Sound wave type tension detection method based on signal gradient attenuation

Abstract

The invention provides a sound wave type tension detection method based on signal gradient attenuation, and the method comprises the steps: collecting an audio signal through a digital microphone audio collection module when a belt is in a vibration state; the signal processing and control module performs tension detection according to the collected audio signal to obtain a preprocessed audio signal; detecting to obtain an effective audio signal; framing the signal, and performing fast Fourier transform (FFT) processing on each frame of signal to convert M frames of time domain signals into M frames of frequency domain data; m frames of amplitude-frequency data are obtained; analyzing and searching the main peak values and obtaining frequencies corresponding to the main peak values of the M frames; based on the relationship between signal gradient attenuation and integer multiplying power, after characteristic vibration frequency is extracted from the frequencies corresponding to the main peak values of the M frames, the characteristic vibration frequency is converted to obtain belt tension; the characteristic frequency is determined based on the signal gradient attenuation, then the belt tension is obtained, and the efficiency and accuracy of detecting the belt tension can be improved.

Description

Technical Field

[0001] The present invention relates to a sound wave type tension detection method based on signal gradient attenuation, and belongs to the technical field of tension detection. Background Art

[0002] In the early stage of industrial production, the detection of belt tension mainly relied on manual experience and simple mechanical tools. For example, workers judged the tightness of the belt by pressing the belt with their hands and relying on their sense of touch. The background of this method was that industrial equipment was relatively simple at that time and the accuracy requirements for belt tension were not high. Moreover, this intuitive judgment method could meet the basic production needs in some small-scale and low-speed equipment. With the development of industry, some tension detection tools based on mechanical principles, such as tension meters, emerged. Its principle is to convert the tension of the belt into the deflection of a pointer through a mechanical structure, so that the tension value can be intuitively read. The emergence of this tool was to meet the occasions with certain requirements for tension detection accuracy, such as some textile machinery and simple conveying equipment.

[0003] When industrial production entered the automation stage, the requirements for the accuracy and real-time performance of belt tension detection were greatly improved. Contact electronic tension meters came into being. The background was the booming development of electronic technology, which could use electronic components such as strain gauges to convert the tension of the belt into an electrical signal. The strain gauge was installed in the probe of the tension meter. When the probe contacted the belt and was subjected to tension, the resistance of the strain gauge changed, and this change was converted into a voltage or current signal through a circuit, and then the tension value was displayed. This technology was widely used in industries such as automobile manufacturing and printing. For example, on an automobile production line, the tension of the conveyor belt needs to be precisely controlled to ensure the accurate conveyance of automobile parts, and the contact electronic tension meter can meet this high-precision and real-time detection requirement.

[0004] As industrial equipment develops towards high speed, high precision, and complex environments, traditional contact detection technologies have encountered bottlenecks. On the one hand, contact detection may cause damage to the belt, affecting the service life of the belt and the performance of the equipment. On the other hand, in some high-speed rotating belt systems or belt positions that are difficult to access, contact detection is difficult to operate. In response to the above problems, non-contact detection technologies, such as sound wave type and light wave type belt tension detection technologies, emerged under the background of the development of multi-disciplinary technologies such as acoustics, optics, and signal processing. Among them, the sound wave type detection technology uses the sound waves generated by the vibration of the belt to detect tension, and its background is the progress of acoustic measurement technology, which can accurately capture and analyze acoustic signals. However, the efficiency and accuracy of existing sound wave type detection methods are both low. Summary of the Invention

[0005] The object of the present invention is to provide a sound wave - type tension detection method based on signal gradient attenuation to solve the problems of how to improve the efficiency and accuracy of detecting belt tension in the prior art.

