A method for enhancing rolling bearing fault features
By optimizing the filter length and fault cycle parameters of MOMEDA using the WOA algorithm and combining it with the permutation entropy objective function, the problem of weak early fault characteristics in rolling bearings was solved, and the fault signal was effectively enhanced and accurately diagnosed.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING RESEARCH INSTITUTE OF MECHANICAL & ELECTRICAL TECHNOLOGY CO LTD CAM
- Filing Date
- 2022-09-26
- Publication Date
- 2026-05-19
AI Technical Summary
In existing technologies, rolling bearing fault diagnosis is difficult, especially in the early stages when fault features are weak and easily drowned out by noise. It is difficult to accurately determine the filter length and fault cycle parameters in the MOMEDA algorithm, which makes fault feature extraction difficult.
The WOA algorithm is used to optimize the filter length and fault cycle parameters of MOMEDA. The minimum permutation entropy is used as the objective function. The optimal parameter combination is found by using the whale algorithm and combined with the MOMEDA algorithm to enhance fault features.
It enables rapid and accurate identification of early fault signals in rolling bearings, enhances fault characteristics, and improves the accuracy and robustness of fault diagnosis.
Smart Images
Figure CN115586004B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of reliability engineering PHM (Prognostics and Health Management), specifically involving a method for enhancing the fault characteristics of rolling bearings based on permutation entropy optimization WOA-MOMEDA. Background Technology
[0002] Rolling bearings (hereinafter referred to as bearings), as core components of rotating machinery, operate in high-speed, high-load environments for extended periods. The reliability and long lifespan of bearings directly determine the service life of rotating machinery. Vibration is an external representation of the dynamic characteristics of mechanical equipment. By analyzing and processing the vibration signals generated by mechanical equipment, the operating status of the equipment and the state information of system components can be obtained. Due to their rotating operating characteristics, bearings exhibit periodic impact information within their vibration signals when internal cracks or defects occur. However, on the one hand, the transmission path of vibration signals within bearings is time-varying, and there are many internal vibration sources; the vibration components corresponding to local faults are often submerged in strong non-Gaussian noise interference, making fault diagnosis difficult. On the other hand, in the early stages of rolling bearing failure, the fault characteristics are weak, making it difficult to extract state features. Therefore, fault feature extraction and enhancement for bearings has significant theoretical and engineering practical value.
[0003] When collecting bearing operating status information, the Multipoint Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA) algorithm is commonly used to extract the periodic impact component from the vibration signal for fault diagnosis. However, when using the MOMEDA method for deconvolution, two parameters need to be set in advance: the filter length *l* and the fault period *τ*. Filters that are too long or too short will result in poor generalization ability of the model, and the determination of the fault period directly affects the accuracy of MOMEDA deconvolution. Therefore, it is necessary to perform combined optimization of the filter length *l* and the fault period *τ* in the MOMEDA algorithm to improve the enhancement effect of rolling bearing fault features. Summary of the Invention
[0004] To address the technical problem of difficulty in determining the optimal values of filter length and fault period when using the MOMEDA algorithm for fault diagnosis in existing technologies, this invention provides a rolling bearing fault feature enhancement method. By using the WOA optimization algorithm to optimize the combination of filter length and fault period parameters of MOMEDA, and taking the minimum permutation entropy as the objective function, the optimal parameter combination of MOMEDA can be obtained quickly and accurately, resulting in the enhanced fault signal.
[0005] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:
[0006] A method for enhancing the fault characteristics of rolling bearings includes the following steps.
[0007] S1. The WOA algorithm initializes the filter length and fault period range to determine the whale's feeding range;
[0008] S2. Using the one-dimensional vibration signal generated during the operation of the rolling bearing as the input signal, a set of filter length and fault cycle parameter combinations are randomly generated.
[0009] S3. Input the parameter combination and one-dimensional vibration signal into the MOMEDA algorithm to calculate the fault feature enhancement signal;
[0010] S4. Calculate the permutation entropy of the enhanced signal and update the permutation entropy value;
[0011] S5. Determine if the permutation entropy is minimized. If so, terminate the iteration and save the optimal parameter combination; otherwise, update the whale position, return to step S3, and recalculate the fault feature enhancement signal.
