the motor itself, and cannot prove that the hoist is malfunctioning. Therefore, it is necessary to determine the actual cause of the motor vibration;

[0143] A high-pass filter is used to eliminate static signals and low-frequency signals in the signal, separate high-frequency vibration signals, and obtain motor vibration signals. Since the swing of the sky wheel is low-frequency and has a large amplitude, the interference of the sky wheel swing can be easily eliminated by using a high-pass filter; the motor vibration signal is processed in the fast and slow time dimensions by the range Doppler algorithm to calculate the motor position, and then the spectrum at the corresponding position is observed. Since the motor vibration signal has the characteristics of high frequency and small amplitude, the spectrum can be used to intuitively analyze whether the motor is vibrating, such as Figure 4 and Figure 5 As shown, in the case of motor vibration, the vibration intensity and speed of the motor are further calculated, and the cause of the motor vibration is inferred;

[0144] Since the motor is located in the radial direction of the millimeter-wave radar, when the motor vibrates, the motor vibration displacement can be obtained by calculating the phase change of the millimeter-wave radar signal. The distance resolution is much smaller than the phase resolution. If the displacement of the moving target is 1mm (for 60GHz, radar λ=5mm, B=4GHz), the phase change of the IF signal is 0.8π, that is, 144°, so the phase change can reflect the tiny vibration signal of the motor.

[0145] S6: Analyze the characteristics of the overhead sheave swing signal and the motor vibration signal based on the spectrum diagram, and build a multi-feature fusion framework based on the attention mechanism to obtain a mine hoist safety detection model;

[0146] S7: Input the enhanced signal obtained in real time into the mine hoist safety detection model, and use the mine hoist safety detection model to monitor the operating status of the mine hoist in real time. When it is determined that the sheave swing exceeds the sheave swing threshold, the abnormally operating mine hoist is first marked with a risk logo, and it is noted that there is a risk of aging and breakage of the sheave traction rope. Then, the high-frequency signal, that is, the motor vibration signal, is further extracted from the millimeter-wave radar single channel. If it is confirmed that there is a continuous motor vibration signal, an early warning signal is sent to the data processing terminal. After receiving the early warning signal, the data processing terminal controls the abnormally marked mine hoist to stop working, and sends early warning information to relevant personnel, so that relevant personnel can carry out subsequent inspection and maintenance work in time.

[0147] The spectrum obtained by fast Fourier transform is a time-frequency diagram that can reflect the speed change in time, such as Figure 5 As shown in the figure, the sheave swing characteristic diagram reflects the speed of the sheave swing process, so as to analyze and judge the intensity of the sheave swing. Therefore, the speed of the sheave swing process can be captured by STFT to judge the sheave swing amplitude and possible causes of the sheave swing. By observing the sheave swing motion characteristic diagram, it is not difficult to find the sheave phase change and speed characteristics in the time domain. Combined with the motor vibration spectrum diagram corresponding to the sheave swing image, it can be determined that the mine hoist is working abnormally and it is necessary to issue a countermeasure instruction immediately.

[0148] As a preferred embodiment, in step 4 S1, the process of denoising the signal and enhancing the target signal is as follows:

[0149] S11: Obtain the output amplitude of the n-order Butterworth filter according to formula (1);

[0150]

[0151] Where j is the imaginary unit, ω is the angular frequency of the signal; ω c is the cut-off frequency;

[0152] S12: Based on formula (2), the square amplitude function of the n-order Butterworth filter is obtained;

[0153]

[0154] S13: Due to According to formula (3), the transfer function H with phase frequency shift characteristics is n (s)H n Derivation of (-s);

[0155]

[0156] In the formula, H n(s) There are n extreme points s1, s2, ...s n , H n The n poles of (-s) are H n (s) corresponds to the negative symmetric point, that is, -s1 = s1e jπ , -s2=-s2e jπ ,…,-s n =s n e jπ ,

[0157]

[0158] S14: It is not difficult to find that s1, s2, …s n , and s1e jπ , -s2e jπ ,…,s n e jπ It also constitutes H n (s)H n (-s) has 2n poles. Solve for the transfer function H n (s)H n All extreme points of (-s) are the gains of each order that researchers are looking for. n (s)H n (-s) is solved to obtain the gain s of each order. Therefore, the gain s can be obtained as shown in formula (4);

[0159]

[0160] S15: The signal after clutter removal can be obtained by subtracting the increment from the signal. After filtering by the Butterworth filter, the signal gradually decreases to zero in the stop band (noise clutter).