[0006] The technical solution of the present invention is as follows:

[0007] A sound wave - type tension detection method based on signal gradient attenuation, comprising the following steps:

[0008] S1. When the belt is in a vibrating state, the digital microphone audio acquisition module acquires an audio signal and sends it to the signal processing and control module;

[0009] S2. The signal processing and control module performs tension detection according to the acquired audio signal. Specifically:

[0010] S21. Pre - process the acquired audio signal to obtain the pre - processed audio signal;

[0011] S22. After detecting the effective audio signal through endpoint detection (EPD) for the pre - processed audio signal, enter the next step S23; otherwise, return to step S1;

[0012] S23. Frame the effective audio signal through a first - in - first - out memory (FIFO) to obtain M frames of time - domain signals s(n), and then perform fast Fourier transform (FFT) processing on each frame of the signal to convert the M frames of time - domain signals into M frames of frequency - domain data;

[0013] S24. Use the coordinate rotation digital computer (CORDIC) method to take the modulus of the complex results of the M frames of frequency - domain data to obtain M frames of amplitude - frequency data;

[0014] S25. Analyze each frame of the amplitude - frequency data through peak detection to find the main peak and obtain the corresponding frequencies of the main peaks of the M frames;

[0015] S26. Based on the relationship between signal gradient attenuation and integer multiple, extract the characteristic vibration frequency from the corresponding frequencies of the main peaks of the M frames, and then convert the characteristic vibration frequency to obtain the belt tension.

[0016] Furthermore, the digital microphone audio acquisition module uses a micro - electro - mechanical system microphone, i.e., a MEMS microphone.

[0017] Furthermore, in step S22, when detecting the effective audio signal through endpoint detection (EPD) for the pre - processed audio signal, specifically:

[0018] Set a threshold V th as: V th =(V MAX -V MIN )*K + V MIN, where V MAX is the maximum value in the preprocessed audio signal, and V MIN is the minimum value in the preprocessed audio signal, and K is the sensitivity parameter;

[0019] When the audio signal is greater than the set threshold V th , it is a valid audio signal; otherwise, it is a silent and invalid audio signal.

[0020] Furthermore, in step S23, the fast Fourier transform FFT processing formula is as follows:

[0021]

[0022] where F[k] is the frequency-domain data, k represents the frequency component, s(n) is the time-domain signal of the nth sampling point, N is the total number of sampling points of the fast Fourier transform FFT, e is the natural constant, and j is the imaginary unit.

[0023] Furthermore, step S24 is specifically

[0024] S241. The complex result of the frequency-domain data F[k] is specifically expressed as:

[0025] F[k] = x(k) + jy(k), k = 0, 1,..., N - 1

[0026] where x(k) is the real component of the frequency-domain data F[k], j is the imaginary unit, and y(k) is the imaginary component of the frequency-domain data F(k);

[0027] S242. The corresponding vector of the frequency-domain data F[k] using the vector mode of the coordinate rotation digital calculation method CORDIC is (x(k) i , y(k) i ), where x(k) i and y(k) i are the x coordinate and y coordinate after rotating i times respectively. The vector (x(k) i , y(k) i ) is successively rotated in the positive direction of the X axis, and the rotation angle is θ i ,:

[0028]

[0029] where x(k) i+1 and y(k) i+1 are the x coordinate and y coordinate after rotating i + 1 times respectively, and d i is the rotation direction;

[0030] S243. For the convenience of calculation, let tanθ i = 2 -i, and cosθ is omitted i After that, we get:

[0031]

[0032] Among them, if the y coordinate y(k) after rotating i times i is greater than 0, the rotation direction d i = -1, and perform clockwise rotation; if the y coordinate y(k) after rotating i times i y(k) i is less than 0, the rotation direction d i = 1, and perform counterclockwise rotation;

[0033] S244. During pseudo-rotation, cosθ i is ignored, and the vector is scaled by 1 / cosθ i times. When the total number of rotations q is known, calculate the scaling factor K q ;

[0034] S245. As the number of rotations increases to q times, the vector (x(k) q , y(k) q ), where x(k) q , y(k) q are the x coordinate and y coordinate after rotating q times respectively, approaches the positive direction of the X-axis. At this time, the amplitude-frequency data A(k) is approximately expressed as:

[0035] A(k) = |F[k]| ≈ x(k) q *K q , k = 1, 2, …, N.

[0036] Furthermore, in step S244, calculate the scaling factor K q :

[0037]

[0038] Among them, q is the total number of rotations, and the number of rotations i = 0, 1, … q. When q -> ∞, the scaling factor K q -> 1.64676.

