Rotating machine fault diagnosis method based on SCNGO-FMD

Abstract

The invention provides a rotating machinery fault diagnosis method based on SCNGO-FMD, and the method comprises the steps: collecting a vibration signal as the input of feature mode decomposition, employing an improved northern eagle optimization algorithm, taking a related composite Gini index CCGI as an optimization target, and carrying out the optimization of the length L of a filter, the number K of filters, and the number n of modes in the feature mode decomposition; performing characteristic mode decomposition on the collected vibration signals by using the optimized parameters, and selecting a mode component with the maximum CCGI as an optimal mode component; then, an envelope spectrum analysis is performed on the optimal mode component to extract a fault frequency. According to the method, the problem of performance reduction caused by improper parameter setting of existing characteristic mode decomposition is effectively solved, so that the fault characteristics are extracted from the vibration signals of the rotating machinery more effectively, and fault diagnosis of the rotating machinery is realized.

Description

Technical Field

[0001] The invention relates to the technical field of rotating machinery fault diagnosis, and in particular to a rotating machinery fault diagnosis method based on SCNGO-FMD. Background Art

[0002] Rotating machinery is widely used in modern industrial production and is one of the key equipment. Monitoring the health status of rotating machinery and performing fault diagnosis are crucial to ensuring industrial production safety and improving equipment reliability.

[0003] The vibration signals of rotating machinery often have nonlinear and non-stationary characteristics. It is difficult to effectively extract the fault features based on the original signal alone. To this end, researchers have proposed a signal processing method called characteristic mode decomposition. FMD is an adaptive signal decomposition technology that uses an adaptive FIR filter group to decompose the original signal, establishes the correlation kurtosis (CK) criterion as the objective function, and iteratively updates the filter group. Compared with the traditional EMD and VMD methods, the FMD algorithm takes into account the impulse and periodic characteristics of the signal, and has good robustness to noise and interference.

[0004] The performance and efficiency of FMD decomposition depend on three key parameters: the number of modes n, the number of filters K, and the filter length L. Reasonable selection of these parameters is crucial to improving the decomposition effect and overall performance of FMD. Too large n may lead to modal redundancy, while too small n may lead to under-decomposition of the signal; increasing K can improve the decomposition performance, but it will also increase the calculation time; L has a significant impact on both the calculation efficiency and the decomposition effect. Summary of the invention

[0005] In view of the deficiencies in the background technology, the present invention proposes a rotating machinery fault diagnosis method based on SCNGO-FMD, which can improve the accuracy and reliability of fault diagnosis.

[0006] The above-mentioned object of the present invention is achieved by the following technical solutions:

[0007] The rotating machinery fault diagnosis method based on SCNGO-FMD includes the following steps:

[0008] S1: Collect vibration signals of rotating mechanical equipment;

[0009] S2: Based on the SCNGO optimization algorithm, the FMD decomposition parameters are optimized with the relevant composite Gini index as the fitness function to obtain the optimal combination of decomposition parameters, where the FMD decomposition parameters include the filter length L, the number of filters K and the number of modes n;

[0010] S3: Perform FMD decomposition on the vibration signal of the rotating equipment according to the optimal parameter combination obtained in S2 to obtain n mode components, calculate the CCGI value of the mode components after FMD decomposition, and select the mode component with the largest value as the optimal component;

[0011] S4: Extract fault characteristic frequency by performing envelope analysis on the best mode component;

[0012] Moreover, in step S2, the step of optimizing the FMD decomposition parameters based on the SCNGO optimization algorithm with the relevant composite Gini index as the fitness function includes:

[0013] S201: Establishing a fitness function for parameter optimization for optimizing the SCNGO algorithm;

[0014] The maximum relevant composite Gini index of the signal is used as the fitness function for parameter optimization, and its expression is:

[0015]

[0016] Where CCGI is the relevant composite Gini index of the signal;

[0017] S202: Establish a random population matrix: The first step of the NGO algorithm is to take a northern goshawk as a vector, and a group of northern goshawks constitute the random population matrix of the algorithm, as shown in the following formula; in the initial stage, each goshawk is randomly distributed in the population matrix X;

[0018]

[0019] X k is the kth northern goshawk individual; x k,d is the value of the dth variable of the kth feasible solution; N is the number of populations, and D is the variable dimension;

[0020] S203: Use the lens refraction reverse learning strategy to optimize the northern goshawk individuals and form a refraction reverse population matrix:

[0021] For a random population matrix X of northern goshawks, select N matrix elements X i,j is the population vector; generate the refraction reverse population matrix, the specific generation strategy is:

[0022]

[0023] Where: X i,j is the position of the i-th northern goshawk in j (i=1,2,…,N; j=1,2,…D), lb j ,ub j are the lower and upper limits of the search space respectively; p is the scaling factor, which is related to the refraction principle.