[0161] As a preferred embodiment, in step 4 S2, the process of using the average cancellation algorithm to filter out stationary clutter is as follows:

[0162] S21: The target object with stationary clutter is fixed to the millimeter wave radar transmitting antenna, so the delay between the received signal and the transmitted signal is also fixed. The mean of all received pulses is calculated and then the mean is subtracted from each received pulse to obtain the target echo signal. Specifically, the mean of all received pulses is calculated according to formula (5);

[0163]

[0164] In the formula, m and i represent the sampling points of the fast and slow time dimensions respectively;

[0165] S22: Obtain the target echo signal using the average cancellation algorithm according to formula (6);

[0166] R[m,n]=R[m,n]-C[m] (6).

[0167] In step 4 S2, the process of filtering the signal using the differential judgment algorithm is as follows:

[0168] S23: Use the signal vector to intuitively represent the superposition of multiple signals in the received echo signal, as shown in formula (7);

[0169] S=S v +S l +S d +S b (7);

[0170] Where S represents all reflected signals at the receiver, S b Represents the background static interference signal, S d Represents other dynamic interference signals around the mine hoist, S l Represents the swing signal of the sky wheel, S v Represents the motor vibration signal;

[0171] S24: Because S b It is a static background component, which is directly offset according to the variability of the reflected signal. Since the sky wheel swing signal is a low-frequency signal with significant amplitude changes, a low-pass filter is used and the phase signal in the range box is resampled to eliminate the motor vibration interference. All reflected signals S are obtained by formula (8): l+d ;

[0172] S l+d =S l +S d (8).

[0173] As a preferred embodiment, in step 4 S3, the process of preliminarily processing the sky wheel swing signal is as follows:

[0174] A31: According to formula (9) and formula (10), define the FMCW signal emitted by the millimeter wave radar to sense the target motion and the received reflected signal;

[0175]

[0176] S rx (t) = Aexp[j2πf c (t-τ)+jπB(t-τ) 2 / T (10);

[0177] In the formula, fc represents the scanning start frequency, B represents the bandwidth of the chirp, T represents the period of the chirp, and τ represents the delay of the received signal;

[0178] A 32: Use mixer and filter to remove carrier S from received signal rx (t), and obtain the intermediate frequency signal, and obtain S according to formula (11) IF (t);

[0179]

[0180] Where A is the path loss, c is the speed of light, and λ is the wavelength;

[0181] A 33: By transforming the intermediate frequency signal using Fourier transform, basic information such as the frequency amplitude of the target signal, the distance from the target to the radar, the target transformation speed, and the initial phase angle can be obtained. Specifically, the distance d from the wheel to the millimeter wave radar is obtained according to formula (12);

[0182]

[0183] Where τ is the delay between the transmitted signal and the received signal;

[0184] A 34: The phase change of the sky wheel is calculated according to formula (13):

[0185]

[0186] Where λ is the wavelength and Δd is the change in the distance from the wheel to the millimeter-wave radar.

[0187] As a preferred embodiment, in step 4 S3, the process of analyzing whether the shaking amplitude of the sky wheel is within the normal range by combining the range Doppler algorithm is as follows:

[0188] B31: The millimeter-wave radar transmits a signal that is reflected and generates an intermediate frequency signal. This signal is a sinusoidal signal. After a series of radar signal processing, the received sinusoidal signal is converted into an N*M complex matrix (N is the chirp number, M is the number of sample points) for pulse compression, and the distance pulse is straightened into a vector.

[0189] Convert the sinusoidal wave signal into an N*M complex matrix for pulse compression, and straighten the distance pulse into a vector, where N is the chirp number and M is the number of sample points;

[0190] B32: First, the processed echo signal is transformed in the fast time dimension. The fast time corresponds to the distance frequency domain, so the horizontal axis frequency corresponding to the spectrum peak is extracted to solve the distance of the target;

[0191] First, the pulse compressed signal is autocorrelated with the transmitted signal, that is, matched filtering (range-dimensional FFT), and then the horizontal axis frequency corresponding to the spectrum peak is extracted. The processed echo signal is transformed in the fast time dimension (Range-FFT), and the target distance R is obtained according to formula (14);

[0192]

