An agricultural nozzle clogging detection method based on acoustic signals

Abstract

The present invention relates to a method for detecting the state of agricultural equipment using acoustic signals, mainly applied to the detection of clogging faults of agricultural nozzles. It includes: Ⅰ. Acquisition of the acoustic time-domain signal of the agricultural nozzle. Ⅱ. Preprocessing of the acoustic signal. Ⅲ. Estimation of the power spectral density. Ⅳ. Estimation of the band power. Ⅴ. Removal of outliers. Ⅵ. Calculation of the band power difference. Ⅶ. Sum of the absolute differences of the band power (SAD). Ⅷ. By comparing with the sum of the absolute power differences without outliers of the acoustic signals of nozzles in different clogging states, the clogging degree of the nozzle is detected and distinguished. The present invention provides a real-time and efficient detection method for the clogging faults that often occur during the spraying operation of agricultural nozzles, and can provide certain guidance for improving the spraying operation quality of agricultural nozzles. The present invention has broad application prospects.

Description

Technical Field

[0001] The present invention relates to a method for detecting the state of agricultural equipment using acoustic signals, and is mainly applied to the detection of clogging faults of agricultural nozzles. Background Art

[0002] As the most critical component of the spraying system, various fault states of agricultural nozzles pose a serious threat to the spraying quality during the spraying operation. Among them, nozzle clogging is the most common nozzle fault. Nozzle clogging will lead to insufficient liquid medicine spraying volume, which will further affect the atomization effect of the liquid medicine by the nozzle and reduce the spraying operation quality. Therefore, the research on detecting whether the agricultural nozzle is clogged and the degree of clogging has important value and significance. At present, there are many ways to detect various equipment, but for working places with a very high humidity like spraying, most detection equipment cannot adapt well. And using acoustic signals for equipment monitoring can well overcome the limitations of these adverse factors. The present invention proposes a method for detecting agricultural nozzle clogging based on acoustic signals. Summary of the Invention

[0003] The purpose of the present invention is to propose a method for detecting agricultural nozzle clogging to detect whether the agricultural nozzle is clogged and the degree of clogging.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions:

[0005] Step 1: Use a microphone array to collect the acoustic signals of the agricultural nozzle in the working state;

[0006] Step 2: Signal preprocessing, that is, use a low-pass filter to preprocess the collected data, including data selection and noise reduction processing;

[0007] Step 3: Use the Welch algorithm to estimate the power spectral density of the nozzle sound signal; in order to reduce the error caused by spectral leakage, a window function is added to the Welch algorithm, and the Welch algorithm is as follows:

[0008]

[0009] In the formula: x(n) is a certain segment of data of the observed data of the nozzle sound signal, and the number of data is N; h(n) is the added window function; e -j2πfn is a complex function, f is the Fourier frequency of the nozzle time-domain sound signal sequence, j is the imaginary unit, j 2 =-1; PSD(f) is the estimated value of the power spectral density of the sound signal sequence calculated by the Welch algorithm; n represents the number of data points for performing the fast Fourier transform; Δt is the sampling period of the nozzle sound signal;

[0010] Step 4. To reduce the error between the power spectral density estimates calculated in Step 3 for the data volumes of different degrees of nozzle blockage and the unblocked data, the nozzle acoustic signal is evenly divided into M frequency bands according to frequency, and the acoustic power in the M frequency bands is calculated:

[0011]

[0012] where, f a and f b respectively represent the lower frequency limit and the upper frequency limit; b[k] is the average acoustic power of the k-th frequency band defined by the lower frequency limit f a and the upper frequency limit f b ; Δf is the frequency resolution;

[0013] Step 5. Considering that the estimated value of the band power of the signal data set will be interfered by various external sudden noises (such as bird calls, cicada chirps, etc.), it is necessary to remove outliers from the band power to reduce the interference caused by sudden noises. To achieve the purpose of removing outliers, a threshold is set for the average power estimates of all frequency bands, and the threshold is determined by the Median Absolute Deviation (MAD) of the average power estimates. The power when the nozzle is blocked is specified in a threshold frequency band, and the frequency band can be expressed as:

[0014]

[0015] where, T d is the set threshold of the k-th frequency band of the nozzle with different degrees of blockage DL under a specific spray pressure (0.2 - 0.5 MPa), DL = 0, 1, 2, and 3, respectively representing that the nozzle filter is unblocked, the nozzle is blocked by 1 / 3, the nozzle is blocked by 1 / 2, and the nozzle is blocked by 3 / 4; B d [k] represents the vector of all average power estimates with a blockage degree of DL in the k-th frequency band; med(B d [k]) is the median of B d [k]; MAD(B d [k]) is the absolute median difference of B d [k].

