A method and device for monitoring the health status of mechanical equipment resistant to strong noise interference

Abstract

The present invention provides a method and device for monitoring the health status of mechanical equipment resistant to strong noise interference, the method comprising: step 1: collecting vibration signals of mechanical equipment at different positions and selecting input signals; step 2: decomposing the vibration signal based on the CEEMDAN algorithm, and distinguishing the noise-dominated connotation modal component IMF and the signal-dominated connotation modal component IMF according to the KL divergence; step 3: based on the wavelet threshold denoising method, performing denoising on the noise-dominated connotation modal component IMF; step 4: reconstructing the denoised noise-dominated connotation modal component IMF and the signal-dominated connotation modal component IMF into a denoised vibration signal; step 5: judging the health status of the mechanical equipment. The present invention can perform denoising on the vibration signal while ensuring the integrity of the fault characteristics, so as to extract the fault characteristics of the mechanical equipment faster and more accurately and judge the health status of the mechanical equipment.

Description

Technical Field

[0001] The present invention relates to the field of mechanical equipment fault diagnosis, and in particular to a method and device for monitoring the health status of mechanical equipment resistant to strong noise interference. Background Art

[0002] With the rapid development of my country's manufacturing industry, various mechanical equipment are constantly being innovated. my country's aerospace vehicles, large generator sets, gas turbine units, ships, deep-diving equipment and other major mechanical equipment are showing a trend of large-scale, complex and integrated components, which is bound to put forward more stringent requirements on their operating indicators. The health status of key components in mechanical equipment, such as reducers in helicopters and gas turbines in gas turbine units, is directly related to the working status of the entire equipment, so it is very important to monitor the health of key parts in mechanical equipment. When mechanical equipment starts working, a large number of vibration signals will be generated, accompanied by strong noise. The vibration signal of mechanical equipment contains rich health information about mechanical equipment. When a mechanical equipment fails, the amplitude of its vibration signal will increase significantly, and its signal characteristics will also change accordingly. Therefore, it is very important to extract the characteristics of the vibration signal and then realize the health monitoring of mechanical equipment. At present, the vibration signal feature extraction based on vibration signals is mainly divided into two methods: time domain and spectrum. The time domain vibration signal feature extraction method includes EMD (empirical mode decomposition) and other methods, and the spectrum vibration signal feature extraction method mainly includes inverse spectrum and other methods. However, when a mechanical device fails, its vibration signal has the characteristics of multi-signal coupling, weak signal and strong noise interference. The traditional vibration signal feature extraction method cannot effectively extract the fault features. The EMD method is prone to problems such as modal aliasing and endpoint effect, and cannot decompose the vibration signal well to extract the time domain fault features. Due to the characteristics of noise and multi-signal coupling, extracting frequency domain fault features through methods such as inverse spectrum is not ideal, and it is difficult to perform health monitoring through the extracted features. Therefore, a new health monitoring method is needed that can eliminate the influence of noise while enhancing the fault characteristics, so as to judge the health status of mechanical equipment earlier and more accurately to avoid greater losses. Summary of the invention

[0003] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method for monitoring the health status of mechanical equipment that is resistant to strong noise interference in view of the deficiencies in the prior art. The method specifically comprises the following steps:

[0004] The following steps are involved:

[0005] Step 1, by using more than two acceleration sensors placed at different positions of the mechanical equipment, obtain vibration signals of the mechanical equipment when it is running at different positions, analyze the collected vibration signals, calculate the effective value (RMF component), select the one with the largest effective value as the vibration sensitive position, and use the vibration signal of the sensitive position as the input signal;

[0006] Step 2: Process the input signal to obtain the intrinsic mode components IMF (Intrinsic Mode Functions, IMF) dominated by noise and the intrinsic mode components IMF dominated by signal;

[0007] Step 3, using the wavelet threshold denoising method to denoise each intrinsic modal component IMF;

[0008] Step 4, combining the noise-dominated intrinsic modal component IMF after denoising with the signal-dominated intrinsic modal component IMF to form a denoised vibration signal;

[0009] Step 5: Use the enhanced envelope spectrum method to extract fault features from the denoised vibration signal to determine the health status of the mechanical equipment.