[0193] In the formula, f staticBeat is the frequency of the beat signal when the target is stationary, f movingBeat is the frequency of the beat signal when the target is in motion, f c is the initial frequency of the sweep, C is the speed of light, t c is the frequency sweep period, f is the center frequency of the Chirp signal;

[0194] B33: After obtaining the target distance, the azimuth frequency domain is processed, that is, the slow time dimension is processed, and the Doppler frequency f is obtained according to formula (15) d ;

[0195]

[0196] Where, f is the center frequency of the Chirp signal;

[0197] B34: Transform formula (15) to obtain formula (16), and obtain the target speed v through formula (16);

[0198]

[0199] B35: Based on the distance d from the sky wheel to the millimeter wave radar, the target distance R, and the phase change of the sky wheel Doppler frequency f d The target speed v is used to determine whether the swing amplitude of the sky wheel exceeds the set sky wheel swing threshold. When it exceeds the set sky wheel swing threshold, it is determined that the sky wheel is in an abnormal state.

[0200] The difference frequency signal containing distance and speed information can be obtained through the above two formulas. In the algorithm design, windowing operation is usually performed before Fourier transform and range Doppler transform to make the result more accurate. The range Doppler heat map obtained by range Doppler processing reflects the time-dependent distance transformation relationship of the reflected target motion signal. After determining the position of the hoist sheave, the signal is further Fourier transformed to observe the phase change of the sheave to infer whether the sheave is shaking, and the magnitude of the shaking to determine whether there is an abnormal risk.

[0201] The fast Fourier transform (STFT) change can observe the speed change relationship of the overhead sheave swing signal in time. Different reasons lead to different degrees of overhead sheave swing. Therefore, the intensity of overhead sheave swing can be determined through STFT and the cause of mine hoist abnormality can be preliminarily inferred. The core principle of STFT is to use the window function to intercept the signal, perform Fourier transform on each signal segment, and then move the window to process different segmented signals to obtain a complete time-frequency diagram.

[0202] As a preferred embodiment, in step 4 S4, the process of determining the intensity of the sheave swing and preliminarily inferring the cause of the mine hoist abnormality is as follows:

[0203] S41: define a window function ω(t) to intercept the signal and divide the signal into small signal segments;

[0204] S42: Perform short-time Fourier transform on the original signal x(t), place the window function ω(t) at the beginning of the signal, and make the center position of the window function ω(t) correspond to t=τ0, and obtain y(t) after processing, as shown in formula (17);

[0205] y(t)=x(t)·w(t-τ0) (17);

[0206] S43: further performing Fourier transform on the windowed signal to obtain the spectrum distribution X(ω) according to formula (18);

[0207]

[0208] S44: The signal is a discretely distributed point sequence, so the obtained X(ω) is finally arranged into a spectrum sequence X[N], and the defined function S(ω,τ) is used to represent the result when the center of the window function is at τ, as shown in formula (19);

[0209]

[0210] Each column of S[ω,τ] represents a two-dimensional matrix of the results of windowing the signal at different positions and performing Fourier transform on the segments.

[0211] S45: eliminating the dynamic interference caused by the swing of the sheave and other parts of the hoist, and obtaining a spectrum diagram of the sheave swing signal without dynamic interference;

[0212] S46: Continuously move the window function ω(t) to repeat the operations from S41 to S45 for different segmented signals, and finally obtain the spectrum result of the complete signal. The intensity of the sheave swing can be determined through the temporal speed change reflected in the spectrum diagram, and the cause of the mine hoist abnormality can be further analyzed.

[0213] As a preferred method, in order to obtain the time-frequency diagram of the sheave swing signal, it is necessary to first eliminate the dynamic interference such as shaking of other parts of the hoist caused by the sheave swing; the obtained reflection signal S l+d , through the range Doppler analysis, it can be found that the dynamic interference is contained in the distance box where the sheave swings, and has obvious changes in its adjacent boxes. By adopting the differential-based dynamic interference removal method, the dynamic interference such as shaking of other parts of the elevator caused by the sheave swing can be effectively eliminated; in step 4 S45, the process of eliminating the dynamic interference caused by the sheave swing to other parts of the elevator is as follows:

[0214] S451: Obtain an estimated signal of dynamic interference according to formula (20)

[0215]

[0216] Where n, N∈Z, n≠0, S l+d (R) is the signal of the sky wheel swing;

[0217] S452: Use short-time Fourier transform (STFT) to obtain the Doppler spectrum of the signal related to the movement of the sky wheel. l+d ) and the Doppler spectrum of the dynamic interference signal According to formula (21), the time-frequency diagram STFT (S L );