[0039] Furthermore, step S25 is specifically

[0040] S251. Traverse the amplitude-frequency data of the first frame, and find all the maximum values A 极 (k), k ∈ [1, N], to form a maximum value group;

[0041] S252. Find the maximum value A 极 (k) max in the maximum value group, and save it as a main peak P(kp ), where k p is the k value corresponding to the largest maximum. According to the peak spacing d, remove all the maxima in the maximum value group within the range of the maximum value A 极 (k) max in the range of [k p - d, k p + d];

[0042] S253. Repeat the above step S252 for m - 1 times to obtain m main peaks P(k p ), and the k values k p corresponding to the m main peaks;

[0043] S254. Sort the k values k p corresponding to the m main peaks obtained in step S253 from small to large to get k p (1), k p (2),... k p (i),..., k p (m).

[0044] Furthermore, step S26 is specifically

[0045] S261. Based on the integer multiple relationship between the fundamental frequency and the harmonic frequency of the signal, and the relationship between the frequency f i and the k value k p (i) corresponding to the i-th main peak:

[0046]

[0047] where F s is the sampling frequency of the microphone, and N is the total number of sampling points of the fast Fourier transform FFT;

[0048] The k value k 1 corresponding to the first main peak of the characteristic vibration frequency f p (1) should satisfy condition 1:

[0049] k p (i) = Z * k p (1), i = 1, 2,... m

[0050] where Z is an integer;

[0051] S262. Based on the signal vibration gradient attenuation relationship, the high-frequency signal decays faster with time. Then the amplitude change of the main peaks corresponding to the M-th frame and the first frame

[0052]

[0053] where A[kp (·)] 1 The amplitude-frequency data corresponding to the main peak of the first frame, A[k p (i)] M is the k value corresponding to the main peak of the Mth frame, k p (i); the corresponding amplitude-frequency data

[0054] The characteristic vibration frequency f 1 The corresponding amplitude change should also satisfy Condition 2:

[0055]

[0056] S263. After satisfying Condition 1 in Step S261 and Condition 2 in Step S261, the characteristic vibration frequency f 1 is obtained, and the next step S264 is entered; otherwise, return to Step S1;

[0057] S264. The characteristic vibration frequency f 1 is converted to obtain the belt tension T by the following formula:

[0058]

[0059] where M is the unit mass of the belt, W is the width of the belt, and L is the effective length of the belt interval.

[0060] The beneficial effects of the present invention are as follows: This acoustic wave-based tension detection method based on signal gradient attenuation determines the characteristic frequency based on signal gradient attenuation, and then obtains the belt tension, which can improve the efficiency and accuracy of detecting the belt tension. Further, the MEMS microphone is used to accurately collect sound signals, which can capture tiny vibration signals, improve the detection accuracy, can achieve non-contact measurement, reduce belt wear, improve measurement safety, is easy to operate, can adapt to various belt types, and can be used in various environments, solving the problems of complex operation and low efficiency of traditional detection methods. Description of the Drawings

[0061] Figure 1 is a flowchart of the acoustic wave-based tension detection method based on signal gradient attenuation according to an embodiment of the present invention;

[0062] Figure 2 is an explanatory diagram of the audio signal acquisition by the digital microphone audio acquisition module in the embodiment;

[0063] Figure 3 is a comparison diagram of the audio signal before pre-emphasis in the embodiment, where (a) is the audio signal before pre-emphasis and (b) is the audio signal after pre-emphasis;

[0064] Figure 4Schematic diagram of the amplitude-frequency data of the first frame in a specific example of the embodiment;

[0065] Figure 5 Schematic diagram of the amplitude-frequency data of the Mth frame in a specific example of the embodiment. Detailed implementation manners

[0066] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0067] Embodiment A sound wave type tension detection method based on signal gradient attenuation, as Figure 1 , includes the following steps,

[0068] S1. When the belt is in a vibrating state, the digital microphone audio acquisition module acquires an audio signal and sends it to the signal processing and control module, as Figure 2 .

[0069] In step S1, the digital microphone audio acquisition module uses a microelectromechanical system microphone, i.e., a MEMS microphone. Using a MEMS microphone can accurately acquire sound signals, capture tiny vibration signals, and improve the detection sensitivity and detection range.

[0070] S2. The signal processing and control module performs tension detection based on the acquired audio signal. Specifically:

[0071] S21. Preprocess the acquired audio signal to obtain the preprocessed audio signal.

