Method and device for detecting quality of key component of acoustic non-contact precision dibbler

Abstract

The invention discloses an acoustic non-contact precision dibbler key component quality detection method and device, and relates to the technical field of mechanical fault detection, and the method comprises the steps: 1, collecting acoustic signals generated in the operation process of a precision dibbler key component; 2, decomposing the acquired acoustic signals by adopting a resonance decomposition method, and optimizing resonance decomposition parameters of the resonance decomposition method by adopting a grey wolf optimization algorithm to obtain an optimal high-resonance component and an optimal low-resonance component; 3, envelope spectrum extraction is carried out based on the optimal high resonance component and the optimal low resonance component, Fourier transform is carried out on the extracted envelope spectrum, a frequency spectrum is obtained, fault diagnosis is carried out according to the frequency spectrum, and a quality detection result is obtained. According to the invention, the detection efficiency of the precision dibbler can be effectively improved, so that high-coverage detection is realized, and meanwhile, the detection precision can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault detection, and more specifically, to a method and device for detecting the quality of key components of an acoustic non-contact precision hill-drop planter. Background Art

[0002] Currently, with the wide application of precision seeding technology, the market demand for mechanical hill-drop planters continues to grow. As the core component of precision seeding, the performance and quality of the hill-drop planter directly determine the seeding effect. Due to the strong timeliness of seeding, higher requirements are put forward for the production cycle and quality of hill-drop planter manufacturers, and ensuring the ex-factory quality of hill-drop planters has become a key link in enterprise production. However, the ex-factory quality inspection of hill-drop planters relies on the detection method of simulating seeding operation with seeds in the seedbed. This method has the deficiencies of long time-consuming for loading and taking seeds and relying on manual tracking detection. Therefore, sampling inspection is mostly used in actual production, which leads to low detection coverage and high repair costs.

[0003] In addition, when the mechanical hill-drop planter works in the traditional detection method, a large amount of non-linear pulse impact noise will be generated by the repeated impact of the crank arm and the keel plate, and the fault signal will be submerged by the non-linear pulse noise generated by the impact. Such a situation brings great challenges to the quality inspection link of the mechanical hill-drop planter and affects the accuracy of detection.

[0004] Therefore, how to improve the detection coverage and accuracy of hill-drop planters is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and device for detecting the quality of key components of an acoustic non-contact precision hill-drop planter, which effectively improves the detection efficiency of the precision hill-drop planter, thereby realizing high-coverage detection, and at the same time can improve the detection accuracy.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for detecting the quality of key components of an acoustic non-contact precision hill-drop planter includes the following steps:

[0008] Step 1: Collect the acoustic signals generated during the operation of the key components of the precision hill-drop planter;

[0009] Step 2: Decompose the collected acoustic signals by using the resonance decomposition method, and optimize the resonance decomposition parameters of the resonance decomposition method by using the grey wolf optimization algorithm to obtain the optimal high-resonance component and the optimal low-resonance component;

[0010] Step 3: Based on the periodic noise pulse components in the optimal high resonance component and the periodic fault components with non-linear noise suppression in the optimal low resonance component, perform envelope spectrum processing extraction, and perform a fast Fourier transform on the extracted envelope spectrum to obtain its frequency spectrum. Conduct fault diagnosis based on the frequency spectrum, and finally obtain the quality inspection result.

[0011] The technical effect of the above technical solution is to use the resonance decomposition method to separate the high and low frequency resonance components of the oscillation signal, and separate the weak fault signal generated by the assembly fault from the background of the continuous impact noise generated by the crank arm hitting the keel piece during the normal operation of the precision hill-drop planter, improving the accuracy and efficiency of quality inspection. The fault-free pulse impact noise signal in the acoustic signal is decomposed into high resonance components using the difference in resonance attributes, and the weak coupled fault pulse signal in the acoustic signal is also decomposed into low resonance components using the difference in resonance attributes; the corresponding fault frequency is found according to the frequency with a high peak in the frequency spectrum, and the fault corresponding to the fault frequency is accurately identified.

[0012] Preferably, the resonance decomposition parameters include the quality factor Q, redundancy r, and decomposition layer number J. Set the initial high quality factor Q H , high decomposition layer number J H , low quality factor Q L , low decomposition layer number J L and redundancy r; the resonance decomposition method performs two-channel filter bank sparse decomposition, adjustable Q-factor wavelet transform, and morphological component analysis on the acoustic signal according to the resonance decomposition parameters, controls the frequency domain and time domain characteristics of the wavelet, and separates the high resonance component and the low resonance component. According to the set resonance decomposition parameters, signal decomposition can be achieved at different scales, and the control of the frequency response frequency domain overlap at adjacent scales can be obtained, so as to achieve the precise separation of different frequency components.

