Bearing fault diagnosis method based on cluster discharge resonance neuron model

Abstract

The invention provides a bearing fault diagnosis method based on a cluster discharge resonance neuron model, and the method comprises the steps: collecting 2D mechanical signals of an axial channel and a radial channel of a turbine, constructing a self-adaptive cluster discharge resonance neuron model, optimizing the parameter configuration, inputting the collected data into the self-adaptive cluster discharge resonance neuron model, and carrying out the fault diagnosis of the turbine. And outputting a fault diagnosis result. According to the method, the self-adaptive cluster discharge resonance neuron model is adopted, early fault detection is improved, the effectiveness of the self-adaptive cluster discharge resonance neuron model is enhanced by coupling multiple resonance neurons, the neurons can generate cluster discharge behaviors, and the fault detection accuracy is improved. Signal-to-noise ratios under different noise intensities, discharge thresholds and neuron numbers help evaluate multiple effects of reactivity of the adaptive cluster discharge resonance neuron model. According to the invention, acquisition and processing of two-dimensional vibration signals are integrated, fault feature extraction of two-channel direction early signals is realized, and the limitation that one-dimensional signals are interfered by complex noise is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of mechanical fault detection and signal processing, and in particular relates to a bearing fault diagnosis method based on a cluster discharge resonance neuron model. Background Art

[0002] As a renewable, clean energy source, wind power boasts advantages such as environmental friendliness and wide distribution. Consequently, wind power generation technology is attracting increasing attention and is being widely adopted worldwide. Rolling bearings are key components of wind turbines and are also vulnerable to wear. Failures in these bearings can increase operating and maintenance costs. Therefore, diagnosing the operating status of wind turbine rolling bearings, detecting early failures, and promptly repairing them can effectively improve wind turbine utilization.

[0003] With the rapid development of science and technology, nonlinear science has garnered increasing attention. Bifurcations and chaos have been found to reveal the nonlinear nature of matter from a dynamical perspective. Generally speaking, signal transmission can be manipulated or overwhelmed by noise. However, noise in nonlinear systems, particularly neural systems, plays a significant role. Research on nonlinear resonance has revealed that noise can induce novel phenomena in systems, such as stochastic resonance, autoresonance, parametric resonance, phase-locked resonance, and vibration resonance. Among these, the enhancement of signal characteristic frequencies based on stochastic resonance can violate the concept of noise immunity, triggering nonlinear system dynamics, achieving spatiotemporal synchronization between weak useful signals to be detected and energy sources, and enhancing weak characteristic signals by harvesting random energy. While traditional stochastic resonance methods for processing vibration signals have been extensively studied, further research is needed to determine how to optimally detect them in complex noise environments.

[0004] Furthermore, it is widely believed that noise not only affects the dynamic behavior of nonlinear systems but also exists at all levels of the nervous system, influencing the generation, transmission, and response of neural signals. Consequently, researchers have recently focused their attention on the impact of noise on neural dynamic behavior. Numerous studies have shown that in the nervous system, subthreshold signals can be effectively transmitted with the aid of certain noise levels, leading to stochastic resonance. Neurons are highly differentiated cells and the fundamental units of the nervous system. Fluctuations or randomness are inevitable in the nervous system, and noise plays a positive and important role in the pattern switching of neural discharge activity and the stochastic resonance of external signals.

[0005] Therefore, the energy transfer of noise and the influence of coupled discharge pulse energy are used to enhance the ability of neurons to perceive subtle changes in the external environment. Inspired by this mechanism and based on the Langevin equation representation of stochastic resonance, this paper proposes an adaptive discharge resonance neuron (CDRN) model to improve the ability to extract characteristic frequencies from multidimensional bearing fault vibration signals. Through a comprehensive study of experiments on the inner and outer races of wind turbine bearings and the CDRN method, the model demonstrates sufficient enhancement for the accurate detection of early mechanical faults. Summary of the Invention

[0006] In view of this, the present invention aims to overcome the shortcomings of the above-mentioned problems in the prior art and proposes a bearing fault diagnosis method based on a cluster discharge resonance neuron model to enhance the detection capability of early faults in the mechanical operating environment noise.