[0072] In step S21, the preprocessing includes pre-emphasis. The formula for pre-emphasis is: V[n] = v[n] - αv[n - 1], where V[n] is the nth audio signal after pre-emphasis, v[n] and v[n - 1] are the nth and (n - 1)th audio signals before pre-emphasis respectively, and α is the pre-emphasis coefficient, with a value range of: 0.9 - 1. In this embodiment, α is taken as 0.97. The audio signal before pre-emphasis is as Figure 3 (a), and the audio signal after pre-emphasis is as Figure 3 (b).

[0073] S22. After obtaining the effective audio signal by endpoint detection EPD for the preprocessed audio signal, enter the next step S23; otherwise, return to step S1.

[0074] In step S22, for the preprocessed audio signal, obtaining the effective audio signal by endpoint detection EPD is specifically:

[0075] Set the threshold V th as: V th =(V MAX -V MIN )*K + VMIN , where V MAX is the maximum value in the preprocessed audio signal, and V MIN is the minimum value in the preprocessed audio signal, and K is the sensitivity parameter;

[0076] When the audio signal is greater than the set threshold V th , it is a valid audio signal; otherwise, it is a silent invalid audio signal.

[0077] S23. Frame the valid audio signal through a first-in first-out memory FIFO to obtain M frames of time-domain signals s(n), and then perform fast Fourier transform (FFT) processing on each frame of the signal to transform the M frames of time-domain signals into M frames of frequency-domain data.

[0078] In step S23, the fast Fourier transform (FFT) processing formula is as follows:

[0079]

[0080] where F[k] is the frequency-domain data, k represents the frequency component, s(n) is the time-domain signal at the nth sampling point, N is the total number of sampling points of the fast Fourier transform (FFT), e is the natural constant, and j is the imaginary unit.

[0081] S24. Use the coordinate rotation digital calculation method (CORDIC) to take the modulus of the complex results of the M frames of frequency-domain data to obtain M frames of amplitude-frequency data. As Figure 4 and Figure 5 , Figure 4 is a schematic diagram of the amplitude-frequency data of the first frame in a specific example of the embodiment; Figure 5 is a schematic diagram of the amplitude-frequency data of the Mth frame in a specific example of the embodiment.

[0082] S241. The complex result of the frequency-domain data F[k] is specifically expressed as:

[0083] F[k] = x(k) + jy(k), k = 0, 1,..., N - 1

[0084] where x(k) is the real component of the frequency-domain data F[k], j is the imaginary unit, and y(k) is the imaginary component of the frequency-domain data F(k);

[0085] S242. The corresponding vector of the frequency-domain data F[k] using the vector mode of the coordinate rotation digital calculation method (CORDIC) is (x(k) i , y(k) i ), where x(k) i , y(k) i are the x-coordinate and y-coordinate after rotating i times respectively. The vector (x(k) i , y(k)i ) Rotate successively in the positive direction of the X-axis by an angle of θ i ,:

[0086]

[0087] where x(k) i+1 and y(k) i+1 are the x-coordinate and y-coordinate after the (i + 1)-th rotation respectively, and d i is the rotation direction;

[0088] S243. For the convenience of calculation, let tanθ i = 2 -i and omit cosθ i to obtain:

[0089]

[0090] where if the y-coordinate y(k) i after the i-th rotation is greater than 0, the rotation direction d i = -1, and perform a clockwise rotation; if the y-coordinate y(k) i y(k) i after the i-th rotation is less than 0, the rotation direction d i = 1, and perform a counterclockwise rotation;

[0091] S244. During pseudo-rotation, cosθ i is ignored, and the vector is scaled by 1 / cosθ i times. When the total number of rotations q is known, calculate the scaling factor K q : where q is the total number of rotations, the number of rotations i = 0, 1, …, q. When q -> ∞, the scaling factor K q -> 1.64676.

[0092] S245. As the number of rotations increases to q times, the vector (x(k) q , y(k) q ), where x(k) q and y(k) q are the x-coordinate and y-coordinate after the q-th rotation respectively, approaches the positive direction of the X-axis. At this time, the amplitude-frequency data A(k) is approximately expressed as:

[0093] A(k) = |F[k]| ≈ x(k) q * K q , k = 1, 2, …, N.