[0013] The technical effects of the above technical solution are as follows: The resonance decomposition method is used to decompose the continuous oscillation signal according to the differences of different components in the original oscillation signal. The high and low quality factors are reasonably selected, and the decomposition layer number J under the high and low quality factors ensures the decomposition effect, ensuring that the high resonance component contains high-frequency fault feature information, or using the high resonance component to suppress the low resonance component, while the low resonance component suppresses the non-linear noise interference and contains low-frequency fault feature information. Two wavelet bases with different quality factors are generated through morphological component analysis and tunable Q-factor wavelet transform, and the non-linear complex signal containing high and low quality factors in the original continuous oscillation signal is decomposed into high resonance components and low resonance components, so that the high and low resonance components contain rich fault information. The tunable Q-factor wavelet transform is a flexible fully discrete wavelet transform. By adjusting the values of the quality factor Q and the redundancy r, signal decomposition can be achieved at different scales, and the control of the frequency response frequency domain overlap at adjacent scales can be obtained, so as to achieve the precise separation of different frequency components. Morphological component analysis performs sparse representation on the decomposed signal, minimizes the coupling degree of the high resonance component and the low resonance component, effectively separates the components with different resonance attributes in the signal, and extracts fault feature information. Taking the autocorrelation function of the high resonance component decomposed by resonance and the envelope spectrum kurtosis of the low resonance component as the objective function, the grey wolf optimization algorithm is used to adaptively search for the optimal objective function, so as to obtain the optimal quality factor Q and the optimal decomposition layer number J corresponding to the high resonance component and the low resonance component corresponding to the optimal objective function, so as to decompose the optimal high resonance component and the optimal low resonance component.

[0014] Preferably, the specific process of step 2 is as follows:

[0015] Step 21: According to the high decomposition layer number J H and the low decomposition layer number J L perform two-channel filter bank sparse decomposition on the acoustic signal respectively to obtain a high-quality factor decomposition matrix W 1 and a low-quality factor decomposition matrix W 2 ; the decomposition layer number is equal to the number of two-channel filter banks. The two-channel filter bank includes a high-pass filter and a low-pass filter. Several two-channel filter banks are connected in sequence. Each filter outputs a quality factor decomposition matrix. The low-pass filter obtains the output of the previous low-pass filter as the input of the next two-channel filter bank. The acoustic signal passes through several two-channel filter banks to obtain multiple groups of high-quality factor decomposition matrices W 1 and low-quality factor decomposition matrices W 2 ;

[0016] Step 22: According to the high-quality factor Q H and the low-quality factor Q LPerform tunable Q-factor wavelet transform on the acoustic signals respectively. Convert each high-frequency component and low-frequency component of the acoustic signals to the frequency domain through discrete Fourier transform, and obtain multiple groups of high-quality factor sparse basis functions S H and low-quality factor sparse basis functions S L ;

[0017] Step 23: Perform morphological component analysis based on multiple groups of high-quality factor decomposition matrices W 1 、low-quality factor decomposition matrices W 2 、high-quality factor sparse basis functions S H and low-quality factor sparse basis functions S L to construct a dissipation function, and use the split augmented Lagrangian shrinkage algorithm to solve the dissipation function to obtain high-resonance components W 1 S H and low-resonance components W 2 S L ;

[0018] Step 24: Construct an autocorrelation function AFM based on the high-resonance components W 1 S H and calculate the envelope spectrum kurtosis KES according to the low-resonance components W 2 S L ;

[0019] Step 25: Construct an objective function G based on the autocorrelation function AFM and the envelope spectrum kurtosis KES;

[0020] Step 26: Use the grey wolf optimization algorithm to optimize the objective function G to obtain better autocorrelation function AFM and envelope spectrum kurtosis KES;

[0021] Step 27: Calculate better high-resonance components W 1 S H according to the better autocorrelation function AFM, and calculate better low-resonance components W 2 S L ;

[0022] Step 28: Calculate better high-quality factor Q 1 S H and high decomposition fraction J H H according to the better high-resonance components W; Calculate better low-quality factor Q 2 S L L and low decomposition fraction J L H ;

[0023] Step 29: Return to Step 21 until the optimal high-quality factor Q H, high decomposition level J H , low quality factor Q L , low decomposition level J L , or the grey wolf optimization algorithm reaches the maximum iteration condition during optimization;

[0024] Step 210: Resonantly decompose the acoustic signal according to the optimal quality factor, low decomposition level, and redundancy to obtain the optimal high resonance component and the optimal low resonance component.