[0012] Furthermore, the filter length is set to a range of [250, 500], and the fault period is set to a range of [100, 170].
[0013] Furthermore, the whale position update method in step S5 is as follows:
[0014] The position update method is determined by the value of P, which is a random value from [0,1]. If P < 0.5, each whale chooses to surround the prey; if P ≥ 0.5, each whale chooses to release a bubble net to drive away the prey.
[0015] If the prey is surrounded, calculate...
[0016]
[0017] Where A is the coefficient vector, δ max Let be the optimal generation number, s be the current time, and r be a random number uniformly distributed between [0,1].
[0018] When |A| < 1, the whale's position update formula is as follows:
[0019]
[0020] When |A|≥1, the whale's position update formula is as follows:
[0021]
[0022] in, This indicates the whale's position at the current moment and its next position. This is the optimal position for the whale at present. Let C be the current random position of the whale, and let C be a random number between (0,2).
[0023] If the whale uses a bubble net to drive away prey, the whale's location will be updated as follows:
[0024]
[0025] Where b is a constant coefficient and q is a random number in [-1,1].
[0026] The beneficial effects of this invention compared to the prior art are as follows:
[0027] This invention proposes a fault feature enhancement method based on the joint optimization of MOMEDA using Permutation Entropy (PE) and Whale Optimization Algorithm (WOA) for MOMEDA, building upon the deconvolution of vibration signals using the MOMEDA algorithm. The WOA optimization algorithm optimizes the combination of MOMEDA parameters, including filter length l and fault period τ, using minimum PE as the objective function. When the minimum PE is found, the optimal parameter combination for MOMEDA is extracted, ultimately yielding the enhanced fault signal. Experimental results validate the method, demonstrating its ability to accurately identify the fault period of vibration signals and enhance fault features. The method exhibits good deconvolution performance and robustness.
[0028] The fault feature enhancement algorithm proposed in this invention combines the advantages of WOA (Wide Aspect Analysis) for fast optimization, permutation entropy for sensitivity to abrupt signals, and MOMEDA (Modified Oscillator) for enhancing fault signals. It plays an important role in the detection and prevention of early-stage weak faults in bearings. Attached Figure Description
[0029] The accompanying drawings, which form part of this specification, are provided to further illustrate embodiments of the invention and, together with the textual description, explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.
[0030] Figure 1 A WOA-optimized MOMEDA flowchart is provided for a specific embodiment of the present invention;
[0031] Figure 2 A schematic diagram of the optimized parameters of the inner ring early fault signal WOA algorithm provided in a specific embodiment of the present invention;
[0032] Figure 3 A comparison diagram of the early fault signal in the inner ring before and after time-domain enhancement provided for a specific embodiment of the present invention;
[0033] Figure 4 A comparison diagram of the early fault signal of the inner ring before and after frequency domain enhancement provided for a specific embodiment of the present invention;
[0034] Figure 5 A schematic diagram of the optimized parameters of the whale algorithm for early fault signals in the outer ring provided in a specific embodiment of the present invention;
[0035] Figure 6 A comparison diagram of the early fault signal of the outer ring before and after time-domain enhancement provided for a specific embodiment of the present invention;
[0036] Figure 7 A comparison diagram of the early fault signal of the outer ring before and after frequency domain enhancement provided for a specific embodiment of the present invention. Detailed Implementation
[0037] Specific embodiments of the present invention will now be described in detail. In the following description, specific details are set forth for purposes of explanation and not limitation, in order to aid in a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced in other embodiments departing from these specific details.
[0038] It should be noted that, in order to avoid obscuring the invention with unnecessary details, only the device structure and / or processing steps closely related to the solution of the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.
[0039] To facilitate understanding, we will first explain the working principle of the MOMEDA algorithm for enhancing the fault characteristics of rolling bearings.
[0040] If a rolling bearing degrades during operation and causes misalignment of its center of gravity, it will generate vibration signals that indicate a fault.
[0041] x=h*y+e (1)
[0042] The vibration signal x can be considered as the original vibration signal, which is composed of a regular impact signal corrected by the propagation coefficient, plus noise. In the formula, y is the impact signal actually generated during bearing operation, h is the propagation response of the impact signal during the acquisition stage, and e is the noise interference encountered in the propagation path.