[0010] Step 2 includes:

[0011] Step 2-1, select a set of vibration signals x(t), and transform the standard normally distributed Gaussian white noise ω i (t) is added to the vibration signal x(t) to obtain the signal to be processed x(t)+ζω i (t), where ζ is the standard deviation of the noise;

[0012] Step 2-2, using the empirical mode decomposition algorithm EMD, after repeated decomposition of the signal to be processed I times, arithmetic averaging is performed to obtain the IMF1 component IMF1(t):

[0013]

[0014] Among them, IMF1(t) is the first intrinsic modal component IMF decomposed from the signal, IMF1 i (t) is the first intrinsic modal component IMF obtained after the signal is decomposed repeatedly I times, and there are I in total;

[0015] The residual component r1(t) after removing the first-order IMF1(t) is:

[0016] r1(t)=x(t)-IMF1(t);

[0017] Step 2-3, define the operator of the jth mode function component obtained from the empirical mode decomposition algorithm EMD of the signal to be processed as E j(·), 1≤j≤K, and the Gaussian white noise ω of the standard normal distribution is i (t) is added to the component r1(t), and the signal r1(t)+ζE1[ω i (t)] is decomposed again by the empirical mode decomposition algorithm EMD, and the IMF2 component is obtained as follows:

[0018]

[0019] The residual component r2(t) after removing the second-order component IMF2 is obtained:

[0020] r2(t)=r1(t)-IMF2(t)

[0021] By analogy, the original vibration signal is decomposed into K modal function components, and finally the original vibration signal is expressed as:

[0022]

[0023] Among them, IMF k (t) represents the decomposed K modal function components;

[0024] Step 2-5, determine the critical point of the noise-dominated intrinsic modal component IMF and the signal-dominated intrinsic modal component IMF.

[0025] Steps 2-4 include:

[0026] Step 2-4-1, suppose the discrete n of the intrinsic modal component IMF to be calculated i Expressed as: n i =[x1,x2,x3,…,x n ], where the discrete N of the original vibration signal is expressed as: N = [y1, y2, y3, ..., y n ],y n is the amplitude of the vibration signal collected by the sensor for the nth time; the probability distribution p(x) of the intrinsic modal component IMF is calculated using the following formula:

[0027]

[0028] Where k is the Gaussian kernel function, is the mean of the signal, h is a known positive number, and its calculation formula is as follows:

[0029]

[0030] Among them, e represents a natural constant, u is x i and difference;

[0031] Similarly, the probability distribution q(y) of the original vibration signal is obtained;

[0032] Step 2-4-2: Calculate the KL distances δ(p,q) and δ(q,p) between the intrinsic mode function IMF and the original vibration signal:

[0033]

[0034] The calculation formula of δ(q,p) is as follows:

[0035]

[0036] Step 2-4-3: Obtain the KL divergence value KL(p||q) between the original vibration signal and the intrinsic mode function IMF:

[0037] KL(p||q) = δ(p,q) + δ(q,p).

[0038] Step 2-5 includes:

[0039] Step 2-5-1: Subtract the KL divergences of each layer of the intrinsic mode function IMF in turn to obtain the difference curve of the KL divergence;

[0040] Step 2-5-2: Based on the difference curve, select the first difference point greater than 0 as the critical point between the intrinsic mode function IMF dominated by noise and the intrinsic mode function IMF dominated by signal. The intrinsic mode function IMF before the first difference point greater than 0 is used as the intrinsic mode function IMF dominated by noise, and the intrinsic mode function IMF after the first difference point greater than 0 is used as the intrinsic mode function IMF dominated by signal.