[0025] Preferably, the expression for calculating the quality factor according to the resonance component is:

[0026]

[0027] where f o is the center frequency of the resonance component, and BW is the bandwidth of the resonance component;

[0028] The expression for calculating the decomposition level according to the quality factor is:

[0029]

[0030] where N is the length of the acoustic signal, Q represents the quality factor, and r represents the redundancy; is the rounding symbol, rounding down; the calculation result of J ensures that the decomposition level meets the signal complexity requirements.

[0031] Preferably, the redundancy r≥3.5 to balance the time-domain localization performance and the frequency-domain decomposition accuracy. In the adjustable Q-factor wavelet transform, r = 4.5. In the same frequency range, the larger the redundancy value r, the more serious the frequency-domain overlap of the wavelet frequency response on adjacent scales. When r approaches 1, the time-domain localization performance of the wavelet is poor, while when the r value is large, the frequency-domain overlap degree of the wavelet on adjacent scales increases, thus achieving accurate separation of high and low quality factors in the time-frequency domain. The number of levels of wavelet change during the decomposition process is represented by J. This change consists of a series of two-channel filter banks. The multi-level decomposition process is to iteratively decompose the low-pass output of the low-scale filter bank through high- and low-pass filters, and finally obtain the multi-scale decomposition wavelet coefficients through discrete Fourier transform.

[0032] Preferably, morphological component analysis performs sparse representation on the decomposed signal, minimizes the coupling degree between the high resonance component and the low resonance component, effectively separates the components with different resonance attributes in the signal, extracts fault feature information, and the smaller the coupling degree between the two separated parts, the better; use the dissipation function to perform sparse decomposition on the signal, and decompose the obtained high-quality factor decomposition matrix W 1 , low-quality factor decomposition matrix W 2 , high-quality factor sparse basis function S H and low-quality factor sparse basis function S LFurther optimize the construction of the dissipation function, expressed as:

[0033]

[0034] Among them, W 1 and W 2 are the high-quality factor decomposition matrix and the low-quality factor decomposition matrix respectively; S H and S L are the high-quality factor sparse basis function and the low-quality factor sparse basis function respectively; λ 1 and λ 2 are the regularization parameters respectively, ensuring the minimum coupling degree between the high and low resonance components after sparse decomposition; y represents the acoustic signal.

[0035] By using the split augmented Lagrangian shrinkage algorithm to solve W 1 and W 2 in the dissipation function, the minimum coefficient sparse matrix is obtained. At this time, a comprehensive minimization state is achieved between the coupling degree of the high resonance component and the low resonance component and the residual component in the decomposition result. Obtain the low resonance component (W 2 S L ) containing the seeder fault information from the original acoustic signal y, decompose the continuous impact pulse noise background into the high resonance component (W 1 S H ), or extract other high-frequency fault characteristics generated by other seeder assembly mechanical fit errors from the high resonance component (W 1 S H ).

[0036] Preferably, the process of constructing the autocorrelation function AFM in step 24 is as follows:

[0037] Step 2411: Extract the autocorrelation maximum value p 1 S H from the high resonance component W u ;

[0038] Step 2412: Construct the autocorrelation function AFM according to the autocorrelation maximum value p u , expressed as:

[0039]

[0040] In the formula, U is the number of delays, and the impact characteristics of the fault signal are enhanced by maximizing the value of AFM. The autocorrelation function is the root mean square of the autocorrelation maximum value. This function is used as a metric for fault information to reflect the useful weak periodic fault influence components in the signal. The autocorrelation function of noise decays rapidly to 0, and its AFM value will be very small. If the high resonance component contains useful periodic fault influence components, then its autocorrelation function is periodic, and its AFM value will increase significantly.

[0041] Preferably, the process of calculating the envelope spectrum kurtosis KES in step 24 is as follows:

[0042] Step 2421: Extract the low resonance component W 2 S L of the envelope spectrum;

[0043] Step 2422: Calculate the envelope spectrum kurtosis KES according to the envelope spectrum. The expression is:

[0044]

[0045] where A(f t ) represents the amplitude at frequency f t in the envelope spectrum, is the average value of the amplitudes of the envelope spectrum, N A is the signal length of the low resonance component W 2 S L . The decomposition effect of the non - linear noise of the low resonance component is measured by the envelope spectrum kurtosis. KES has second - order cyclostationarity and can suppress the non - linear continuous pulse noise or other fault diagnosis information generated when the hill - drop seeder is working normally.