[0016] The signals outside the frequency band range are marked as outliers and then deleted and not used for the blocked state detection signal. The average processing of the power spectral density estimate PSD of the acoustic signal after removing outliers and the set of band power estimates is used for subsequent analysis.

[0017] Step 6. Further, in order to quantify the difference in the average power of the frequency band caused by different degrees of blockage, the average power of the acoustic signal when the unblocked nozzle is working is used as the average baseline, and the average baseline is subtracted from the estimated value of the average power of the frequency band of the acoustic signal to be classified. The specific calculation is as follows:

[0018]

[0019] Among them, Δb d [k] represents the difference in the frequency band power under different degrees of blockage of the nozzle, and b d [k] is the estimated value of the average power of the frequency band of the nozzle under different degrees of blockage DL, represents the estimated value of the average power of the average frequency band in the unblocked state of the nozzle;

[0020] Step 7. In order to distinguish the acoustic signals of nozzles with different degrees of blockage, the sum of the absolute values of the frequency band power differences in Step 6 is used as an evaluation index for blockage, that is, the sum of the absolute differences of the frequency band power (Sum of AbsoluteDifference (SAD)). The specific calculation is as follows:

[0021]

[0022] In the formula, SAD d represents the sum of the absolute differences of the average power of the frequency band of the acoustic signals of nozzles with different degrees of blockage;

[0023] Step 8. By comparing with the sum of the absolute differences of the power without outliers of the acoustic signals of the nozzles in different blocked states, the degree of blockage of the nozzle can be detected and distinguished.

[0024] Compared with the existing nozzle blockage state detection technology, the beneficial effects of the present invention are as follows:

[0025] 1. Different from using a flow meter and a pipeline pressure gauge or microwave imaging technology to judge the blockage state of the nozzle. The present invention uses a microphone to obtain the acoustic signal of the nozzle to infer the blockage state of the nozzle, which has the advantage of low cost, greatly reduces the cost, and has the advantage of popularization.

[0026] 2. A new detection method and processing steps for detecting nozzle blockage are proposed, providing a reference for popularization and application. Brief Description of the Drawings

[0027] Figure 1 is the structural block diagram of the agricultural nozzle blockage detection method of the present invention;

[0028] Figure 2 is the schematic diagram of the acoustic signal processing program for different degrees of blockage of the agricultural nozzle.

[0029] Table 1 Parameter Table of Test Agricultural Nozzles

[0030] Table 2 Detection results of different clogging degrees of agricultural nozzle filters under different spray pressures Detailed implementation manners

[0031] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0032] The present invention proposes a method for detecting clogging of agricultural nozzles based on sound signals. Specifically, by collecting the working sound signals of agricultural nozzles in different clogging states, and taking the sound power difference analysis as an evaluation index, the nozzle clogging and the detection of the clogging degree are judged.

[0033] The object of the present invention is achieved through the following technical solutions. The implementation steps of the clogging detection of agricultural nozzles based on sound signals are as follows:

[0034] In this paper, 6 different models of 3 types of agricultural nozzles, namely conventional fan nozzles, hollow cone nozzles and drift reduction nozzles, are selected, and the specific technical parameters are shown in Table 1.

[0035] Step 1: Use a microphone sensor array to collect the sound signals of agricultural nozzles in the working state. In this paper, for 3 clogging degrees (D1, D2, D3, as shown in Table 3) of the nozzle filter, 150 data blocks are collected under 3 spray pressures (0.2 MPa, 0.3 MPa, 0.4 MPa).

[0036] Step 2: Use the low-pass filter function (py_lowpass) of python to preprocess the sampled data in Step 1.