[0043] The MOMEDA algorithm can find the original impulse signal y by iterating through the optimal filter FIR.
[0044]
[0045] Where: k = 1, 2, ..., NL, f are filter parameters, N represents the deconvolution range, and L represents the filter order.
[0046] The core idea of MOMEDA is to design an optimal filter to filter impulse signals, which contain rich fault information. The process of finding the optimal filter involves maximizing the multi-point D-norm.
[0047]
[0048]
[0049] In the formula: t represents the position and weight of the target impact component in the deconvolution. The deconvolution effect is optimal when t matches the impact signal. Therefore, the position of the impact signal can be determined and the impact signal can be separated using the constant t.
[0050]
[0051]
[0052]
[0053] Therefore, equation (7) can be expressed as:
[0054]
[0055] Assume t1M1 + t2M2 + ... + t N-L M N-L =X0t, let the derivative be zero, equation (8) can be written as:
[0056] ||y|| -1 X0t-||y|| -3 t T yX0y=0 (9)
[0057] Equation (9) can be simplified to:
[0058]
[0059] Because y = X0 T f, assume there exists an inverse matrix (X0X0) T ) -1 :
[0060]
[0061] Since the impact component also generates impact at corresponding cycle multiples as the roller rotates, the multiple of f is also a solution of equation (11). The optimal filter and output solution of MOMEDA can be summarized as follows:
[0062] f = (X0X0) T ) -1 X0t (12)
[0063] Let y = X0 T Substituting f into the equation allows for the recovery of the original impact signal. The MOMEDA algorithm can be used to deconvolve the vibration signal, thereby obtaining the impact signal containing fault information. Furthermore, as shown in the above equation, determining the optimal filter and the fault period is crucial for finding the impact signal.
[0064] This invention, based on the deconvolution of vibration signals using the MOMEDA algorithm, proposes a fault feature enhancement method that jointly optimizes MOMEDA using Permutation Entropy (PE) and the Whale Optimization Algorithm (WOA). WOA is used to optimize the filter length l and fault period τ in the MOMEDA algorithm, and minimizing PE is set as the objective function during the optimization process. Figure 1 As shown, it includes the following steps:
[0065] Step 1: Initialize the filter length l and fault period τ using the WOA algorithm to determine the range of whale predation.
[0066] Preferably, in order to improve computational efficiency, the range of values for filter length l and fault period τ is set to l = [250, 500] and τ = [100, 170].
[0067] Step 2: Using the one-dimensional vibration signal generated during the operation of the rolling bearing as the input signal, the WOA algorithm is used to randomly generate a set of different combinations of filter length L and fault period τ parameters within the initial value range.
[0068] Step 3: Input the randomly generated parameter combination and the one-dimensional vibration signal into the MOMEDA algorithm, and output the enhanced signal after the fault characteristics are enhanced.
[0069] Step 4: Calculate the permutation entropy PE of the enhanced signal and update the permutation entropy value.
[0070] The principle of the permutation entropy algorithm can be found in the literature "Fault Feature Extraction of Rolling Bearing Based on Improved Permutation Entropy Algorithm" by Chen Xianglong et al., Journal of Vibration Engineering, 2018, 31(5).
[0071] The minimum permutation entropy T is obtained by updating the permutation entropy value. min .
[0072] Step 5: Determine whether to terminate the iteration by checking if the permutation entropy obtained in the last calculation is the minimum. If the permutation entropy T obtained in the last calculation is greater than Tmin... min Once the whale's position is updated, return to step 3 and input the obtained parameter combination and one-dimensional vibration signal into the MOMEDA algorithm to calculate the fault feature enhancement signal; otherwise, the iteration terminates and the optimal parameter combination is saved.
[0073] The fault feature enhancement algorithm proposed in this invention combines the advantages of WOA (Wide Aspect Analysis) for fast optimization, permutation entropy for sensitivity to abrupt signals, and MOMEDA (Modified Oscillator) for enhancing fault signals. It plays an important role in the detection and prevention of early-stage weak faults in bearings.