[0046] Preferably, the grey wolf optimization algorithm is used to adaptively search for the quality factor Q and the decomposition layer number J in the resonance decomposition method, forming an objective function of the maximum value of the autocorrelation function of the high resonance component and the envelope spectrum kurtosis of the low resonance component. The objective function G of the grey wolf optimization algorithm is expressed as:

[0047] G = AFM + KES

[0048] By maximizing the objective function G, the optimal autocorrelation function AFM and envelope spectrum kurtosis KES are adaptively searched, so as to infer the optimal quality factor Q and decomposition layer number J, and obtain the optimal combination of resonance decomposition parameters [Q H , J H , Q L , J L , to enhance the decoupling ability of the fault signal in the complex noise background, ensure the effective extraction of the fault signal characteristics in the non - linear impact noise background, and save the calculation cost.

[0049] For the technical effect of the above - mentioned technical solution, under the condition that the redundancy r = 4.5, the grey wolf optimization algorithm is used to search the solution space, and the sparse resonance decomposition combination [Q H , J H , Q L , J L when the index G is the largest is solved, saving the calculation cost and improving the decoupling ability of the effective fault signal in the non - linear continuous noise background.

[0050] An acoustic non-contact precision acupuncture seeding device key component quality detection device comprises a frequency converter, a three-phase AC asynchronous motor, an aluminum profile bracket, an acoustic sensor, a data acquisition card and a PC computer; the frequency converter is respectively connected to the three-phase AC asynchronous motor and the PC computer, and the three-phase AC asynchronous motor is fixed on the aluminum profile bracket; the precision acupuncture seeding device to be detected is fixed on the aluminum profile bracket and is drivingly connected to the three-phase AC asynchronous motor; the acoustic sensor is connected to the data acquisition card, and the data acquisition card is connected to the PC computer; the acoustic sensor collects acoustic signals of the precision acupuncture seeding device and transmits them to the data acquisition card, the data acquisition card transmits the acoustic signals to the PC computer, the PC computer performs signal decomposition and fault diagnosis according to the acoustic signals, and outputs quality detection results.

[0051] Preferably, a PC computer controls a frequency converter to drive a three-phase AC asynchronous motor, driving the precision hole-seeker to be tested to run at a constant speed; an acoustic sensor collects acoustic signals when the precision hole-seeker to be tested is in operation, and transmits them to the PC computer through a data acquisition card; the PC computer resonates and optimizes the acoustic signals, generates time-frequency spectra of high and low resonance components, and extracts fault characteristic frequencies through envelope spectrum analysis for fault diagnosis, thereby generating quality inspection results.

[0052] Preferably, the precision seed drill to be tested is fixed on the aluminum profile bracket through a spherical bearing and an end fixed bearing, and the three-phase AC asynchronous motor is connected to the spherical bearing through a coupling.

[0053] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a method and device for quality inspection of key components of an acoustic non-contact precision seeding device. In order to overcome the shortcomings of traditional detection methods and improve the efficiency and accuracy of factory quality inspection of seeding devices, a diagnostic method based on acoustic signals is proposed. For the acoustic signals generated during the operation of the seeding device, the original acoustic signals are decomposed by resonance decomposition technology, and the acoustic signals are decomposed and reconstructed by adjustable Q factor wavelet transform, and the weak coupling fault characteristic frequencies under the complex noise background are separated. Then, morphological component analysis is used to separate fault information from the complex noise background, thereby achieving rapid and accurate fault diagnosis. The present invention uses the acoustic signal of the normal operation of the precision seeding device as a benchmark. During normal operation, the seeding device to be tested has 15 holes, and the rotation speed is set to 60 rpm. The operating rotation frequency is 1Hz. The 15 holes contain 15 keel pieces, and each rotation generates 15 pulse shocks, so the frequency when normal and without fault is 15Hz. Analyze the characteristic differences of the test sample signals in the time domain and frequency domain to achieve rapid and accurate diagnosis of the assembly quality of the mechanical seed drill, so as to improve the efficiency and accuracy of factory quality inspection of mechanical seed drills, and quickly and efficiently detect the factory quality of mechanical seed drills in a non-contact manner.

[0054] The present invention particularly aims at the following two scenarios: one is that there is a large amount of non-linear noise interference during the normal operation of the equipment; the other is that it is impossible to install an intrusive acceleration sensor on complex rotating machinery. By using non-contact detection technology, it diagnoses the weak fractures of key components of rotating machinery and extremely weak coupling fault characteristics, and realizes quality detection. Description of the Drawings

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0056] Figure 1 Schematic diagram of the process for the quality detection method of key components of the acoustic non-contact precision seed drill provided by the present invention;

[0057] Figure 2 Schematic diagram of the structure of the dual-channel decomposition and synthesis filter for the adjustable Q-factor wavelet transform provided by the present invention;

[0058] Figure 3 Schematic diagram of the method for resonance decomposition of the original data provided by the present invention;

[0059] Figure 4 Schematic diagram of the comparison between the original time-domain data collected under the background of complex continuous impact noise and the diagnostic frequency for extracting the weak coupling fault characteristic frequency provided by the present invention;