[0024] S204: Arrange the random population matrix generated in S202 and the refracted reverse population matrix generated in S203 in descending order of fitness values, and take the first N as the initial population matrix;

[0025] S205: A mathematical model is established based on the hunting behavior of the northern goshawk, and a two-stage iterative process is performed, wherein the first stage is the recognition and attack stage; the second stage is the escape and pursuit stage; and a nonlinear step search factor and a weight factor are used to update the position of the northern goshawk individual in the first stage of the iterative process. Through the iteration of these two stages, the optimal position of the northern goshawk individual is obtained, that is, the optimal combination of the corresponding FMD decomposition parameters.

[0026] Furthermore, in step S201, the signal-correlated composite Gini index CCGI is defined as:

[0027]

[0028] Where, signal s=(s1,s2,…,s N ), SE = s 2 , whose elements are arranged in ascending order; CK is the correlation kurtosis; SE(n) is the nth element of SE sorted in ascending order; ||·||1 represents the 1 norm; SES=abs(FFT(s 2 )); SES(n) is the nth element of SES sorted in ascending order.

[0029] Moreover, in step S205, the mathematical model of the first stage of iteration is:

[0030] G=X s

[0031]

[0032] In the formula, X k,1 is the new position of the kth feasible solution in the first stage; α is a random number in [0,1]; F is the fitness function; s is a random positive number in [1,N]; I is 1 or 2; G is the prey corresponding to the sth northern goshawk;

[0033] The mathematical model of the second stage of iteration is:

[0034] R=ω(1-it / T max )

[0035] X k,2 =X k +R(2α-1)X k

[0036]

[0037] Where: it is the current iteration number; T max is the maximum number of iterations; ω is the step size; X k,2 is the new position of the kth feasible solution in the second stage; R is the set hunting radius of the northern goshawk.

[0038] Furthermore, in step S205, the individual position update formula is as follows:

[0039]

[0040] Where η is the adjustment coefficient, η≥1; a=1; t is the current iteration number, T max is the maximum number of iterations; w is the nonlinear weight factor; X best is the current overall optimal position; r2 determines the distance the northern goshawk moves, r2∈[0,2π]; r3 controls the influence of the optimal individual on the northern goshawk's next position.

[0041] Moreover, in step S3, the characteristic mode decomposition expression after parameter optimization is:

[0042] u=FMD(fs,s,L best ,K best ,n best ,M)

[0043] In the formula, fs represents the sampling frequency of the signal, s is the acquisition signal, M is the preset maximum number of iterations, and L is best , K best 、n best are the filter length, number of filters and number of modes after parameter optimization.

[0044] The advantages and positive effects of the present invention are:

[0045] The present invention uses the SCNGO optimization algorithm to find the best parameter combination in a shorter number of iterations. The related composite Gini index CCGI is constructed as an evaluation index, which can more accurately evaluate the effectiveness of each mode component after FMD decomposition, which is conducive to selecting the best mode component for subsequent analysis. By optimizing the FMD parameters and selecting the best mode component, the fault characteristics can be more effectively extracted from the vibration signal of the rotating machinery, and the accuracy and reliability of fault diagnosis are improved. This method can overcome the problem of performance degradation caused by improper parameter setting of the existing FMD method, and provides an effective fault diagnosis solution for engineering practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a rotating machinery fault diagnosis method based on SCNGO-FMD in an embodiment of the present invention;

[0047] Figure 2 is a time domain waveform of a fault signal in an embodiment of the present invention;

[0048] Figure 3 is an envelope spectrum of a fault signal in an embodiment of the present invention;

[0049] Figure 4 This is a curve diagram showing the fitness function change of the SCNGO optimization algorithm for optimizing the FMD parameters in an embodiment of the present invention;

[0050] Figure 5 : is a CCGI amplitude diagram of each component after FMD decomposition in an embodiment of the present invention;

[0051] Figure 6 is the envelope spectrum of the optimal component in the embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the above features and advantages of the present invention more clearly understood, specific embodiments will be described in detail below in conjunction with the accompanying drawings.

