Simulation-driven BCSM deconvolution fault diagnosis method and system

Abstract

The invention belongs to the field of mechanical equipment fault diagnosis, and relates to a simulation-driven BCSM deconvolution fault diagnosis method and system. Due to the fact that interference of a rotating shaft exists in the wind turbine generator, early-stage faults of the wind turbine generator are not obvious and can be covered by working noise very easily. Considering that the carrier frequency generated during damage is often similar to the bending frequency in modal analysis, the method comprises the following steps of: iteratively updating the most suitable filter through a band-pass filter designed by finite element simulation and by utilizing BCSM, and then filtering wind power vibration data; according to the method, the health condition of the unit can be effectively diagnosed, the accuracy is high, and the method is easy to operate and has certain industrial value.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical equipment fault diagnosis, and in particular to a simulation-driven BCSM deconvolution fault diagnosis method and system. Background Art

[0002] As a common mechanical equipment, wind turbines require specialized maintenance due to their complex operating requirements and harsh environmental conditions. Regular maintenance is essential to ensure optimal performance and long-term use of the system. Components such as rolling bearings play a vital role in maintaining rotational accuracy and reducing friction, thereby improving efficiency and reliability. However, harsh environments can cause bearing failures, potentially leading to catastrophic failures and significant economic losses. Therefore, in the operation of wind turbines, effective fault detection and maintenance strategies are crucial and have attracted widespread attention in the industry. Due to their impact on operational safety and economic feasibility, long-term condition monitoring of their key parts is indispensable. Among them, rolling bearings are the connecting hub of wind turbines. Throughout the service life of wind turbines, rolling bearings are widely used to maintain rotational accuracy and reduce surface friction. However, in challenging and harsh working environments, these bearings are prone to failure. Such failures have the potential to cause a complete collapse of the mechanical system, causing significant economic losses and endangering operational safety. Therefore, identifying the failure of wind turbine bearings is crucial and has attracted widespread attention in the industry.

[0003] Fault diagnosis technology provides an effective way to improve the reliability and maintainability of equipment. Due to the large noise interference in the working environment, it is very difficult to diagnose mechanical equipment. With the advancement of science and technology, the development of modern new theories and technical applications such as information technology, sensor technology, and monitoring technology has helped the application of bearing fault diagnosis, bringing fresh blood to bearing diagnosis. At the same time, it provides theoretical support for working conditions in different environments and becomes an important part of equipment monitoring and diagnosis. Signal processing technology is the key to the actual processing method of mechanical equipment diagnosis technology and the analysis of fault characteristics.

[0004] Blind convolution is considered to be an effective method to mitigate the impact of the transfer path. The convolution process involves creating an inverse FIR filter, which can be iteratively adjusted according to a pre-set objective function. Therefore, the FIR coefficients obtained by deconvolution can be regarded as superior filters. The exploration of deconvolution technology in fault diagnosis and condition monitoring can be traced back to the 1960s, where the traditional minimum entropy deconvolution (MED) is expressed by mathematical equations. Through practical applications, the present invention has a good effect in processing rolling bearing fault signals, and the method is more suitable for removing noise from signals in mechanical systems. Among them, the traditional minimum entropy deconvolution (MED) is expressed by mathematical equations. Due to the characteristics of kurtosis, MED is very sensitive to individual outliers. To address this problem, McDonald et al. developed the maximum correlation kurtosis deconvolution (MCKD), which makes full use of the periodic characteristics of fault-induced pulses. Due to the sliding phenomenon in rolling bearings, MCKD cannot be successfully applied when the theoretical fault frequency may be different from the actual fault frequency. Multi-point optimal minimum entropy deconvolution adjustment (MOMEDA) is proposed to separate the excitation source from the vibration signal. It is proposed to combine the average kurtosis (AK) with the BD method (MAKD) to solve the situation of time-varying speed. In recent years, the objective function of deconvolution has been extended from time domain indicators to square envelope spectrum (SES) indicators. Peeters et al. introduced traditional indicators to measure the sparsity of SES, such as l2 / l1 norm, spectral negentropy and Hoyer index. These indicators are formulated under the assumption that discrete peaks will appear in SES when a fault occurs. Therefore, as the fault progresses, SES tends to show greater sparsity. The Box-cox sparsity metric (BCSM) was developed to extend the l2 / l1 norm and spectral negentropy in a unified framework and has been further applied to the deconvolution method. However, this method is affected by the unreasonable setting of the initial filter. Therefore, the present invention designs a bandpass filter to replace its undefined initial filter through finite element simulation technology. Through practical application, the present invention has a good effect in extracting the fault characteristics of wind turbines. The method is more suitable for denoising and fault detection in mechanical systems, and provides technical support for monitoring the health status of wind turbines. Summary of the invention

[0005] In view of the above problems existing in the prior art, the present invention is proposed.

[0006] Therefore, the present invention provides a simulation-driven BCSM deconvolution fault diagnosis method. The technical problem to be solved by the method is to provide a wind turbine fault detection method. A wind turbine fault diagnosis method jointly driven by finite element simulation and BCSM deconvolution is designed. The method can utilize the damaged physical properties to realize fault detection of the wind turbine. The method is simple to operate, fast to run, and can realize online monitoring.