[0218]

[0219] As a preferred embodiment, in S6 of step 4, a multi-feature fusion framework is constructed based on the attention mechanism to obtain a mine hoist safety detection model as follows:

[0220] S61: Establish motor vibration encoder;

[0221] Use embedding in machine learning to transform the input segment x=(X1,...,X n ) is converted to a continuous representation output segment y = (Y1,..., Y n ), where x and y are both real numbers. For the embedded segmentation vector, the attention function is calculated using formula (22);

[0222]

[0223] In the formula, Q represents a set of query matrices, and the keys and values ​​are packed together into matrices K and V. represents the scale factor;

[0224] S62: Establishing a head wheel swing encoder;

[0225] According to the time dimension, the spectrum graph L∈R H×W×C Perform segmentation to form a series of flattened 2D surface molecular formulas L p ∈R M×(H×W×C) , where (H, W) is the resolution of the original spectrogram, (H, P) is the resolution of each patch, C is the number of channels, and M = W / P is the number of patches obtained; the temporal logic in the wheel swing spectrogram is preserved through a standard learnable one-dimensional position embedding;

[0226] S63: Multimodal fusion;

[0227] In order to reasonably and effectively fuse the motor vibration and wheel swing features, a multi-feature fusion method based on attention is proposed, which connects the motor vibration features and the wheel swing features alternately through cross attention and merge attention. s =att((K l ,V l ),Q s ) and A l =att((K s ,V s ),Q l ), is to realize the characteristic association between the swing of the sky wheel and the vibration of the motor through cross-attention information; the essence of full attention, that is, AF=att((K F ,V F ),Q s )+att((K F ,V F ),Q l ), is to achieve feature complementarity between the sky wheel swing and the motor vibration by combining attention information.

[0228] In the present invention, the millimeter wave radar uses beamforming to focus energy on the mine hoist sheave and motor part, which can ensure the maximization of the energy of the transmitted signal and the reflected signal, thereby facilitating the accuracy of the detection result. The radar is used to transmit a linear frequency modulation continuous wave, which is reflected by the target part and processed by the mixer to generate an intermediate frequency signal, and then the intermediate frequency signal is subjected to radar signal analysis and processing, so as to determine the position of the mine hoist, whether the sheave swings, and the slight vibration of the motor. Further combined with the RDA algorithm and the STFT algorithm to convert the signal frequency domain to the time domain and segment the processed signal, the distance from the hoist sheave swinging to the radar in the time domain, the phase change, and the speed change of the motor vibration signal can be determined. Therefore, the abnormal swing of the sheave and the abnormality of the motor can be accurately judged. In the process of processing the signal, for the reflected signal from other strong stationary targets in the underground tunnel of the coal mine, the average phase cancellation algorithm is used to filter out the stationary clutter, and for the signal from other dynamic interference signals, the differential judgment algorithm is used to filter the signal, thereby effectively removing clutter and interference signals, ensuring the accuracy of the subsequent analysis results.

[0229] Abnormal operation of the hoist will cause a series of minor equipment activities such as the swing of the sheave and the vibration of the motor. A single feature is not enough to prove that the hoist has failed, so combining multiple features for judgment and analysis can improve the detection accuracy, avoid waste of resources, and save maintenance costs. This application innovatively combines the sheave shaking characteristics with the motor vibration characteristics by building a fusion module, and obtains a mine hoist safety detection model. The mine hoist safety detection model effectively integrates the corresponding characteristics of the sheave and the motor, effectively improving the accuracy of mine hoist working status recognition. At the same time, it fully combines the inherent correlation and complementarity between the two modes, and can accurately and efficiently determine whether the mine hoist is working normally. It is particularly suitable for detection operations in complex mines with a lot of dust and obstructions.