[0094] S25. Analyze the amplitude-frequency data of each frame through peak detection to find the main peaks and obtain the frequencies corresponding to the main peaks of M frames.

[0095] S251. Traverse the amplitude-frequency data of the first frame, and find all the maximum values A 极 (k), k ∈ [1, N], to form a group of maximum values;

[0096] S252. Find the maximum value A 极 (k) max in the group of maximum values, and save it as a main peak P(k p ), where k p is the k value corresponding to the largest maximum value. According to the peak spacing d, remove all the maximum values in the group of maximum values within the range [k 极 (k) max of the maximum value A p - d, k p + d];

[0097] S253. Repeat the above step S252 for m - 1 times, and m main peaks P(k p ) and the k values k p corresponding to the m main peaks can be obtained;

[0098] S254. Sort the k values k p corresponding to the m main peaks obtained in step S253 from small to large to get k p (1), k p (2),..., k p (i),..., k p (m).

[0099] Figure 4 is a schematic diagram of the amplitude-frequency data of the first frame in a specific example of the embodiment. After the data in Figure 4 undergoes steps S251 - S254, the k values k p corresponding to the m main peaks are sorted from small to large to obtain 65, 130, 196, 260, and 325 respectively.

[0100] S26. Based on the relationship between signal gradient attenuation and integer multiple, extract the characteristic vibration frequency from the frequencies corresponding to the main peaks of M frames, and then convert the characteristic vibration frequency to obtain the belt tension.

[0101] S261. Based on the integer multiple relationship between the fundamental frequency and harmonic frequency of the signal, and the relationship between the frequency f i and the k value k p (i) corresponding to the i-th main peak:

[0102]

[0103] where, F sis the sampling frequency of the microphone, and N is the total number of sampling points of the Fast Fourier Transform (FFT);

[0104] Characteristic vibration frequency f 1 The corresponding first main peak corresponds to the k value k p (1) It should satisfy Condition 1:

[0105] k p (i) = Z * k p (1), i = 1, 2, … m

[0106] where Z is an integer;

[0107] S262. Based on the signal vibration gradient attenuation relationship, the high-frequency signal decays faster with time, so the amplitude change corresponding to the main peak between the Mth frame and the first frame

[0108]

[0109] where A[k p (i)] 1 is the amplitude-frequency data corresponding to the main peak of the first frame, and A[k p (i)] M is the amplitude-frequency data corresponding to the k value k p (i) corresponding to the main peak of the Mth frame;

[0110] Characteristic vibration frequency f 1 The corresponding amplitude change should also satisfy Condition 2:

[0111]

[0112] S263. After satisfying Condition 1 in step S261 and Condition 2 in step S261, the characteristic vibration frequency f 1 is obtained and proceeds to the next step S264; otherwise, return to step S1;

[0113] S264. The characteristic vibration frequency f 1 is converted to obtain the belt tension T by the following formula:

[0114]

[0115] where M is the unit mass of the belt, W is the width of the belt, and L is the effective length of the belt interval, such as the belt length between two adjacent gears.

[0116] This acoustic tension detection method based on signal gradient attenuation determines the characteristic frequency based on signal gradient attenuation, and then obtains the belt tension, which can improve the efficiency and accuracy of belt tension detection. Further, a MEMS microphone is used to accurately collect sound signals, which can capture tiny vibration signals, improve detection accuracy, enable non-contact measurement, reduce belt wear, improve measurement safety, is easy to operate, can adapt to various belt types, and can be used in various environments, solving the problems of complex operation and low efficiency of traditional detection methods.

[0117] This acoustic tension detection method based on signal gradient attenuation includes a digital microphone audio acquisition module and a signal processing and control module. The digital microphone audio acquisition module is connected to the signal processing and control module. The signal processing and control module controls the digital microphone audio acquisition module to collect the sound signals generated by the vibration of the external belt, and the signal processing and control module is used to perform tension detection based on the sound signals collected by the digital microphone audio acquisition module. Among them, the digital microphone audio acquisition module includes a MEMS sensor module, an amplifier module, an analog-to-digital conversion module ADC, and an IIS interface module. The collected audio signals are amplified and converted from analog to digital and then sent to the signal processing and control module. The MEMS sensor module converts the sound signals into electrical signals, the amplifier module is responsible for increasing the signal amplitude, enhancing the signal driving ability, and optimizing the signal quality, the analog-to-digital conversion module ADC is responsible for converting the analog signals into digital signals, and the IIS interface module is responsible for communicating with the signal processing and control module.