[0060] Figure 5 Schematic diagram of the structure of the quality detection device for key components of the acoustic non-contact precision seed drill provided by the present invention. Detailed Embodiments

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0062] The embodiment of the present invention discloses a method for detecting the quality of key components of an acoustic non-contact precision seed drill. By using the resonance decomposition method to decompose the acoustic signals generated during the operation of the seed drill, combined with the adjustable Q-factor wavelet transform and morphological component analysis, the weak fault characteristics are effectively separated from the complex noise background, and the factory quality of the mechanical seed drill can be detected non-contact, quickly and efficiently. As Figure 1 shown, it includes the following steps:

[0063] S1: Collect the acoustic signals generated during the operation of the key components of the precise hole seeder;

[0064] S2: Decompose the collected acoustic signals using the resonance decomposition method, and optimize the resonance decomposition parameters of the resonance decomposition method using the grey wolf optimization algorithm to obtain the optimal high-resonance component and the optimal low-resonance component;

[0065] S3: Perform envelope spectrum processing based on the periodic noise pulse components in the optimal high-resonance component and the periodic fault components of non-linear noise suppression in the optimal low-resonance component, and perform fast Fourier transform on the extracted envelope spectrum to obtain its frequency spectrum, and conduct fault diagnosis based on the frequency spectrum to finally obtain the quality inspection result.

[0066] Furthermore, the resonance decomposition parameters include the quality factor Q, redundancy r, and decomposition layer J. Set the initial high-quality factor Q H , high decomposition layer J H , low-quality factor Q L , low decomposition layer J L and redundancy r; The resonance decomposition method performs two-channel filter bank sparse decomposition, tunable Q-factor wavelet transform, and morphological component analysis on the acoustic signals according to the resonance decomposition parameters, controls the frequency domain and time domain characteristics of the wavelet, and separates the high-resonance component and the low-resonance component.

[0067] Furthermore, the redundancy r ≥ 3.5 to balance the time domain localization performance and the frequency domain decomposition accuracy, and r = 4.5 in the tunable Q-factor wavelet transform.

[0068] Furthermore, the specific process of S2 is as Figure 3 shown, specifically:

[0069] S21: Perform two-channel filter bank sparse decomposition on the acoustic signals respectively according to the high decomposition layer J H , low decomposition layer J L to obtain the high-quality factor decomposition matrix W 1 and the low-quality factor decomposition matrix W 2 ; The decomposition layer is equal to the number of two-channel filter banks. The two-channel filter bank includes a high-pass filter and a low-pass filter. A number of two-channel filter banks are connected in sequence. Each filter outputs a quality factor decomposition matrix. The low-pass filter obtains the output of the previous low-pass filter as the input of the next two-channel filter bank. The acoustic signals pass through a number of two-channel filter banks to obtain multiple groups of high-quality factor decomposition matrices W 1 and low-quality factor decomposition matrices W 2 ;

[0070] S22: According to the high-quality factor Q Hand low quality factor Q L Perform an adjustable Q-factor wavelet transform on the acoustic signal respectively, convert each high-frequency component and low-frequency component of the acoustic signal to the frequency domain through discrete Fourier transform, and obtain multiple groups of high-quality factor sparse basis functions S H and low-quality factor sparse basis functions S L ;

[0071] S23: According to multiple groups of high-quality factor decomposition matrices W 1 , low-quality factor decomposition matrices W 2 , high-quality factor sparse basis functions S H and low-quality factor sparse basis functions S L perform morphological component analysis, construct a dissipation function, and use the split augmented Lagrangian shrinkage algorithm to solve the dissipation function to obtain high-resonance components W 1 S H and low-resonance components W 2 S L ;

[0072] S24: Construct an autocorrelation function AFM according to the high-resonance components W 1 S H and calculate the envelope spectrum kurtosis KES according to the low-resonance components W 2 S L ;

[0073] S25: Construct an objective function G according to the autocorrelation function AFM and the envelope spectrum kurtosis KES;

[0074] S26: Use the grey wolf optimization algorithm to optimize the objective function G to obtain better autocorrelation function AFM and envelope spectrum kurtosis KES;

[0075] S27: Calculate better high-resonance components W 1 S H according to the better autocorrelation function AFM, and calculate better low-resonance components W 2 S L ;

[0076] S28: Calculate better high-quality factor Q 1 S H and high decomposition fraction J H ; Calculate better low-quality factor Q H and low decomposition fraction J 2 S L according to the better low-resonance components W L and low decomposition fraction J L ;

[0077] S29: Return to S21 until the optimal high-quality factor Q is obtainedH , high decomposition level J H , low quality factor Q L , low decomposition level J L , or the gray wolf optimization algorithm reaches the maximum iteration condition during optimization;

[0078] S210: Resonantly decompose the acoustic signal according to the optimal quality factor, low decomposition level, and redundancy to obtain the optimal high resonance component and the optimal low resonance component.