[0037] Step 3: Use the Welch algorithm to calculate the estimated value of the nozzle sound power spectrum. More specifically, it can be divided into 3 processes: a. Divide the collected nozzle sound signals into n equal-length data segments, a total of N segments, and each segment contains M data points; b. Apply a window function to each segment of data, that is, assign corresponding weights to the data. The window functions include Hanning window, Hamming window, rectangular window and flat-top window, etc. In this example, the Hamming window is selected; c. Then perform a fast Fourier transform on each segment of data and take the square of the modulus to obtain the power spectrum; d. Finally, add up the power spectra of each segment and divide by the total number of segments N to obtain the estimated value of the average power spectral density, that is

[0038]

[0039] In the formula, PSD[f] is the estimated value of the power spectral density of the time-domain signal sequence, h[n] represents the Hamming window, and x[n] is the time-domain sequence of the signal.

[0040] Step 4: Based on the calculated average power spectral density estimate, divide the average power spectral density evenly into 39 frequency bands according to the frequency range (0 - 19.5 kHz), and calculate the average power in each 500 Hz frequency band. These 39 frequency bands contain more than 90% of the energy of the signal. Calculate the sound power in the 39 frequency bands, aiming to reduce the data volume and the error between the power spectral lines of the unclogged data. Secondly, it can also reduce the data volume and enhance data analysis and visualization: Under different spray pressures and different nozzle clogging degrees, repeat 150 times:

[0041]

[0042] In the formula, b[k] represents the average power of the k-th frequency band defined by frequencies f a and f b where k belongs to one of the 39 frequency bands (for example, k th = 2, f a = 0.5 kHz, f b = 1 kHz), and Δf (Hz) is the frequency resolution; as shown in Appendix Figure 2 Step 4.

[0043] Step 5: Remove outliers from the acoustic signals obtained under all conditions. To identify outliers, a threshold is set for each of the 39 frequency bands of each data set. The threshold is calculated based on the Median Absolute Deviation (MAD) of the power estimate. The threshold corresponding to the clogging degree of a given nozzle and frequency band is defined as three scales of MAD above and below the median of all power estimates corresponding to the given clogging degree and frequency band. The specific calculation is as follows:

[0044]

[0045] In the formula, T d represents the set threshold of the k-th frequency band of the frequency band power corresponding to a certain clogging degree under a certain spray pressure, and B d [k] represents the vector of all power estimates with a clogging degree of DL in the k-th frequency band. The frequency band power estimates above or below the given threshold are marked as outliers, as shown in Appendix Figure 2 Step 5.

[0046] Step 6. To quantify the difference in sound transmission rate caused by the blockage degree of the nozzle filter screen, subtract the estimated value of the band power from the average baseline, where the average baseline corresponds to the average band power of the data subset of the unblocked filter screen measured by the microphone array under the same sampling conditions (filter screen blockage state). The calculation method is the same as that of the average band power in the non-steady flight attitude, and both are calculated according to Step 4. Then, subtract the power of the 39 frequency bands of the unblocked nozzle filter screen calculated from each data block from the power of the 39 frequency bands with different blockage degrees of the nozzle filter screen. The different blockage degrees under each spray pressure are repeated 150 times:

[0047]

[0048] In the formula, Δb d [k](W) is the difference in band power for different blockage degrees d th (d th is D0, D1... D3), is the estimated average band power of the data subset of the unblocked filter screen (D0) under the same sampling conditions, and b d [k] is the estimated value of the band power with different blockage degrees of the filter screen. The so-called average band power estimate is the average of 150 groups of data, as shown in Figure 2 Step 6.

[0049] Step 7. Take the absolute value of the power difference of the 39 frequency bands of the unblocked nozzle filter screen (150 groups each) calculated in Step 6 and the blocked filter screen, and then sum them up, as shown in the following formula:

[0050]

[0051] In the formula, SAD d is the sum of the absolute differences of the data with different blockage degrees. k = 39, and there are 150 sums of absolute differences SAD d data points for each blockage degree, forming a data curve of the sum of the absolute differences with different blockage degrees of the nozzle filter screen. As shown in Figure 2 Step 7.

[0052] Step 8. Store the processed SAD values without outliers and compare them with the SAD data subset in the unblocked state of the nozzle filter screen to detect and distinguish the unblocked nozzle filter screen from different blockage degrees. A nozzle filter screen blockage detection threshold is established, which is equal to three standard deviations above the average value of the data subset in the unblocked state of the nozzle filter screen. Exceeding this set threshold indicates that the nozzle filter screen is blocked.

[0053] According to the test conclusion as Figure 2As shown in Table 3, the proposed method for detecting clogging of agricultural nozzles based on sound signals can well complete the detection of the clogging degree of the filter screens of agricultural nozzles under different spraying pressures, and also has the advantages of real-time efficiency and small sample size, providing a basis for improving spraying efficiency and operation quality.