[0118] The signal processing and control module uses an FPGA as a controller and processor to implement signal processing and control. The signal processing and control module includes a tension conversion module, a signal processing module, and a microphone control module. The microphone control module is connected to the IIS interface module to achieve the control of the digital microphone acquisition module and obtain data from the digital microphone acquisition module. The signal processing module is responsible for processing and analyzing the data to obtain the characteristic vibration frequency of the belt, and the tension conversion module combines the obtained characteristic vibration frequency to convert and obtain the belt tension. The present invention is based on the FPGA platform and quantifies the belt tension by collecting the frequency of the sound signals generated by the vibration of the belt.

[0119] This acoustic tension detection method based on signal gradient attenuation can improve the efficiency and accuracy of belt tension detection by improving the way of testing the belt tension. By collecting the frequency of the sound signals generated by the vibration of the belt, quantifying the belt tension, and verifying the characteristic frequency, the detection accuracy can be improved. It can achieve non-contact measurement, reduce belt wear, improve measurement safety, is easy to operate, adapts to various belt types, and can be used in various environments, solving the problems of complex operation and low efficiency of traditional detection methods.

[0120] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A sonic tension detection method based on signal gradient attenuation, characterized in that: The following steps are included: S1. When the belt is in a vibrating state, the digital microphone audio acquisition module collects audio signals and sends them to the signal processing and control module; S2, the signal processing and control module performs tension detection according to the collected audio signal, specifically: S21, preprocessing the collected audio signal to obtain a preprocessed audio signal; S22, after obtaining a valid audio signal through endpoint detection EPD detection of the pre-processed audio signal, proceed to the next step S23; Otherwise, return to step S1; S23, dividing the effective audio signal into frames through a first-in-first-out memory FIFO to obtain M frames of time domain signals s(n), and then performing fast Fourier transform FFT processing on each frame signal to convert the M frames of time domain signals into M frames of frequency domain data; S24, using a coordinate rotation digital calculation method CORDIC to modulo the complex result of the frequency domain data of the M frames to obtain the amplitude-frequency data of the M frames; S25, analyzing the amplitude-frequency data of each frame through peak detection to find the main peak and obtain the main peak corresponding frequency of M frames; S26. Based on the relationship between signal gradient attenuation and integer multiples, after extracting the characteristic vibration frequency from the frequency corresponding to the main peak of the M frame, the characteristic vibration frequency is converted to obtain the belt tension.

2. The method for ultrasonic tension detection based on signal gradient attenuation according to claim 1, characterized in that: The digital microphone audio acquisition module uses a micro-electromechanical system microphone, namely a MEMS microphone.

3. The method for ultrasonic tension detection based on signal gradient attenuation according to claim 1, characterized in that: In step S22, the pre-processed audio signal is subjected to endpoint detection (EPD) to obtain a valid audio signal, specifically, Set the threshold V th =V th =(V MAX -V MIN )*K+V MIN , where V MAX is the maximum value of the preprocessed audio signal, V MIN is the minimum value in the preprocessed audio signal, and K is the sensitivity parameter; When the audio signal is greater than the set threshold V th , it is a valid audio signal; otherwise, it is a muted invalid audio signal.

4. The method for ultrasonic tension detection based on signal gradient attenuation according to claim 1, characterized in that: In step S23, the fast Fourier transform FFT processing formula is as follows: Wherein, F[k] is the frequency domain data, k represents the frequency component, s(n) is the time domain signal of the nth sampling point, N is the total number of sampling points of the fast Fourier transform FFT, e is a natural constant, and j is an imaginary unit.