[0041] Step 3 includes:

[0042] Step 3-1: Use the Morlet wavelet as the wavelet basis for the wavelet threshold denoising method;

[0043] Step 3-2: Preset the decomposition layer number as X (usually taking the value of 5);

[0044] Step 3-3: Set the wavelet coefficient threshold function:

[0045]

[0046] Where, represents the wavelet coefficient threshold function; ω j,k represents the wavelet coefficient; a is a constant; sgn() is the sign function; λ1 and λ2 are thresholds, and λ1 = aλ2, 0 < a ≤ 1. λ1 is obtained from the following formula:

[0047] λ1 = σ 2 lgM

[0048] Where M represents the length of the signal denoised by wavelet thresholding; σ represents the standard deviation of the k-th layer noise, which is estimated by the following formula:

[0049]

[0050] Among them, median(|ω j,k |) is the median of the absolute values ​​of the wavelet coefficients decomposed at the kth level.

[0051] Step 5 includes:

[0052] Step 5-1, let the denoised vibration signal be x(n), and its autocorrelation function r x (τ) is:

[0053] r x (τ)=E[x(n)x * (n+τ)]

[0054] Where E is the correlation function, τ=1,2,…,n-1 is the time delay, and x * is the conjugate of x, x is the denoised vibration signal x(n);

[0055] Step 5-2, find the autocorrelation envelope spectrum:

[0056] By Hilbert-Huang transform, find the autocorrelation envelope R x (n+τ), and then perform Fourier transform to obtain the autocorrelation envelope spectrum H(k);

[0057] Step 5-3, find the enhanced envelope spectrum:

[0058] The generalized Shannon entropy M(k) is defined as:

[0059] M(k)=H(k)log2 H(k)

[0060] Where H(k) is the autocorrelation envelope spectrum;

[0061] Defining the Enhanced Envelope Spectrum for:

[0062]

[0063] Wherein, k = 1, 2, ..., n; s is the standard deviation of the autocorrelation envelope spectrum H(k).

[0064] The present invention also provides a mechanical health status monitoring device resistant to strong noise interference, comprising a data acquisition module, a signal processing module and a feature extraction module;

[0065] The data acquisition module obtains vibration signals of the mechanical equipment at different positions when the mechanical equipment is running through more than two acceleration sensors installed at different positions of the mechanical equipment, analyzes the collected vibration signals, and then selects n vibration sensitive positions, and uses n groups of vibration signals of the sensitive positions as input signals.

[0066] The signal processing module is used to pre-process the collected n groups of vibration signals to obtain denoised vibration signals;

[0067] The feature extraction module is used to process the noise-reduced data and extract fault features.

[0068] The signal processing module includes a decomposition module, an IMF differentiation module and a denoising module;

[0069] The decomposition module is based on the complete set empirical mode decomposition CEEMDAN algorithm, which uses the empirical mode decomposition algorithm EMD to decompose the original vibration signal with adaptive white noise to obtain the intrinsic mode components IMF and residual r from the 1st layer to the nth layer. h (n);

[0070] The IMF differentiation module obtains the IMF dominated by noise and the IMF dominated by signal based on the KL divergence algorithm;

[0071] The denoising module performs denoising on the noise-dominated intrinsic modal component IMF based on the wavelet basis, decomposition layer number, threshold and threshold function of the preset wavelet threshold algorithm.

[0072] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for monitoring the health status of mechanical equipment that is resistant to strong noise interference is implemented.

[0073] The present invention has the following beneficial effects: firstly, the present invention screens vibration signals at different positions of mechanical equipment based on the effective value of the signal, selects input signals, and improves the accuracy of fault identification; secondly, the present invention removes noise in the signal by constructing a CEEMDAN-wavelet threshold denoising method and a signal reconstruction-wavelet threshold denoising method; finally, the present invention proposes a method based on CEEMDAN-wavelet threshold and enhanced envelope spectrum, which can obtain a clearer frequency spectrum diagram and improve the accuracy of fault feature extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments, and the above and / or other advantages of the present invention will become more clear.

[0075] Figure 1The figure is an overall flow chart of the method of the present invention.

[0076] Figure 2 This is the IMF diagram obtained after the vibration signal is decomposed by CEEMDAN.

[0077] Figure 3 It is the noise-dominated intrinsic modal component IMF obtained after being processed by wavelet threshold denoising method.