[0007] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0008] A first aspect of the present invention provides a bearing fault diagnosis method based on a burst discharge resonance neuron model, comprising the following steps:

[0009] Step 1: Collect 2D mechanical signals of the turbine in both axial and radial channels;

[0010] Step 2: Construct an adaptive bursting resonance neuron model and optimize parameter configuration;

[0011] Step 3: Input the data collected in step 1 into the adaptive cluster discharge resonance neuron model and output the fault diagnosis result.

[0012] Furthermore, the step 1 specifically includes:

[0013] Accelerometers are set at the axial and radial dual-channel measuring points of the turbine to monitor the vibration of the generator in different directions. The two-dimensional mechanical signals of the inner and outer rings of the bearing are collected using a data acquisition system, and the envelope of the signal is extracted through Hilbert transform.

[0014] Furthermore, in step 2, the equation of the adaptive burst discharge resonance neuron model is described as follows:

[0015]

[0016] Where U(x i ) represents the stochastic resonance control system, I in Represents the external input signal, A, f0 and Represent the amplitude, frequency and phase of the input signal respectively, x th is the discharge threshold potential, N is the number of resonant neurons, I syn,irepresents the conducted synaptic current received by the i-th neuron from other neurons at time t, c s represents the synaptic conductance, E syn represents the synaptic reversal potential, w ij represents the synaptic weight, s j (t) represents whether the synaptic ion channel gate of the i-th neuron is open at time t, t jk represents the kth discharge moment of the jth neuron, and the noise intensity value is ε x , the Wiener process is represented by dω x (t), it is a continuous time random process, for any t and τ, dω x (t)-dω x (τ)~N(0,1) obeys the standard normal distribution.

[0017] Furthermore, in step 2, scaling and Chebyshev filtering are combined to optimize parameter configuration. Scaling is used to adjust the dynamic range of the signal, and Chebyshev filtering is used to filter out unnecessary frequency components and retain useful signals.

[0018] Furthermore, it also includes optimizing the optimal parameters by using the signal-to-noise ratio indicator.

[0019] A second aspect of the present invention provides a bearing fault diagnosis device based on a cluster discharge resonance neuron model, comprising:

[0020] A signal acquisition unit is used to collect 2D mechanical signals of the turbine in both axial and radial channels;

[0021] A model building unit, used to build an adaptive cluster discharge resonance neuron model and optimize parameter configuration;

[0022] The result output unit is used to input the data collected by the signal acquisition unit into the adaptive cluster discharge resonance neuron model and output the fault diagnosis result.

[0023] A third aspect of the present invention provides an electronic device, comprising a processor and a memory communicatively connected to the processor and used to store instructions executable by the processor, wherein the processor is used to execute the above-mentioned bearing fault diagnosis method based on the cluster discharge resonance neuron model.

[0024] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned bearing fault diagnosis method based on the cluster discharge resonance neuron model.

[0025] Compared with the existing technology, the bearing fault diagnosis method based on the cluster discharge resonance neuron model described in the present invention has the following advantages:

[0026] This study uses an adaptive clustering resonant neuron model to improve early fault detection. Results show that the effectiveness of the adaptive clustering resonant neuron model is enhanced by coupling multiple resonant neurons, which can produce clustered discharge behavior. The signal-to-noise ratio at different noise intensities, discharge thresholds, and numbers of neurons helps evaluate the multiple effects of the adaptive clustering resonant neuron model's responsiveness.

[0027] The present invention integrates the acquisition and processing of two-dimensional vibration signals through an adaptive cluster discharge resonance neuron model, realizes the fault feature extraction of early signals in dual-channel directions, and solves the limitation of one-dimensional signals being interfered by complex noise.

[0028] This invention improves the detection capability of engineering vibration signals. Under background noise interference, the adaptive cluster discharge resonance neuron model can accurately identify the two-dimensional fault characteristics of wind turbine bearings. Its output signal-to-noise ratio and the recognition rate of the neural network classifier are both higher than those of the original axial signal, the original radial signal, and the two-dimensional signal processed by the SOSR method. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0030] Figure 1 Schematic diagram of the CDRN model for mechanical fault diagnosis of the present invention. DETAILED DESCRIPTION

[0031] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0032] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0033] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0034] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0035] Example 1:

[0036] like Figure 1 As shown, the bearing fault diagnosis based on the cluster discharge resonance neuron model of the present invention includes the following steps:

[0037] (1) Obtaining 2D mechanical signals. Accelerometers are installed at the turbine's axial and radial dual-channel measurement points to monitor the generator's vibration in different directions. The data acquisition system collects the 2D mechanical signals from the bearing's inner and outer rings and extracts the signal envelope using a Hilbert transform.