Claims

1. A non-contact vibration detection device for a hoist motor based on a millimeter wave radar, comprising a mine hoist and a millimeter wave radar, characterized in that: It also includes a data acquisition module and a data processing terminal; The millimeter wave radar includes a transmitting antenna, a receiving antenna, a radio frequency component, a digital-to-analog converter and a digital signal processor; The transmitting antenna is used to convert the received outgoing electrical signal into an electromagnetic wave signal and transmit it in a specific direction; The receiving antenna is used to convert the received reflected electromagnetic waves into reflected electrical signals and send them to the radio frequency component; The radio frequency component includes a synthesizer and a mixer, wherein the mixer is used to synthesize the received reflected electrical signal and the transmitted electrical signal into an intermediate frequency signal and send the intermediate frequency signal to a digital-to-analog converter; The synthesizer is used to generate a linear frequency modulated transmission electrical signal from the received single frequency transmission electrical signal, and then send it to the transmission antenna and the mixer; The digital-to-analog converter is used to perform analog-to-digital conversion on the received intermediate frequency signal and then send it to the digital signal processor, which is used to perform analog-to-digital conversion on the received transmission electrical signal and then send it to the radio frequency component; The digital signal processor is used to send a transmission signal to the digital-to-analog converter according to a set frequency, and is used to process the received intermediate frequency signal; The number of the millimeter wave radars is multiple and the number is the same as the number of mine hoists. The millimeter wave radars are installed in the radial direction of the corresponding mine hoists, and the wireless signal energy emitted by the millimeter wave radars is focused on the sheave and motor of the corresponding mine hoist. The data processing terminal is connected to a plurality of millimeter radars respectively through data acquisition modules.

2. A non-contact vibration detection method for a hoist motor based on a millimeter wave radar, using a non-contact vibration detection device for a hoist motor based on a millimeter wave radar as claimed in claim 1, characterized in that: The following steps are involved: Step 1: Conduct on-site surveys of each mine hoist in the coal mine, and determine the position with good signal detection effect in the radial direction of the mine hoist as the installation position of the corresponding millimeter wave radar; mark the installation position, and record the specific coordinates and environmental description of the installation position; Step 2: Install millimeter-wave radars at the installation positions of each mine hoist in the coal mine, and adjust the angle of each millimeter-wave radar transmitting antenna to ensure that the wireless signal energy emitted by the millimeter-wave radar is focused on the sheave and motor of the corresponding mine hoist. At the same time, adjust the angle of each millimeter-wave radar receiving antenna to ensure that the reflected wireless signal energy is focused on the location of the receiving antenna; Step 3: Set the working parameters and sampling parameters of each millimeter-wave radar, and start the millimeter-wave radar; During the operation of the mine hoist, the synthesizer in the millimeter-wave radar is used to generate a linear frequency modulated continuous wave, which is transmitted to the target part through the transmitting antenna. After being reflected by the target part and processed by the mixer, an intermediate frequency signal is generated, which is received and recorded in real time by the digital signal processor. Step 4: The received intermediate frequency signal is processed in chronological order by a digital signal processor. The processing process is as follows: S1: De-noise the intermediate frequency signal and enhance the target signal strength to obtain an enhanced target signal; S2: After receiving the enhanced signal, the digital signal processor first determines whether there is a sheave swing signal. If there is no sheave swing signal, S1 is executed again. If there is a sheave swing signal, the reflection source of the signal is first determined. If the signal is a reflection signal from other strong stationary targets in the underground tunnel of the coal mine, the average cancellation algorithm is used to filter out the stationary clutter. If the signal comes from other dynamic interference signals, the differential judgment algorithm is used to filter the signal. S3: First, obtain the sky wheel swing signal through filtering, and then preliminarily process the sky wheel swing signal to obtain the distance from the sky wheel to the millimeter wave radar and the phase change of the sky wheel. Then, combine the range Doppler algorithm to analyze whether the sky wheel swing amplitude is within the normal range. When the sky wheel swing amplitude is within the normal range, it is determined that the sky wheel is in a normal working state, and S1 is re-executed. When the sky wheel swing amplitude exceeds the normal range, it is determined that the sky wheel is in an abnormal state, and S4 is executed; S4: Process the signal to obtain a frequency spectrum, observe the speed change of the overhead sheave swing signal over time through the frequency spectrum, and then determine the intensity of the overhead sheave swing, and preliminarily infer the cause of the mine hoist abnormality; S5: Use a high-pass filter to eliminate static signals and low-frequency signals in the signal, separate high-frequency vibration signals, and obtain motor vibration signals. Use the range Doppler algorithm to process the motor vibration signals in fast and slow time dimensions to calculate the motor position. Then observe the spectrum at the corresponding position, and intuitively analyze whether the motor is vibrating through the spectrum. If there is motor vibration, further calculate the vibration intensity and speed of the motor, and infer the cause of the motor vibration. S6: Analyze the characteristics of the overhead sheave swing signal and the motor vibration signal based on the spectrum diagram, and build a multi-feature fusion framework based on the attention mechanism to obtain a mine hoist safety detection model; S7: Input the enhanced signal obtained in real time into the mine hoist safety detection model, and use the mine hoist safety detection model to monitor the operating status of the mine hoist in real time. When it is determined that the sheave swing exceeds the sheave swing threshold, the abnormally operating mine hoist is first marked with a risk logo, and it is noted that there is a risk of aging and breakage of the sheave traction rope. Then, the high-frequency signal, that is, the motor vibration signal, is further extracted from the millimeter-wave radar single channel. If it is confirmed that there is a continuous motor vibration signal, an early warning signal is sent to the data processing terminal. After receiving the early warning signal, the data processing terminal controls the abnormally marked mine hoist to stop working, and sends early warning information to relevant personnel, so that relevant personnel can carry out subsequent inspection and maintenance work in time.

3. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 2 is characterized in that: In step 4 S1, the process of denoising the signal and enhancing the target signal is as follows: S11: Obtain the output amplitude of the n-order Butterworth filter according to formula (1); Where j is the imaginary unit, ω is the angular frequency of the signal; ω c is the cut-off frequency; S12: Based on formula (2), the square amplitude function of the n-order Butterworth filter is obtained; S13: Due to According to formula (3), the transfer function H with phase frequency shift characteristics is n (s)H n Derivation of (-s); In the formula, H n (s) There are n extreme points s1, s2, ...s n , H n The n poles of (-s) are H n (s) corresponds to the negative symmetric point, that is, -s1 = s1e jπ , -s2=-s2e jπ ,…,-s n =s n e jπ , S14: Transfer function H n (s)H n The extreme value point of (-s) is solved to obtain the gain s of each order, as shown in formula (4); S15: The signal after clutter removal can be obtained by subtracting the increment from the signal. After filtering by the Butterworth filter, the signal gradually decreases to zero in the stop band.

4. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 3 is characterized in that: In step 4 S2, the process of using the average cancellation algorithm to filter out stationary clutter is as follows: S21: Calculate the mean of all received pulses according to formula (5); In the formula, m and i represent the sampling points of the fast and slow time dimensions respectively; S22: Obtain the target echo signal using the average cancellation algorithm according to formula (6); R[m,n]=R[m,n]-C[m] (6).

5. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 3 is characterized in that: In step 4 S2, the process of filtering the signal using the differential judgment algorithm is as follows: S23: Use the signal vector to intuitively represent the superposition of multiple signals in the received echo signal, as shown in formula (7); S=S v +S l +S d +S b (7); Where S represents all reflected signals at the receiver, S b Represents the background static interference signal, S d Indicates other dynamic interference signals around the mine hoist, S1 represents the swing signal of the sheave, S v Represents the motor vibration signal; S24: Use a low-pass filter and resample the phase signal in the range box to eliminate motor vibration interference, and obtain all reflected signals S through formula (8) l+d ; S l+d =S l +S d (8)。 6. A non-contact vibration detection method for a hoist motor based on millimeter wave radar according to claim 4 or 5, characterized in that: In step 4 S3, the process of performing preliminary processing on the overhead sheave swing signal is as follows: A31: According to formula (9) and formula (10), define the FMCW signal emitted by the millimeter wave radar to sense the target motion and the received reflected signal; S rx (t)=Aexp[j2πf c (t-τ)+jπB(t-τ) 2 / T (10); In the formula, f c represents the scanning start frequency, B represents the bandwidth of the chirp, T represents the period of the chirp, and τ represents the delay of the received signal; A 32: Use mixer and filter to remove carrier S from received signal rx (t), and obtain the intermediate frequency signal, and obtain S according to formula (11) IF (t); Where A is the path loss, c is the speed of light, and λ is the wavelength; A 33: Use Fourier transform to transform the intermediate frequency signal, and obtain the distance d from the wheel to the millimeter wave radar according to formula (12); Where τ is the delay between the transmitted signal and the received signal; A 34: The phase change of the sky wheel is calculated according to formula (13): Where λ is the wavelength and Δd is the change in the distance from the wheel to the millimeter-wave radar.

7. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 6 is characterized in that: In step 4 S3, the process of analyzing whether the sway amplitude of the sky wheel is within the normal range by combining the range Doppler algorithm is as follows: B31: Convert the sinusoidal wave signal into an N*M complex matrix for pulse compression, and straighten the distance pulse into a vector, where N is the chirp number and M is the number of sample points; B32: First, autocorrelate the pulse compressed signal with the transmitted signal, then extract the horizontal axis frequency corresponding to the spectrum peak, transform the processed echo signal in the fast time dimension, and obtain the target distance R according to formula (14); In the formula, f staticBeat is the frequency of the beat signal when the target is stationary, f movingBeat is the frequency of the beat signal when the target is in motion, f c is the initial frequency of the frequency sweep, C is the speed of light, t c is the frequency sweep period, f is the center frequency of the Chirp signal; B33: Process the azimuth frequency domain, i.e., the slow time dimension, and obtain the Doppler frequency fd according to formula (15); Where, f is the center frequency of the Chirp signal; B34: Transform formula (15) to obtain formula (16), and obtain the target speed v through formula (16); B35: Based on the distance d from the sky wheel to the millimeter wave radar and the phase change of the sky wheel Doppler frequency f d The target speed v is used to determine whether the swing amplitude of the sky wheel exceeds the set sky wheel swing threshold. When it exceeds the set sky wheel swing threshold, it is determined that the sky wheel is in an abnormal state.

8. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 7 is characterized in that: In step 4 S4, the process of determining the severity of the sheave swing and preliminarily inferring the cause of the mine hoist abnormality is as follows: S41: define a window function ω(t) to intercept the signal and divide the signal into small signal segments; S42: Perform short-time Fourier transform on the original signal x(t), place the window function ω(t) at the beginning of the signal, and make the center position of the window function ω(t) correspond to t=τ0, and obtain y(t) after processing, as shown in formula (17); y(t)=x(t)·w(t-τ0) (17); S43: further performing Fourier transform on the windowed signal to obtain the spectrum distribution X(ω) according to formula (18); S44: The signal is a discretely distributed point sequence, so the obtained X(ω) is finally arranged into a spectrum sequence X[N], and the defined function S(ω, T) is used to represent the result when the center of the window function is τ, as shown in formula (19); S45: eliminating the dynamic interference caused by the swing of the sheave and other parts of the hoist, and obtaining a spectrum diagram of the sheave swing signal without dynamic interference; S46: Continuously move the window function ω(t) to repeat the operations from S41 to S45 for different segmented signals, and finally obtain the spectrum result of the complete signal. The intensity of the sheave swing can be determined through the temporal speed change reflected in the spectrum diagram, and the cause of the mine hoist abnormality can be further analyzed.

9. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 8 is characterized in that: In step 4 S45, the process of eliminating the dynamic interference caused by the swing of the hoist to other parts of the hoist is as follows: S451: Obtain an estimated signal of dynamic interference according to formula (20) Where n, N∈Z, n≠0, S l+d (R) is the signal of the sky wheel swing; S452: Use short-time Fourier transform to obtain the Doppler spectrum of the sky wheel motion related signal STFT (S l+d ) and the Doppler spectrum of the dynamic interference signal According to formula (21), the time-frequency diagram STFT (S L ); 10. The non-contact vibration detection method of a hoist motor based on millimeter wave radar according to claim 9, characterized in that: In step 4, S6, a multi-feature fusion framework is constructed based on the attention mechanism to obtain the mine hoist safety detection model as follows: S61: Establish motor vibration encoder; Use embedding in machine learning to transform the input segment x=(X1,...,X n ) is converted into a continuous representation output segment y = (Y1, ..., Yn), where x and y are both real numbers. For the embedded segmentation vector, the attention function is calculated using formula (22); In the formula, Q represents a set of query matrices, and the keys and values ​​are packed together into matrices K and V. represents the scale factor; S62: Establishing a head wheel swing encoder; According to the time dimension, the spectrum graph L∈R H×W×C Perform segmentation to form a series of flattened 2D surface molecular formulas L p ∈R M×(H×W×C) , where (H, W) is the resolution of the original spectrogram, (H, P) is the resolution of each patch, C is the number of channels, and M = W / P is the number of patches obtained; S63: Multimodal fusion; Through the attention-based multi-feature fusion method, the motor vibration features and the wheel swing features are effectively fused; s =att((K l , V l ), Q s ) and A l =att((K s , V s ), Q l ), realize the feature association between the sky wheel swing and the motor vibration through cross attention information; let AF = att((K F , V F ), Q s )+att((K F , V F ), Q l ), by combining attention information, feature complementation between the sky wheel swing and the motor vibration is achieved.

Images