[0078] Figure 4 It is the enhanced envelope spectrum obtained by the enhanced envelope spectrum algorithm.

[0079] Figure 5 It is the KL divergence difference curve.

[0080] Figure 6 This is a diagram of the functional modules of the device of the present invention. DETAILED DESCRIPTION

[0081] When using vibration signals to detect mechanical equipment, the time domain and frequency domain characteristic information of the vibration signal will be used to reflect the health status of the mechanical equipment. When a mechanical equipment fails, the frequency domain part of its vibration signal will have a corresponding fault characteristic frequency. However, the vibration signal to be detected is usually mixed with a large number of aliased noise signals, and the fault characteristic frequency is difficult to extract. Therefore, based on the CEEMDAN decomposition algorithm, the present invention introduces white noise to improve the problems of modal aliasing and endpoint effect of the traditional EMD decomposition method, and distinguishes the IMF signal dominated by noise and the IMF signal dominated by signal by calculating the KL divergence of each component after CEEMDAN decomposition and the original vibration signal. Specifically, the first zero crossing point in the differential curve of the KL divergence of the IMF is used to determine the distinction point of the IMF. The connotation modal component IMF before the zero crossing point of the differential curve is regarded as the connotation modal component IMF dominated by noise, and the connotation modal component IMF after the zero crossing point of the differential curve is regarded as the connotation modal component IMF dominated by signal. Based on the preset wavelet threshold denoising function, the connotation modal component IMF dominated by noise is subjected to wavelet threshold denoising respectively. Then, the denoised intrinsic modal components IMF are added to obtain the vibration signal denoised by wavelet threshold. Based on the enhanced envelope spectrum method, the fault characteristics of the vibration signal are enhanced to better extract the fault characteristics of the mechanical equipment.

[0082] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0083] like Figure 1 As shown, the present invention discloses a method and device for monitoring the health status of mechanical equipment resistant to strong noise interference, the specific structure is as follows Figure 6 As shown, the following steps are included:

[0084] Step 1: Vibration signals of the mechanical equipment when it is running at multiple positions are obtained by using multiple acceleration sensors installed at different positions of the mechanical equipment, and the collected vibration signals are analyzed to select vibration sensitive positions, and the vibration signals of the sensitive positions are used as input signals of the vibration signal feature extraction algorithm;

[0085] Step 2: Extract the vibration signal of the sensitive position of the mechanical equipment obtained in step 1. Based on the CEEMDAN (Complete Set Empirical Mode Decomposition) decomposition algorithm, the original vibration signal with adaptive white noise added is modally decomposed using the EMD decomposition algorithm to obtain the intrinsic modal components IMF and residual r(n) from the first layer to the nth layer, as shown in Figure 2 As shown. And by using the KL divergence method, the KL divergence of each intrinsic modal component IMF and the original vibration signal is calculated respectively, and the distinction point of IMF is determined by the first zero-crossing point in the differential curve of the KL divergence of IMF. The intrinsic modal component IMF before the zero-crossing point of the differential curve is regarded as the intrinsic modal component IMF dominated by noise, and the intrinsic modal component IMF after the zero-crossing point of the differential curve is regarded as the intrinsic modal component IMF dominated by signal. The specific implementation method is as follows: Figure 5 shown.

[0086] Step 201: Select a set of vibration signals x(t) and transform the Gaussian white noise ω i (t) is added to the vibration signal x(t) to obtain the signal to be processed x(t)+ζω i (t). Where ζ is the standard deviation of the noise.