[0054] Table 1

[0055]

[0056] Table 2

[0057]

Claims

1. A method for detecting clogging of a plant protection fan-shaped nozzle based on acoustic signals, characterized in that, It includes the following steps: Step 1: Use a microphone array to collect the acoustic signal of an agricultural nozzle in a working state; Step 2: Signal preprocessing, that is, use a low-pass filter to preprocess the collected data, including data selection and noise reduction processing; Step 3: Use the Welch algorithm to estimate the power spectral density of the nozzle acoustic signal; in order to reduce the error caused by spectral leakage, a window function is added to the Welch algorithm, and the Welch algorithm is as follows: Where: x(n) is a certain section of the observed data of the nozzle acoustic signal, and the number of data is N; h(n) is the applied window function; e -j2πfn is a complex function, f is the Fourier frequency of the nozzle time-domain acoustic signal sequence, j is the imaginary unit, j 2 = -1; PSD(f) is the estimated value of the power spectral density of the acoustic signal sequence calculated by the Welch algorithm; n represents the number of data points for the fast Fourier transform; Δt is the sampling period of the nozzle acoustic signal; Step 4: In order to reduce the error between each spectral line of the data volume used to detect different clogging degrees of the nozzle and the data of unclogged nozzles, based on the power spectral density estimation value calculated in Step 3, the acoustic signal of the nozzle is evenly divided into M frequency bands according to frequency, and the acoustic power on M frequency bands is calculated: where f a and f b represent the lower frequency limit and the upper frequency limit respectively; b[k] is the average sound power of the k-th frequency band defined by the lower frequency limit f a and the upper frequency limit f b ; Δf is the frequency resolution. Step 5: Considering that the estimated value of the band power of the signal data set will be interfered by various external sudden noises, it is necessary to remove outliers from the band power to reduce the interference caused by sudden noises; in order to achieve the purpose of removing outliers, thresholds are set for the average power estimation values of all bands, and the setting of the threshold is determined by the Median Absolute Deviation (MAD) of the average power estimation value; the power when the nozzle is clogged is specified in a threshold band, and the band is expressed as: where, T d is the set threshold of the k-th band of nozzles with different clogging degrees DL under a specific spray pressure of 0.2 to 0.5 MPa; B d [k] represents the vector of all average power estimates with a clogging degree of DL in the k-th band; med(B d [k]) is the median of B d [k]; MAD(B d [k]) is the absolute median difference of B d [k]; Signals outside the band range are marked as outliers and then deleted and not used for detecting the clogging state signal. The power spectral density estimation value PSD and the set of band power estimation values of the acoustic signal after removing outliers are averaged for subsequent analysis; Step 6: Further, in order to quantify the difference in the average band power caused by different clogging degrees, the average band power of the acoustic signal when the unclogged nozzle is working is used as the average baseline, and the average baseline is subtracted from the average band power estimation value of the acoustic signal to be classified. The specific calculation is as follows: Among them, Δb d [k] represents the band power difference under different nozzle blockage conditions, and b d [k] is the estimated value of the band average power of the nozzle under different blockage degrees DL, represents the estimated value of the average band average power in the unblocked state of the nozzle; Step 7: In order to distinguish the acoustic signals of nozzles with different clogging degrees, the absolute value of the band power difference in Step 6 is summed as the evaluation index for clogging, that is, the sum of the absolute differences of the band power (Sum of Absolute Difference (SAD)). The specific calculation is as follows: In the formula, SAD d represents the sum of the absolute differences of the band-average powers of the acoustic signals of nozzles with different clogging degrees; Step 8: By comparing with the sum of the absolute power differences without outliers of the acoustic signals of nozzles in different clogging states, the clogging degree of the nozzle is detected and distinguished.

2. The method for detecting clogging of a plant protection fan-shaped nozzle based on acoustic signals according to claim 1, characterized in that, h(n) includes Hanning window, Hamming window, rectangular window and flat-top window.

3. The method for detecting clogging of a plant protection fan-shaped nozzle based on acoustic signals according to claim 1, characterized in that, DL = 0, 1, 2 and 3, representing that the nozzle filter is unclogged, the nozzle is clogged by 1 / 3, the nozzle is clogged by 1 / 2 and the nozzle is clogged by 3 / 4 respectively.

Images