[0079] Furthermore, the signal decomposition and reconstruction process of the tunable Q-factor wavelet transform is as Figure 2 shown. The quality factor Q is defined as the ratio of the center frequency f of the signal 0 to its bandwidth BW, reflecting the concentration degree of the signal frequency components. The expression for calculating the quality factor based on the resonance component is;

[0080]

[0081] The quality factor Q determines the scaling factor β of the high-pass scale transform, and further affects the scaling factor α of the low-pass scale transform. Both are used to control the scale transform of the filter;

[0082]

[0083] In the formula, 0 < β ≤ 1, 0 < α ≤ 1, α + β > 1

[0084] α and β determine the scaling transform relationship between the low-pass filter and the high-pass filter in the multi-layer filter bank used in the tunable Q-factor wavelet transform, affecting the frequency resolution and time-domain resolution of the filter; Determine the low-pass filter H according to α and β 0 (ω) and the high-pass filter H 1 (ω), which are used in the decomposition and reconstruction process of the tunable Q-factor wavelet transform;

[0085]

[0086] In the formula, θ(ω) is an orthogonal filter with two vanishing moments

[0087]

[0088] The expression for calculating the decomposition level according to the quality factor is:

[0089]

[0090] Among them, N is the length of the acoustic signal, Q represents the quality factor, and r represents the redundancy; is the rounding symbol, rounding down; The calculation result of J ensures that the decomposition level meets the signal complexity requirements.

[0091] Furthermore, morphological component analysis performs sparse representation on the decomposed signal, minimizes the coupling degree between the high-resonance component and the low-resonance component, effectively separates the components with different resonance attributes in the signal, extracts fault feature information, and the smaller the coupling degree of the two separated parts, the better; a dissipation function is used to perform sparse decomposition on the signal, and the obtained high-quality factor decomposition matrix W 1 , low-quality factor decomposition matrix W 2 , high-quality factor sparse basis function S H and low-quality factor sparse basis function S L are further optimized to construct a dissipation function, expressed as:

[0092]

[0093] where W 1 and W 2 are the high-quality factor decomposition matrix and the low-quality factor decomposition matrix respectively; S H and S L are the high-quality factor sparse basis function and the low-quality factor sparse basis function respectively; λ 1 and λ 2 are regularization parameters respectively, ensuring that the coupling degree between the high- and low-resonance components after sparse decomposition is minimized; y represents the acoustic signal.

[0094] Furthermore, the autocorrelation maximum value p 1 S H is extracted from the high-resonance component W u , and an autocorrelation function AFM is constructed according to the autocorrelation maximum value p u ; the envelope spectrum of the low-resonance component W 2 S L is extracted, and the envelope spectrum kurtosis KES is calculated according to the envelope spectrum; the expression is:

[0095]

[0096] In the formula, p is the autocorrelation maximum value, U is the number of delay times, and the impact characteristics of the fault signal are enhanced by maximizing the value of AFM; f t represents the amplitude at the frequency f t in the envelope spectrum, is the mean value of the envelope spectrum, N A is the signal length of the low-resonance component W 2 S L , and the suppression effect of the non-linear noise in the low-resonance component is reflected by the value of KES.

[0097] Furthermore, the objective function G is expressed as:

[0098] G = AFM + KES

[0099] By maximizing the objective function G, the optimal autocorrelation function AFM and envelope spectrum kurtosis KES are adaptively searched, thereby inferring the optimal quality factor Q and decomposition layer number J, and obtaining the combination of optimized resonance decomposition parameters [Q H , J H , Q L , J L ].

[0100] Furthermore, the acoustic sensor placed horizontally 20 cm in front of the mechanical seeding device obtains the time domain waveforms of the normal fault-free signal and the single-arm fracture signal, and uses the optimized resonance algorithm to perform sparse resonance decomposition to decompose the fault impact pulse into high-resonance components and the noise and other residual components into low-resonance components. Figure 4 As shown, the resonance sparse decomposition results are as follows Figure 4 As shown in (a) and (b), the signal of normal function and the signal of single arm fracture reflect the basic impact characteristics; Figure 4 (c) and (d) describe the envelope spectra under normal operating conditions and when a single arm is broken, respectively. In the envelope spectrum, the characteristic frequency and harmonic components of normal operation, as well as the characteristic frequency and harmonic components of a single arm break, are clearly present. The typical frequency and harmonic components of normal operation, as well as the characteristic frequency and harmonic components of a single arm break, are clearly visible in the envelope spectrum. The method proposed in the present invention can accurately extract weak coupling faults under the background of continuous impact noise.