[0087] Step 202: Using the EMD method, the signal to be processed obtained in step 201 is repeatedly decomposed once, and then the arithmetic average is performed to obtain the IMF1 component:

[0088]

[0089] The residual component after removing the first-order IMF1(t) is:

[0090] r1(t)=x(t)-IMF1(t)

[0091] Step 203: Define the operator of the jth modal function component obtained from the EMD decomposition of the signal to be processed as E j (·). Again, the Gaussian white noise ω of the standard normal distribution i (t) is added to the component r1(t), then the signal r1(t)+ζE1[ω i (t)] is decomposed again by EMD method, and the IMF2 component is obtained as follows:

[0092]

[0093] Then the residual component after removing the second-order component IMF2 can be obtained:

[0094] r2(t)=r1(t)-IMF2(t)

[0095] By analogy, the original vibration signal can be decomposed into K modal function components, and finally the original vibration signal can be expressed as:

[0096]

[0097] Where R(t) is the final residual component.

[0098] Step 204: Calculate the KL divergence between each IMF component and the original vibration signal.

[0099] The calculation steps of KL divergence include:

[0100] 1. Assume that the discrete representation of the intrinsic modal component IMF to be calculated is: i =[x1,x2,x3,…,x n ], the original vibration signal is: N = [y1, y2, y3, ..., y n ], whose probability distributions are p(x) and q(y), respectively. The probability distributions of the intrinsic modal component IMF and the original vibration signal are calculated respectively. The calculation formula for the IMF probability distribution is as follows:

[0101]

[0102] Among them, k is the Gaussian kernel function, and its calculation formula is as follows:

[0103]

[0104] Similarly, the probability distribution q(y) of the original vibration signal can be obtained:

[0105] 2. Calculate the KL distance δ(p,q) and δ(q,p) between the intrinsic modal component IMF and the original vibration signal:

[0106]

[0107] Similarly, we can obtain δ(q,p).

[0108] 3. Get the KL divergence value KL(p||q) of the original vibration signal and the intrinsic modal component IMF:

[0109] KL(p||q)=δ(p,q)+δ(q,p)

[0110] Step 3: Extract the intrinsic mode function IMF dominated by noise, and use the wavelet threshold denoising method to denoise each intrinsic mode function IMF dominated by noise; preset the wavelet basis, decomposition level, threshold, and threshold function of the wavelet threshold algorithm, and denoise the intrinsic mode function IMF dominated by noise based on the preset wavelet threshold algorithm.

[0111] Step 301: Set the wavelet basis: Use the Morlet wavelet as the wavelet basis for the wavelet threshold denoising method.

[0112] Step 302: On the basis of Step 301, set the decomposition level: Preset the decomposition level to 5 layers;

[0113] Step 303: On the basis of Step 302, set the wavelet coefficient threshold function, and the expression of the threshold function is as follows:

[0114]

[0115] where λ1 and λ2 are thresholds, and λ1 = aλ2, (0 < a ≤ 1), and λ1 can be obtained from the following formula:

[0116] λ1 = σ 2 lgM

[0117] In the formula, M represents the length of the signal denoised by the wavelet threshold, and σ represents the standard deviation of the noise at the k-th layer, which can be estimated by the following formula:

[0118]

[0119] where median(|ω j,k |) is the median of the absolute values of the wavelet coefficients of the k-th layer decomposition.

[0120] Furthermore, the intrinsic mode function IMF dominated by noise will be denoised by wavelet threshold respectively, and then the denoised intrinsic mode function IMFs will be added together to obtain the vibration signal denoised by wavelet threshold.

[0121] Step 4: Combine the vibration signal formed by the intrinsic mode function IMF dominated by noise denoised by wavelet threshold and the intrinsic mode function IMF dominated by signal by addition to form the denoised vibration signal, and the obtained vibration signal is as Figure 3 shown.

[0122] Step 5:

[0123] Calculate the enhanced envelope spectrum of the denoised vibration signal, as Figure 4 shown, and the specific calculation process of the enhanced envelope spectrum is as follows:

[0124] Step 501 calculates the autocorrelation function: assuming that the denoised vibration signal is x(n), its autocorrelation function is:

[0125] r x (τ)=E[x(n)x * (n+τ)]

[0126] Among them, τ=1,2,…,n-1 is the time delay, x * is the conjugate of x.

[0127] Step 502: Calculate the autocorrelation envelope spectrum

[0128] By Hilbert-Huang transform, find the autocorrelation envelope R x (n+τ), and then perform Fourier transform to obtain the autocorrelation envelope spectrum H(k).