[0074] The principle behind the WOA algorithm's optimization of parameter combinations in the above method is as follows:
[0075] The Whale Optimization Algorithm (WOA) is a typical metaheuristic optimization algorithm, and its optimization process mainly includes three steps:
[0076] 1. Randomly place pods of whales to search for prey.
[0077] Assume the position of each whale in D-dimensional space is:
[0078] X = (x1, x2, ..., x D (13)
[0079] In D-dimensional space, each whale can choose to surround its prey or expel a bubble net to drive it away, and the probability of each whale choosing these two behaviors is equal.
[0080] P(encircle)=P(bubble)=0.5 (14)
[0081] P is a random value from [0,1]. The values of P and the method of updating their positions are shown in Table 1 below:
[0082] Table 1. Correspondence between P-value and location update method
[0083]
[0084] 2. Surround the prey
[0085] When whales surround their prey, they will randomly choose to swim towards the whale closest to the prey, or towards a random point (a random whale).
[0086] 2.1 Swim towards the position closest to the prey.
[0087] If a whale chooses to swim towards the whale closest to its prey, the whale's position update formula is as follows:
[0088]
[0089]
[0090] in, This indicates the whale's position at the current moment and its next position. Let be the optimal position of the whale at present, s be the current time, r be a random number uniformly distributed between [0,1], and C be a random number between (0,2). A is the coefficient vector, δ max This is the optimal number of generations.
[0091] 2.2 Swim towards the location of the random point (random whale).
[0092] If a whale chooses to swim towards a random point (random whale), the whale's position update formula is as follows:
[0093]
[0094] in, This indicates the whale's position at the current moment and its next position. This represents the current random position of the whale.
[0095] Whether a whale chooses to swim towards the optimal or random individual depends on the size of A. When |A| < 1, the whale swims towards the optimal individual; when |A| ≥ 1, the whale swims towards the random individual.
[0096] 3. Dispense the spiral bubble net
[0097] When hunting, whales eject bubbles from the water's surface and spiral, creating a bubble net to drive away their prey. The whale's position updates during the use of this bubble net are as follows:
[0098]
[0099] Where b is a constant coefficient (default value is 1), and q is a random number in [-1, 1].
[0100] The optimal parameter combination is output through equations (15), (17), and (18), which is the global optimal solution, i.e., the optimal parameter combination of filter length L and fault period τ in the MOMEDA function.
[0101] The pseudocode for the whale algorithm is shown in Table 2:
[0102] Table 2. Pseudocode of the WOA Algorithm
[0103]
[0104]
[0105] WOA is merely an optimization algorithm for finding parameters; the real challenge lies in selecting the objective function. When bearings degrade, phenomena such as impacts, velocity cutoffs, and structural deformation often occur. These fault phenomena imbue the vibration signal with impact information. Permutation entropy (PE) is an entropy that estimates the degree of disorder in a time series; the more disordered the internal information, the greater the permutation entropy, while the more regular the internal information, the smaller the permutation entropy. Therefore, when regular impact information is generated within the bearing, PE will decrease significantly. Thus, we can determine whether the bearing is generating impact signals by observing changes in PE values, effectively detecting abrupt changes in the vibration signal. Therefore, minimizing PE is set as the objective function of the WOA algorithm.
[0106] The technical solution of the present invention will be described below with reference to a specific embodiment.
[0107] In this embodiment, the data is used to verify early inner ring faults and early outer ring faults of the bearing, and the WOA algorithm is used to optimize the filter length and fault period in MOMEDA.
[0108] (1) Inner ring fault signal enhancement
[0109] Input early fault data of the inner ring into WOA-MOMEDA.
[0110] Table 3 Optimal Location of Inner Ring Faults After WOA Optimization Table 3 Optimal Location of Inner Ring Faults After WOA Optimization
[0111]
[0112] The WOA optimization parameters are shown in Table 3. Figure 2 (a) It can be seen that, depending on the combination of different filter lengths and fault cycles, the value of the permutation entropy also changes accordingly. Figure 2 (b) It can be seen that the convergence begins to be rapid after 23 iterations. Table 3 shows the optimal combination of parameters and the magnitude of the permutation entropy under this combination.