[0101] On the other hand, in a specific embodiment, an acoustic non-contact precision seeding device key component quality detection device, such as Figure 5 As shown, it includes a frequency converter 1, a three-phase AC asynchronous motor 2, an aluminum profile bracket 13, a first acoustic sensor 7, a second acoustic sensor 8, a SCADAS data acquisition card 14 and a PC computer 9; the frequency converter 1 is connected to the three-phase AC asynchronous motor 2 and the PC computer 9 respectively, and the three-phase AC asynchronous motor 2 is fixed on the aluminum profile bracket 13; the precision cauterization device 5 to be detected is fixed on the aluminum profile bracket 13 and is connected to the three-phase AC asynchronous motor 2 by transmission; the first acoustic sensor 7 and the second acoustic sensor 8 are connected to the SCADAS data acquisition card 14 respectively, and the SCADAS data acquisition card 14 is connected to the PC computer 9; the first acoustic sensor 7 and the second acoustic sensor 8 respectively collect the acoustic signals of the precision cauterization device 5 and transmit them to the SCADAS data acquisition card 14. Since the acoustic wave data collected by the acoustic sensor is transmitted by a gas medium, and the gas medium has a greater attenuation of the vibration wave than the solid medium, two groups of acoustic sensors are used to perform data amplitude correction of the time domain information.

[0102] Further, the precise hill-drop planter 5 is fixed on the aluminum profile bracket 13 through a spherical bearing 4 and an end-face fixed bearing 6, and the three-phase AC asynchronous motor 2 is connected to the spherical bearing 4 through a coupling 3.

[0103] Further, the acoustic sensor is an acoustic sensor with IEPE output.

[0104] The process of detection using the above detection device is as follows:

[0105] The PC computer 9 is connected to the USB interface 11 in the frequency converter 1 through the RS485 bus, and controls the frequency converter 1 to adjust the speed of the three-phase AC asynchronous motor 2 through communication. Among them, the power interface 10 of the frequency converter 1 is connected to 380V alternating current, and the output end 12 of the frequency converter is connected to the three-phase AC asynchronous motor 2; the frequency converter 1 controls the rotation frequency of the mechanical hill-drop planter 5 to be stable at 60 revolutions per minute. Among them, the precise hill-drop planter 5 and the three-phase AC asynchronous motor are fixed on the horizontal ground through the aluminum profile bracket 13 and rotate coaxially through the coupling 3;

[0106] The acoustic signals generated when the precise hill-drop planter 5 rotates are collected by the first acoustic sensor 7 and the second acoustic sensor 8, and the collected acoustic signals are uploaded to the PC computer 9 through the Ethernet interface by using the SCADAS data acquisition card 14 for signal analysis and fault diagnosis.

[0107] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0108] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An acoustic non-contact precision seeding device key component quality inspection method, characterized in that: The following steps are involved: Step 1: Collect the acoustic signals generated during the operation of the key components of the precision seed drill; Step 2: Decompose the collected acoustic signal using the resonance decomposition method, and optimize the resonance decomposition parameters of the resonance decomposition method using the Grey Wolf optimization algorithm to obtain the optimal high resonance component and the optimal low resonance component: Step 3: Extract the envelope spectrum based on the optimal high resonance component and the optimal low resonance component, perform Fourier transform on the extracted envelope spectrum to obtain the frequency spectrum, perform fault diagnosis based on the frequency spectrum, and obtain quality inspection results.

2. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 1 is characterized in that: The resonance decomposition parameters include the quality factor Q, redundancy r and the number of decomposition levels J. The initial high quality factor Q is set H , high decomposition layer J H , low quality factor Q L , low decomposition level J L and redundancy r; the resonance decomposition method performs dual-channel filter bank sparse decomposition, adjustable Q factor wavelet transform and morphological component analysis on the acoustic signal according to the resonance decomposition parameters, controls the frequency domain and time domain characteristics of the wavelet, and separates the high resonance component and the low resonance component.

3. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 2 is characterized in that: The specific process of step 2 is: Step 21: According to the high decomposition layer J H , low decomposition level J L Perform dual-channel filter bank sparse decomposition on the acoustic signal to obtain a high-quality factor decomposition matrix W1 and a low-quality factor decomposition matrix W2; The number of decomposition layers is equal to the number of dual-channel filter groups. The acoustic signal passes through a number of dual-channel filter groups to obtain multiple groups of high-quality factor decomposition matrices W1 and low-quality factor decomposition matrices W2. Step 22: According to the high quality factor Q H and low quality factor Q L The acoustic signals are respectively subjected to adjustable Q factor wavelet transform to obtain multiple groups of high quality factor sparse basis functions S H and low quality factor sparse basis function S L ; Step 23: Based on multiple sets of high-quality factor decomposition matrices W1, low-quality factor decomposition matrices W2, and high-quality factor sparse basis functions S H and low quality factor sparse basis function S L Perform morphological component analysis, construct the dissipation function, and use the split augmented Lagrangian shrinkage algorithm to solve the dissipation function to obtain the high resonance component W1S H and low resonance component W2S L ; Step 24: Based on the High Resonance Component W1S H Construct the autocorrelation function AFM, based on the low resonance component W2S L Calculate the envelope spectrum kurtosis KES; Step 25: construct the target function G according to the autocorrelation function AFM and the envelope spectrum kurtosis KES; Step 26: Use the Grey Wolf optimization algorithm to optimize the objective function G to obtain a better autocorrelation function AFM and envelope spectrum kurtosis KES; Step 27: Calculate a better high resonance component W1S based on a better autocorrelation function AFM H , calculate the better low resonance component W2S according to the better envelope spectrum kurtosis KES L ; Step 28: Based on the more optimal high resonance component W1S H Calculate a better quality factor Q H and high decomposition score J H ; Based on the better low resonance component W2S L Calculate a better low quality factor Q L and low decomposition score J L ; Step 29: Return to step 21 until the optimal high quality factor Q is obtained H , high decomposition layer J H , low quality factor Q L , low decomposition level J L , or the Grey Wolf Optimization Algorithm reaches the maximum iteration condition; Step 210: Perform resonance decomposition on the acoustic signal according to the optimal quality factor, low decomposition layer number and redundancy to obtain an optimal high resonance component and an optimal low resonance component.

4. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 3 is characterized in that: The expression for calculating the quality factor based on the resonance component is: Among them, f o is the center frequency of the resonance component, BW is the bandwidth of the resonance component; The expression for calculating the number of decomposition layers based on the quality factor is: Where N is the length of the acoustic signal, Q represents the quality factor, and r represents the redundancy; Is the rounding sign, round down.

5. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 2 is characterized in that: The redundancy r≥3.5, and r=4.5 in the adjustable Q factor wavelet transform.

6. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 3 is characterized in that: The dissipation function is expressed as: Among them, λ1 and λ2 are regularization parameters respectively; y represents the acoustic signal.

7. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 3 is characterized in that: The process of constructing the autocorrelation function AFM in step 24 is: Step 2411: From the high resonance component W1S H Extract the maximum autocorrelation value p from u ; Step 2412: According to the maximum autocorrelation value p u Construct the autocorrelation function AFM, expressed as: Where U is the delay times, and the impact characteristics of the fault signal are enhanced by maximizing the value of AFM.

8. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 3 is characterized in that: The process of calculating the envelope spectrum kurtosis KES in step 24 is: Step 2421: Extract low resonance component W2S L The envelope spectrum of Step 2422: Calculate the envelope spectrum kurtosis KES according to the envelope spectrum, and the expression is: Among them, A(f t ) represents the frequency f in the envelope spectrum t The amplitude at is the mean amplitude of the envelope spectrum, N A The low resonance component W2S L signal length.

9. The method for quality inspection of key components of an acoustic non-contact precision seeding device according to claim 3 is characterized in that: The objective function G of the gray wolf optimization algorithm is expressed as: G=AFM+KES By maximizing the objective function G, the optimal autocorrelation function AFM and envelope spectrum kurtosis KES are adaptively searched to obtain the optimal combination of resonance decomposition parameters [Q H , J H , Q L , J L ].

10. An acoustic non-contact precision seeding device for key component quality inspection, characterized in that: The method for quality inspection of key components of an acoustic non-contact precision acupuncture point-seeker according to any one of claims 1 to 9 comprises a frequency converter, a three-phase AC asynchronous motor, an aluminum profile bracket, an acoustic sensor, a data acquisition card and a PC; the frequency converter is respectively connected to the three-phase AC asynchronous motor and the PC, and the three-phase AC asynchronous motor is fixed on the aluminum profile bracket; the precision acupuncture point-seeker to be inspected is fixed on the aluminum profile bracket and is transmission-connected to the three-phase AC asynchronous motor; the acoustic sensor is connected to the data acquisition card, and the data acquisition card is connected to the PC; the acoustic sensor collects acoustic signals of the precision acupuncture point-seeker and transmits them to the data acquisition card, the data acquisition card transmits the acoustic signals to the PC, the PC performs signal decomposition and fault diagnosis according to the acoustic signals, and outputs quality inspection results.

Images