[0129] Step 503, find the enhanced envelope spectrum:

[0130] The generalized Shannon entropy M(k) is defined as:

[0131] M(k)=H(k)log2H(k)

[0132] Where H(k) is the autocorrelation envelope spectrum;

[0133] Defining the Enhanced Envelope Spectrum for:

[0134]

[0135] Wherein, k = 1, 2, ..., n; s is the standard deviation of the autocorrelation envelope spectrum H(k).

[0136] According to the calculation results of the enhanced envelope spectrum, the enhanced envelope spectrum is drawn, and the fault information of the mechanical equipment is extracted according to the characteristic frequency in the enhanced envelope spectrum.

[0137] The present invention provides a method and device for monitoring the health status of mechanical equipment that is resistant to strong noise interference. There are many methods and ways to implement the technical solution. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the protection scope of the present invention. All components not specified in this embodiment can be implemented using existing technologies.

Claims

1. A method for monitoring the health status of mechanical equipment resistant to strong noise interference, characterized in that: The following steps are involved: Step 1, using two or more acceleration sensors installed at different positions of the mechanical equipment to obtain vibration signals of the mechanical equipment when it is running at different positions, and analyzing the collected vibration signals to calculate the effective values, selecting the position with the largest effective value as the vibration sensitive position, and using the vibration signal of the sensitive position as the input signal; Step 2, processing the input signal to obtain the noise-dominated intrinsic modal component IMF and the signal-dominated intrinsic modal component IMF; Step 3, using the wavelet threshold denoising method to denoise the IMF dominated by noise; Step 4, combining the noise-dominated intrinsic modal component IMF after denoising with the signal-dominated intrinsic modal component IMF to form a denoised vibration signal; Step 5, using the enhanced envelope spectrum method to extract fault features from the denoised vibration signal, so as to determine the health status of the mechanical equipment; Step 2 includes: Step 2-1, select a set of vibration signals x(t), and transform the standard normally distributed Gaussian white noise ω i (t) is added to the vibration signal x(t) to obtain the signal to be processed x(t)+ζω i (t), where ζ is the standard deviation of the noise; Step 2-2, using the empirical mode decomposition algorithm EMD, after repeated decomposition of the signal to be processed I times, arithmetic averaging is performed to obtain the IMF1 component IMF1(t): Among them, IMF1(t) is the first intrinsic modal component IMF decomposed from the signal, IMF1 i (t) is the first intrinsic modal component IMF obtained after the signal is decomposed repeatedly I times, and there are I in total; The residual component r1(t) after removing the first-order IMF1(t) is: r1(t)=x(t)-IMF1(t); Step 2-3, define the operator of the jth mode function component obtained from the empirical mode decomposition algorithm EMD of the signal to be processed as E j (·), 1≤j≤K, and the Gaussian white noise ω of the standard normal distribution is i (t) is added to the component r1(t), and the signal r1(t)+ζE1[ω i (t)] is decomposed again by the empirical mode decomposition algorithm EMD, and the IMF2 component is obtained as follows: The residual component r2(t) after removing the second-order component IMF2 is obtained: r2(t)=r1(t)-IMF2(t) By analogy, the original vibration signal is decomposed into K modal function components, and finally the original vibration signal is expressed as: Among them, IMF k (t) represents the decomposed K modal function components; Step 2-4: Calculate the KL divergence between each IMF component and the original vibration signal; Step 2-5, determine the critical point of the noise-dominated intrinsic modal component IMF and the signal-dominated intrinsic modal component IMF; Steps 2-4 include: Step 2-4-1, suppose the discrete n of the intrinsic modal component IMF to be calculated i Expressed as: n i =[x1,x2,x3,…,x n ], where the discrete N of the original vibration signal is expressed as: N = [y1, y2, y3, ..., y n ],y n is the amplitude of the vibration signal collected by the sensor for the nth time; the probability distribution p(x) of the intrinsic modal component IMF is calculated using the following formula: Among them, k is the Gaussian kernel function, is the mean of the signal, h is a known positive number, and its calculation formula is as follows: Among them, e represents a natural constant, u is x i and difference; Similarly, the probability distribution q(y) of the original vibration signal is obtained; Step 2-4-2, calculate the KL distance δ(p,q), δ(q,p) between the intrinsic modal component IMF and the original vibration signal: The calculation formula of δ(q,p) is as follows: Step 2-4-3, obtain the KL divergence value KL(p||q) of the original vibration signal and the intrinsic modal component IMF: KL(p||q)=δ(p,q)+δ(q,p); Steps 2-5 include: Step 2-5-1, subtract the KL divergence of each layer of the intrinsic modal component IMF in turn to obtain the difference curve of the KL divergence; Step 2-5-2, based on the differential curve, select the first difference point greater than 0 as the critical point between the intrinsic modal component IMF dominated by noise and the intrinsic modal component IMF dominated by signal, take the intrinsic modal component IMF before the first difference point greater than 0 as the intrinsic modal component IMF dominated by noise, and take the intrinsic modal component IMF after the first difference point greater than 0 as the intrinsic modal component IMF dominated by signal.