5. The method for ultrasonic tension detection based on signal gradient attenuation according to any one of claims 1 to 4, characterized in that: Step S24, specifically, S241, the complex result of the frequency domain data F[k] is specifically expressed as: F[k]=x(k)+jy(k),k=0,1,…,N-1 Wherein, x(k) is the real component of the frequency domain data F[k], j is the imaginary unit, and y(k) is the imaginary component of the frequency domain data F(k); S242, the frequency domain data F[k] uses the coordinate rotation digital calculation method CORDIC vector mode corresponding vector is (x(k) i ,y(k) i ), where x(k) i ,y(k) i are the x-coordinate and y-coordinate after rotation i times respectively, and the vector (x(k) i ,y(k) i ) rotates in the positive direction of the X axis, and the rotation angle is θ i ,: Among them, x(k) i+1 ,y(k) i+1 are the x-coordinate and y-coordinate after rotation i+1 times, d i is the direction of rotation; S243, for the convenience of calculation, let tanθ i =2 -i , and omit cosθ i After that, we get: Among them, if the y coordinate y(k) after rotation i times i If it is greater than 0, the rotation direction is d i = -1, rotate clockwise; if the y coordinate after rotation i times is y(k) i y(k) i Less than 0, the rotation direction d i =1, rotate counterclockwise; S244, pseudo rotation, cosθ i is ignored, the vector is scaled by 1 / cosθ i times, when the total number of rotations q is known, calculate the scaling factor K q ; S245, as the number of rotations increases to q, the vector (x(k) q ,y(k) q ), where x(k) q ,y(k) q are the x-coordinate and y-coordinate after rotation q times, approaching the positive direction of the X-axis. At this time, the amplitude-frequency data A(k) is approximately expressed as: A(k)=|F[k]|≈x(k) q *K q ,k=1,2,…,N。 6. The method for ultrasonic tension detection based on signal gradient attenuation according to claim 5, characterized in that: In step S244, the scaling factor K is calculated q : Where q is the total number of rotations, the number of rotations i = 0, 1, ... q, when q->∞, the scaling factor K q ->1.64676.

7. The method for ultrasonic tension detection based on signal gradient attenuation according to any one of claims 1 to 4, characterized in that: Step S25, specifically, S251, traverse the amplitude-frequency data of the first frame, and find all the maximum values ​​A in the amplitude-frequency data of the first frame 极 (k), k∈[1,N], constitutes a maximum value group; S252. Find the maximum value A in the maximum value group 极 (k) max , saved as a main peak P(k p ), where k p The k value corresponding to the largest maximum value is removed according to the peak spacing d at the maximum value A in the maximum value group. 极 (k) max The range of [k p -d,k p +d]; S253, repeat the above step S252 m-1 times to obtain m main peaks P(k p ), and the m main peaks correspond to k values ​​k p ; S254, the m main peaks obtained in step S253 correspond to k values ​​k p Sort from small to large to get k p (1), k p (2), ...k p (i), ..., k p (m).

8. The method for ultrasonic tension detection based on signal gradient attenuation according to any one of claims 1 to 4, characterized in that: Step S26, specifically, S261, based on the integer multiple relationship between the fundamental frequency and the harmonic frequency of the signal, and the frequency f i The k value corresponding to the i-th main peak is k p (i) The relationship between: Among them, F s is the sampling frequency of the microphone, and N is the total number of sampling points of the fast Fourier transform FFT; The first main peak corresponding to the characteristic vibration frequency f1 corresponds to the k value k p (1) Condition 1 must be met: k p (i)=Z*k p (1),i=1,2,…m Wherein, Z is an integer; S262, based on the signal vibration gradient attenuation relationship, the high-frequency signal decays faster over time, so the amplitude change corresponding to the main peak of the Mth frame and the first frame ▽A[k p (i)]: ▽A[k p (i)]=A[k p (i)]1-A[k p (i)] M ,i=1,2,…,m Among them, A[k p (i)]1 is the amplitude-frequency data corresponding to the main peak of the first frame, A[k p (i)] M The k value corresponding to the main peak of the Mth frame is k p (i) Corresponding amplitude-frequency data; The amplitude change corresponding to the characteristic vibration frequency f1 is ▽A[k p (1)] The second condition must also be met: ▽A[k p (1)]<▽A[k p (i)],i=2,3,…,m; S263, after satisfying the condition 1 in step S261 and the condition 2 in step S261, obtaining the characteristic vibration frequency f1, and proceeding to the next step S264; otherwise, returning to step S1; S264. Convert the characteristic vibration frequency f1 to obtain the belt tension T using the following formula: Where M is the unit mass of the belt, W is the belt width, and L is the effective length of the belt interval.

Images