[0038] (2) An adaptive cluster discharge resonance neuron model was constructed. The cluster discharge model was studied based on the neuron system and combined with the Langevin equation of the stochastic resonance model. The equation of each resonance neuron model can be described as follows:

[0039]

[0040] Where U(x i ) represents the stochastic resonance control system. in Represents the external input signal. A, f0 and Represents the amplitude, frequency and phase of the input signal respectively. th is the discharge threshold potential, and N is the number of resonant neurons. syn,i represents the conducted synaptic current received by the i-th neuron from other neurons at time t, c s represents the synaptic conductance, E syn represents the synaptic reversal potential, w ij represents the synaptic weight, s j (t) represents whether the synaptic ion channel gate of the i-th neuron is open at time t, t jk represents the kth firing moment of the jth neuron. The noise intensity value is represented by ε x , the Wiener process is represented by dω x(t), it is a continuous-time random process characterized by the independence of its increments with time t. Moreover, for any t and τ, dω x (t)-dω x (τ)~N(0,1) obeys the standard normal distribution. The potential function of the stochastic resonance system is expressed as U(x), which can be expressed by the following formula:

[0041]

[0042] Where a and b are model parameters and satisfy a>0, b>0. The discharge process of the model is as follows: once the membrane potential reaches the threshold potential V_th, the action potential is immediately released and reset to the resting potential. After a short refractory period, it is again discharged according to the equation For the sake of simplicity, the influence of the refractory period is ignored in the calculation formula. Therefore, the equivalent potential function of the resonant neuron model is as follows:

[0043]

[0044] The model parameters were optimized, combining scaling and Chebyshev filtering. The two-dimensional vibration signal was input into the clustered resonant neuron model. Scaling adjusted the signal's dynamic range, while the Chebyshev filter effectively filtered out unwanted frequency components, retaining the useful signal. The parameters a, b, and Vth were set to appropriate ranges for effective system performance based on the input signal. In the fault signature detection system, the signal-to-noise ratio (SNR) was used to optimize the optimal parameters. The SNR formula is as follows:

[0045]

[0046] Energy(s) represents the output signal energy, and Energy(n) represents the output noise energy.

[0047] (3) After calculating the CDRN model using the optimal theoretical parameters, the output signal is obtained by scaling according to the adiabatic approximation theory, ensuring the accuracy of the process. Then, by extracting the signal characteristics of the wind turbine bearing signal, potential faults can be identified.

[0048] The following experiments illustrate the effectiveness of this scheme.

[0049] During wind turbine operation, any fault can lead to serious and potentially catastrophic problems. These problems not only pose a significant risk to the operational efficiency and economic viability of the wind farm, but also endanger personnel safety. Therefore, detecting and addressing faults in wind turbine bearings to ensure safe and reliable operation is crucial. To detect bearing anomalies in wind turbines operating in a wind farm environment, the wind turbines are maintained at a consistent rotational speed of 1700 to 1800 rpm. Diagnostic evaluation focuses on the drive end of the generator, with monitoring locations established along the axial and radial dimensions. To capture vibration data from the drive end of the generator, dedicated control software combined with acceleration sensors is used. Following data acquisition, signal processing for fault signature detection is implemented using CDRN and SOSR methods.

[0050] The physical dynamics of bearings in vibration analysis are used to determine the rotational frequencies of bearings within wind turbine generators. Furthermore, theoretical evaluation is used to identify unique fault frequencies corresponding to anomalies in the outer and inner races of the bearings. Vibration data from the wind turbine rolling bearings in both the axial and radial directions are processed and converted into the time and frequency domains. However, even with vibration signals from the measurement points, identifying anomalies, particularly in the outer race, can be challenging when the signals are obscured by significant noise.