[0053] The present invention provides a rotating machinery fault diagnosis method based on SCNGO-FMD, the flow chart is as follows Figure 1 Take the outer ring failure experiment of a rolling bearing in a university as an example, where the speed f r The speed is 1797r / min, the sampling frequency fs is 12000Hz, the number of sampling points N is 9216 points, calculate its fault frequency f i The method comprises the following steps.

[0054] S1: Use acceleration sensors to collect vibration signals from rotating mechanical equipment.

[0055] In the embodiment of the present invention, in step S1, in order to simulate the environmental noise under actual working conditions and make the vibration signal closer to the characteristics of the faulty rotating machinery, -8db Gaussian white noise is added to the signal. It should be emphasized that this step is only used to demonstrate the diagnostic effect of the present invention and is not part of the actual operation. Figure 2 , Figure 3 is the time domain waveform and envelope spectrum of the experimental signal, Figure 2 The impact component of the faulty bearing is masked by the strong noise. Figure 3 The impact frequency component in the image is interfered by high-frequency noise, making it difficult to clearly identify its amplitude and characteristics.

[0056] S2: Based on the SCNGO optimization algorithm (improved northern goshawk optimization algorithm), the FMD decomposition parameters (filter length L, number of filters K and number of modes n) are optimized with the relevant composite Gini index as the fitness function to obtain the optimal combination.

[0057] The steps of the SCNGO optimization algorithm based on the relevant composite Gini index include:

[0058] S201: Establishing a fitness function for parameter optimization for optimizing the SCNGO algorithm;

[0059] The maximum relevant composite Gini index of the signal is used as the fitness function for parameter optimization, and its expression is:

[0060]

[0061] Where CCGI is the relevant composite Gini index of the signal.

[0062] The signal correlation composite Gini index CCGI is defined as:

[0063]

[0064] Where, signal s=(s1,s2,…,s N ), SE = s 2 , whose elements are arranged in ascending order; CK is the correlation kurtosis; SE(n) is the nth element of SE sorted in ascending order; ||·||1 represents the 1 norm; SES=abs(FFT(s 2 )); SES(n) is the nth element of SES sorted in ascending order.

[0065] S202: The first step of the NGO algorithm is to use a northern goshawk as a vector, and a group of northern goshawks constitute the random population matrix of the algorithm. In the initial stage, each goshawk is randomly distributed in the population matrix X.

[0066]

[0067] X k is the kth northern goshawk individual; x k,d is the value of the dth variable of the kth feasible solution; N is the number of populations, and D is the variable dimension.

[0068] S203: Use the lens refraction reverse learning strategy to optimize the northern goshawk individuals and form a refraction reverse population matrix:

[0069] For the random population matrix X of northern goshawks, select N matrix elements X i,j is the population vector;

[0070] Generate the refraction reverse population. The specific generation strategy is:

[0071]

[0072] Where: X i,jis the position of the i-th NGO in j (i=1,2,…,N;j=1,2,…D)lb j ,ub j are the lower and upper limits of the search space respectively; p is the scaling factor, which is related to the refraction principle.

[0073] S204: Arrange the random population matrix generated in S202 and the refraction reverse population matrix generated in S203 in descending order according to fitness values, and take the first N as initial population matrices.

[0074] S205: A mathematical model is established based on the hunting behavior of the northern goshawk. Its iterative process can be divided into two stages: the first stage is the recognition and attack stage; the second stage is the escape and pursuit stage.

[0075] In the first stage, the northern goshawk will randomly select a prey G and then quickly attack it. This stage increases the global search capability of the NGO and can quickly approach the target area of ​​the optimal solution. The mathematical model of this stage is:

[0076] G=X s

[0077]

[0078] X k,1 is the new position of the kth feasible solution in the first stage; α is a random number in [0,1]; F is the fitness function; s is a random positive number in [1,N]; I is 1 or 2. G is the prey corresponding to the sth northern goshawk.