2. The method according to claim 1, characterized in that Step 3 includes: Step 3-1, using Morlet wavelet as the wavelet basis of wavelet threshold denoising method; Step 3-2, preset the number of decomposition layers to be X; Step 3-3, set the wavelet coefficient threshold function: Among them, represents the wavelet coefficient threshold function; ω j,k represents the wavelet coefficient; a is a constant; sgn() is the sign function; λ1 and λ2 are thresholds, and λ1 = aλ2, 0 < a ≤ 1. λ1 is obtained from the following formula: λ1=σ 2 lgM Where M represents the length of the signal denoised by wavelet thresholding; σ represents the standard deviation of the k-th layer noise, which is estimated by the following formula: Among them, median(|ω j,k |) is the median of the absolute values ​​of the wavelet coefficients decomposed at the kth level.

3. The method according to claim 2, characterized in that Step 5 includes: Step 5-1, let the denoised vibration signal be x(n), and its autocorrelation function r x (τ) is: r x (τ)=E[x(n)x * (n+τ)] Where E is the correlation function, τ=1,2,…,n-1 is the time delay, and x * is the conjugate of x, x is the denoised vibration signal x(n); Step 5-2, find the autocorrelation envelope spectrum: By Hilbert-Huang transform, find the autocorrelation envelope R x (n+τ), and then perform Fourier transform to obtain the autocorrelation envelope spectrum H(k); Step 5-3, find the enhanced envelope spectrum.

4. The method according to claim 3, characterized in that Step 5-3 includes: The generalized Shannon entropy M(k) is defined as: M(k)=H(k)log2H(k) Where H(k) is the autocorrelation envelope spectrum; Defining the Enhanced Envelope Spectrum for: Wherein, k = 1, 2, ..., n; s is the standard deviation of the autocorrelation envelope spectrum H(k).

5. A mechanical health status monitoring device resistant to strong noise interference implemented based on the method according to any one of claims 1 to 4, characterized in that: It includes a data acquisition module, a signal processing module and a feature extraction module; The data acquisition module obtains vibration signals of the mechanical equipment at different positions when the mechanical equipment is running through more than two acceleration sensors installed at different positions of the mechanical equipment, analyzes the collected vibration signals, calculates the effective value, selects the position with the largest effective value as the vibration sensitive position, and uses the vibration signal of the sensitive position as the input signal; The signal processing module is used to process the input signal to obtain the noise-dominated intrinsic modal component IMF and the signal-dominated intrinsic modal component IMF; denoise the noise-dominated intrinsic modal component IMF using a wavelet threshold denoising method; combine the denoised noise-dominated intrinsic modal component IMF with the signal-dominated intrinsic modal component IMF to form a denoised vibration signal; The feature extraction module uses the method of enhancing envelope spectrum to extract fault features from the denoised vibration signal, thereby judging the health status of the mechanical equipment.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images