[0051] The CDRN method was used to analyze two-dimensional signals obtained from axial and radial measurement points, while the axial signal was processed using the SOSR method. The signal-to-noise ratio (SNR) of the output signal obtained by applying the CDRN method was -15.8957 dB, while the SNR of the characteristic frequency components of the signal processed using the SOSR method decreased to -21.6854 dB, a decrease of 5.7897 dB. Specifically, a decrease in SNR was observed when compared to the results achieved using the CDRN method. This comparative analysis highlights the superiority of the proposed method in effectively detecting noise in complex real-world signals.

[0052] Example 2:

[0053] A bearing fault diagnosis device based on a cluster discharge resonance neuron model, comprising:

[0054] A signal acquisition unit is used to collect 2D mechanical signals of the turbine in both axial and radial channels;

[0055] A model building unit, used to build an adaptive cluster discharge resonance neuron model and optimize parameter configuration;

[0056] The result output unit is used to input the data collected by the signal acquisition unit into the adaptive cluster discharge resonance neuron model and output the fault diagnosis result.

[0057] Example 3:

[0058] An electronic device includes a processor and a memory communicatively connected to the processor and used to store instructions executable by the processor, wherein the processor is used to execute the above-mentioned bearing fault diagnosis method based on the cluster discharge resonance neuron model.

[0059] Example 4:

[0060] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned bearing fault diagnosis method based on a cluster discharge resonance neuron model.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A bearing fault diagnosis method based on a burst discharge resonance neuron model, characterized by: The steps include: Step 1: Collect 2D mechanical signals of the turbine in both axial and radial channels; Step 2: Construct an adaptive bursting resonance neuron model and optimize parameter configuration; Step 3: Input the data collected in step 1 into the adaptive cluster discharge resonance neuron model and output the fault diagnosis result.

2. The bearing fault diagnosis method based on the burst discharge resonance neuron model according to claim 1 is characterized in that: The step 1 specifically includes: Accelerometers are set at the axial and radial dual-channel measuring points of the turbine to monitor the vibration of the generator in different directions. The two-dimensional mechanical signals of the inner and outer rings of the bearing are collected using a data acquisition system, and the envelope of the signal is extracted through Hilbert transform.

3. The bearing fault diagnosis method based on the burst discharge resonance neuron model according to claim 1 is characterized in that: In step 2, the equation of the adaptive burst discharge resonance neuron model is described as follows: Among them, U(x i ) represents the stochastic resonance control system, I in Represents the external input signal, A, f0 and Represent the amplitude, frequency and phase of the input signal respectively, x th is the discharge threshold potential, N is the number of resonant neurons, I syn,i represents the conducted synaptic current received by the i-th neuron from other neurons at time t, c s represents the synaptic conductance, E syn represents the synaptic reversal potential, w ij represents the synaptic weight, s j (t) represents whether the synaptic ion channel gate of the i-th neuron is open at time t, t jk represents the kth discharge moment of the jth neuron, and the noise intensity value is ε x , the Wiener process is represented by dω x (t), it is a continuous time random process, for any t and τ, dω x (t)-dω x (τ)~N(0,1) obeys the standard normal distribution.

4. The bearing fault diagnosis method based on the burst discharge resonance neuron model according to claim 1 is characterized in that: In step 2, scaling and Chebyshev filtering are combined to optimize parameter configuration. Scaling is used to adjust the dynamic range of the signal, and Chebyshev filtering is used to filter out unnecessary frequency components and retain useful signals.

5. The bearing fault diagnosis method based on the burst discharge resonance neuron model according to claim 1 is characterized in that: It also includes the use of signal-to-noise ratio indicators to optimize the optimal parameters.

6. Bearing fault diagnosis device based on cluster discharge resonance neuron model, characterized by: include: A signal acquisition unit is used to collect 2D mechanical signals of the turbine in both axial and radial channels; A model building unit, used to build an adaptive cluster discharge resonance neuron model and optimize parameter configuration; The result output unit is used to input the data collected by the signal acquisition unit into the adaptive cluster discharge resonance neuron model and output the fault diagnosis result.

7. An electronic device comprising a processor and a memory in communication with the processor and configured to store instructions executable by the processor, wherein: The processor is used to execute the bearing fault diagnosis method based on the cluster discharge resonance neuron model described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the bearing fault diagnosis method based on the cluster discharge resonance neuron model according to any one of claims 1 to 5 is implemented.

Images