[0079] In the second stage, after attacking the prey, the northern goshawk will continue to chase the prey, which will try to escape. Because the northern goshawk is extremely fast, it can catch up with the prey and complete the hunt under any circumstances. This stage enhances the local search ability of the NGO and can quickly approach the optimal solution. In the NGO, assuming that the hunting radius of the northern goshawk is R, the mathematical model of this stage is as follows:

[0080] R=ω(1-it / T max )

[0081] X k,2 =X k +R(2α-1)X k

[0082]

[0083] Where: it is the current iteration number; T max is the maximum number of iterations; ω is the step size; X k,2 is the new position of the kth feasible solution in the second stage.

[0084] SCNGO uses nonlinear step search factors and weight factors to update the positions of individual northern goshawks in the first stage of iterative prey identification and attack phases. The individual position update formula is as follows:

[0085]

[0086] Where η is the adjustment coefficient, η≥1; a=1; t is the current iteration number, T max is the maximum number of iterations; w is the nonlinear weight factor; X best is the current overall optimal position; r2 determines the distance the northern goshawk moves, r2∈[0,2π]; r3 controls the influence of the optimal individual on the northern goshawk's next position.

[0087] The population number and maximum number of iterations of the SCNGO optimization algorithm in the present invention, as well as the maximum number of iterations of the characteristic mode decomposition, can be selected according to the size of the computing task. In the embodiment of the present invention, the population number of the optimization algorithm is set to 20, the maximum number of iterations is set to 50, the maximum number of iterations of the characteristic mode decomposition is set to 20, the search range of L is [20,80], the optimization range of K is [3,8], and the optimization range of n is [2,7]. Figure 4 The figure is the fitness function change curve of SCNGO algorithm for FMD optimization, which shows that SCNGO algorithm converges in the fifth iteration. The best combination of parameters L = 80, K = 8, n = 3 is obtained. The best parameters are input into the FMD decomposition algorithm to process the original signal.

[0088] S3: Perform FMD decomposition on the vibration signal of the rotating equipment according to the obtained optimal parameter combination to obtain n mode components.

[0089] In step S3, the characteristic mode decomposition expression after the parameter optimization is:

[0090] u=FMD(fs,s,L best ,K best ,n best ,M)

[0091] In the formula, fs represents the sampling frequency of the signal, s is the acquisition signal, M is the preset maximum number of iterations, and L is best , K best 、n best are the filter length, the number of filters and the number of modes after parameter optimization. In the embodiment of the present invention, their values ​​are 80, 8 and 3 respectively.

[0092] S4: Calculate the CCGI values ​​of the n mode components after FMD decomposition, select the mode component with the largest value for envelope analysis, and extract the fault characteristic frequency.

[0093] The CCGI of each component is used to evaluate the effectiveness of the model components, calculate the CCGI of each component, and select the component with the largest CCGI index as the best component. Figure 5 As shown in the figure, it can be judged from the figure that component 2 is the best component. Envelope spectrum analysis is performed on component 2. Figure 6 is the envelope spectrum of component 2, and the extracted fault frequency f i It is 108.073 Hz, which is basically consistent with the bearing outer ring fault frequency calculated in the example, and the amplitude is also increased by 2 times than the original one.

[0094] In summary, the main process of the rotating machinery fault diagnosis method based on SCNGO-FMD of the present invention is to use the constructed related composite Gini index as the fitness function of the SCNGO optimization algorithm, and adopt the SCNGO optimization algorithm to realize the adaptation of the FMD decomposition parameters; then input the optimal parameters into the FMD algorithm to realize the decomposition of the rotating machinery vibration signal, and use the related composite Gini index index to select the mode component with the largest value as the analysis object; finally, perform envelope spectrum analysis on the selected mode component to extract the fault characteristic frequency, so as to realize the fault diagnosis of the rotating machinery.

[0095] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A rotating machinery fault diagnosis method based on SCNGO-FMD, characterized in that: The following steps are involved: S1: Collect vibration signals of rotating mechanical equipment; S2: Based on the SCNGO optimization algorithm, the FMD decomposition parameters are optimized with the relevant composite Gini index CCGI as the fitness function to obtain the optimal combination of decomposition parameters, where the FMD decomposition parameters include the filter length L, the number of filters K and the number of modes n; S3: Perform FMD decomposition on the vibration signal of the rotating equipment according to the optimal parameter combination obtained in S2 to obtain n mode components, calculate the CCGI value of the mode components after FMD decomposition, and select the mode component with the largest value as the optimal component; S4: Extract the fault characteristic frequency by performing envelope analysis on the best mode component.