[0113] Input the optimal parameters found after WOA optimization into MOMEDA. Figure 3 This is a time-domain plot of the signal during an early inner ring fault, and the frequency-domain plot of the signal enhanced by MOMEDA is directly compared with the original signal frequency-domain plot. Figure 4It can be seen that, compared with the original signal, the periodic impact characteristics of the enhanced fault signal are significantly enhanced, and the inner ring fault characteristic frequency of 184.38Hz (actual detection data is 184.91Hz) and its amplitude at 1 to 5 harmonics are significantly prominent. The experimental results show that this method can effectively diagnose inner ring faults.
[0114] (2) Enhanced outer ring fault signal
[0115] Early fault data from the outer ring are input into WOA-MOMEDA for WOA-MOMEDA fault feature enhancement. When the Whale Algorithm parameter combination is [322.45, 115.1], the permutation entropy is minimized, with a value of 0.815.
[0116] Input the optimal parameters found after WOA optimization into MOMEDA. Figure 5 A collection of optimal paths for the whale algorithm. Figure 6 The time-domain plot after WOA-MOMEDA processing. Figure 7 It can be observed that the periodic impact characteristics of the fault signal are significantly enhanced, and the outer ring fault characteristic frequency of 115.63Hz (actual diagnostic data is 115.1Hz) and its peak values at 1-6 harmonics are prominent. Experimental results show that this method can effectively diagnose outer ring faults.
[0117] This invention proposes a fault feature enhancement algorithm based on permutation entropy (PE) optimized WOA-MOMEDA. The WOA optimization algorithm is used to identify the optimal parameter combination when PE is minimized, and the optimal parameter combination is input into MOMEDA. This can effectively enhance the features of weak fault signals. Experimental verification was carried out on monitoring data of outer and inner circle faults. The experimental results show that the method can accurately identify the fault cycle, verifying that the new method has good deconvolution performance and good robustness.
[0118] The features described and / or illustrated above with respect to one embodiment may be used in the same or similar manner in one or more other embodiments, and / or in combination with or in lieu of features in other embodiments.
[0119] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, whole, step, or component, but does not exclude the presence or addition of one or more other features, wholes, steps, components, or combinations thereof.
[0120] Many features and advantages of these embodiments are apparent from this detailed description, and therefore the appended claims are intended to cover all such features and advantages of these embodiments that fall within their true spirit and scope. Furthermore, since many modifications and alterations will readily occur to those skilled in the art, the embodiments of the invention are not intended to be limited to the precise structures and operations illustrated and described, but rather to encompass all suitable modifications and equivalents falling within their scope.
[0121] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0122] The parts of this invention not described in detail are techniques known to those skilled in the art.
Claims
1. A method for enhancing the fault characteristics of rolling bearings, characterized in that, Includes the following steps S1. The WOA algorithm initializes the filter length and fault period range to determine the whale's feeding range; S2. Using the one-dimensional vibration signal generated during the operation of the rolling bearing as the input signal, a set of filter length and fault cycle parameter combinations are randomly generated. S3. Input the parameter combination and one-dimensional vibration signal into the MOMEDA algorithm to calculate the fault feature enhancement signal; S4. Calculate the permutation entropy of the enhanced signal and update the permutation entropy value; S5. Determine if the permutation entropy is minimum. If yes, terminate the iteration and save the optimal parameter combination; otherwise, update the whale position, return to step S3, and recalculate the fault feature enhancement signal. The method for updating the whale's position in step S5 is as follows: The position update method is determined by the value of P, which is a random value from [0,1]. If P < 0.5, each whale chooses to surround the prey; if P ≥ 0.5, each whale chooses to release a bubble net to drive away the prey. If the prey is surrounded, calculate... Where A is the coefficient vector. To be the optimal number of generations, At the current time, r is a random number uniformly distributed between [0,1]. when The formula for updating the whale's position is as follows: ， when The formula for updating the whale's position is as follows: ,in, This indicates the whale's position at the current moment and its next position. This is the optimal position for the whale at present. Let C be the current random position of the whale, and let C be a random number between (0,2). If the whale uses a bubble net to drive away prey, the whale's location will be updated as follows: Where b is a constant coefficient. It is a random number in the range [-1, 1].
2. The method for enhancing the fault characteristics of rolling bearings according to claim 1, characterized in that, The filter length is set to a range of [250, 500], and the fault period is set to a range of [100, 170].