2. The rotating machinery fault diagnosis method based on SCNGO-FMD according to claim 1 is characterized in that: In step S2, the steps of optimizing the FMD decomposition parameters based on the improved SCNGO optimization algorithm and taking the relevant composite Gini index as the fitness function include: S201: Establishing a fitness function for parameter optimization for optimizing the SCNGO algorithm; The maximum relevant composite Gini index of the signal is used as the fitness function for parameter optimization, and its expression is: Where CCGI is the relevant composite Gini index of the signal; S202: Establish a random population matrix: The first step of the NGO algorithm is to take a northern goshawk as a vector, and a group of northern goshawks constitute the random population matrix of the algorithm, as shown in the following formula; in the initial stage, each goshawk is randomly distributed in the population matrix X; X k is the kth northern goshawk individual; x k,d is the value of the dth variable of the kth feasible solution; N is the number of populations, and D is the variable dimension; S203: Use the lens refraction reverse learning strategy to optimize the northern goshawk individuals and form a refraction reverse population matrix: For a random population matrix X of northern goshawks, select N matrix elements X i,j is the population vector; generate the refraction reverse population matrix, the specific generation strategy is: Where: X i,j is the position of the i-th northern goshawk in j (i=1,2,…,N; j=1,2,…D), lb j ,ub j are the lower and upper limits of the search space respectively; p is the scaling factor, which is related to the refraction principle; S204: Arrange the random population matrix generated in S202 and the refracted reverse population matrix generated in S203 in descending order of fitness values, and take the first N as the initial population matrix; S205: A mathematical model is established based on the hunting behavior of the northern goshawk, and a two-stage iterative process is performed, wherein the first stage is the recognition and attack stage; the second stage is the escape and pursuit stage; and a nonlinear step search factor and a weight factor are used to update the position of the northern goshawk individual in the first stage of the iterative process; through the iteration of these two stages, the optimal position of the northern goshawk individual is obtained, that is, the optimal combination of the corresponding FMD decomposition parameters.

3. The rotating machinery fault diagnosis method based on SCNGO-FMD according to claim 2 is characterized in that: In step S201, the signal correlation composite Gini index CCGI is defined as: Where, signal s=(s1,s2,…,s N ), SE = s 2 , whose elements are arranged in ascending order; CK is the correlation kurtosis; SE(n) is the nth element of SE sorted in ascending order; ||·||1 represents the 1 norm; SES=abs(FFT(s 2 )); SES(n) is the nth element of SES sorted in ascending order.

4. The rotating machinery fault diagnosis method based on SCNGO-FMD according to claim 2 is characterized in that: In step S205, the mathematical model of the first stage of iteration is: G=X s Where, X k,1 is the new position of the kth feasible solution in the first stage; α is a random number in [0,1]; F is the fitness function; s is a random positive number in [1,N]; I is 1 or 2; G is the prey corresponding to the sth northern goshawk; The mathematical model of the second stage of iteration is: R=ω(1-it / T max ) X k,2 =X k +R(2α-1)X k Where: it is the current iteration number; T max is the maximum number of iterations; ω is the step size; X k,2 is the new position of the kth feasible solution in the second stage; R is the set hunting radius of the northern goshawk.

5. The rotating machinery fault diagnosis method based on SCNGO-FMD according to claim 4 is characterized in that: In step S205, the individual position update formula is as follows: Where η is the adjustment coefficient, η≥1; a=1; t is the current iteration number, T max is the maximum number of iterations; w is the nonlinear weight factor; X best is the current overall optimal position; r2 determines the distance the northern goshawk moves, r2∈[0,2π]; r3 controls the influence of the optimal individual on the northern goshawk's next position.

6. The rotating machinery fault diagnosis method based on SCNGO-FMD according to claim 1, characterized in that: In step S3, the characteristic mode decomposition expression after parameter optimization is: u=FMD(fs,s,L best ,K best ,n best ,M) In the formula, fs represents the sampling frequency of the signal, s is the acquisition signal, M is the preset maximum number of iterations, and L is best , K best 、n best are the filter length, number of filters and number of modes after parameter optimization.

Images