[0007] To solve the above technical problems, the present invention provides the following technical solution: a simulation-driven BCSM deconvolution fault diagnosis method, comprising: obtaining original vibration data from a wind turbine generator set using an acceleration sensor and a data acquisition instrument;

[0008] Based on the finite element simulation, a bandpass initial filter is designed to filter the original vibration data and serve as the initial filter for Box-Cox sparse measure deconvolution;

[0009] Taking BCSM as the objective function, the optimal filter is iterated based on the theory of generalized Rayleigh entropy to filter the original vibration data;

[0010] The designed bandpass filter is used as the initial filter of BCSM deconvolution, and the BCSM deconvolution method is performed to obtain the final filtered signal, and the envelope analysis of the filtered signal is performed to identify the fault characteristics;

[0011] The bandpass initial filter designed according to the finite element simulation is expressed as:

[0012]

[0013] Among them, ω p is the passing frequency, ω c is the cut-off frequency, is the ripple factor, P z is the z-order Chebyshev polynomial, jω represents the complex frequency, R z Expressed as the complex form of the filter, f z (ω) represents a digital z-order Chebyshev type I bandpass filter, f i Denoted as the initial filter;

[0014] According to the digital z-order Chebyshev I-type bandpass filter, find the lowest order z and input ω p The left boundary of w p1 and the right boundary w p2 It is expressed as:

[0015]

[0016] Among them, f s and f c denote the sampling frequency and the center frequency respectively, BW is the bandwidth of the digital z-order Chebyshev type I bandpass filter, and the center frequency is determined based on the finite element simulation.

[0017] As a preferred solution of the fault diagnosis method of simulation-driven BCSM deconvolution described in the present invention, since the pulse caused by the fault cannot be accurately identified due to interference such as the transmission path, random interference, and environmental noise when collecting the original vibration data of the wind turbine, a mathematical model of the original vibration data is constructed to clearly describe the components of the original vibration data. The original vibration data x is obtained as:

[0018] x=s m *p m +d

[0019] d=s h *p h +s r *p r +s n *p n

[0020] Among them, s m It represents the impact signal caused by mechanical failure, and d represents the influence of interference, which includes harmonic interference s h , random shocks m and ambient noise n , original vibration data x,p m 、p h 、p r 、p n They are m ,s h ,s r ,s n The transfer path function of , * is the convolution operator;

[0021] The deconvolution process is expressed as:

[0022] y=x*f i =(s m *p m +d)*f i ≈s m

[0023] Where y represents the deconvolution signal, f i Denoted as the initial filter;

[0024] The deconvolved signal is converted into a vector form and expressed as:

[0025] y=Xf i

[0026]

[0027] Where X is the vector form corresponding to the original vibration data, f iIt is represented as the vector form of the initial filter, y0 represents the first value of the deconvolution signal, x0 represents the first value of the original vibration data, and x L-1 Represented as the L-1th value of the original vibration data, y L-1 Represented as the L-1th value of the deconvolved signal, x N-1 Represented as the N-1th value of the original vibration data, x L+N-1 It is represented as the L+N-1th value of the original vibration data, f0 is represented as the first value of the initial filter, and f L-1 It is represented as the L-1th value of the initial filter, N is represented as the length of the initial filter, and L is represented as the length of the original vibration data;

[0028] The problem of solving the optimal filter is transformed into the maximization problem of the corresponding objective function through the theory of generalized Rayleigh entropy, and the deconvolution process is expressed as:

[0029]

[0030] Among them, O(y) represents the objective function of deconvolution, f o is represented as the optimal filter, Represents the maximization objective function.

[0031] As a preferred solution of the fault diagnosis method of a simulation-driven BCSM deconvolution described in the present invention, the Box-Cox sparse measure deconvolution includes: the square envelope better quantifies the sparse features, the objective function of the deconvolution is extended from the time domain index to the sparse measure of the square envelope spectrum, the concept of the square envelope is introduced, and the square envelope is expressed as:

[0032] ∈ x =|y| 2 =|Xf i | 2

[0033] Among them, ∈ x Expressed as a squared envelope;

[0034] The squared envelope spectrum is transformed into the following form:

[0035]

[0036] Among them, y H It is represented as the conjugate transposed matrix of the deconvolved signal, diag is represented as a diagonal matrix, and the vector form of the square envelope spectrum is E x It is expressed as:

[0037] E x =F H ∈ x =F H diag(yH )Xf i

[0038]

[0039] Where F is the Fourier matrix, n = 0, ..., LN-1; k = 0, ..., K-1; k is the index of the maximum frequency of interest, F H Expressed as the conjugate transpose of the Fourier matrix.

[0040] As a preferred solution of the fault diagnosis method of simulation-driven BCSM deconvolution described in the present invention, the iterative optimal filter includes calculating the sparse measure by Box-Cox transformation according to the kurtosis and entropy of the square envelope spectrum, which is expressed as:

[0041]

[0042] Among them, λ represents the transformation parameter of BCSM, and the vector form of the generalized Rayleigh quotient of BCSM is expressed as:

[0043]

[0044] Among them, f i H Expressed as the conjugate transpose of the initial filter;

[0045] The eigenvalue ω and the corresponding eigenvector are equivalent to f o The maximum value of the generalized Rayleigh entropy and the maximized eigenvalue are expressed as:

[0046]

[0047]

[0048] in, Expressed as a weight matrix of the Box-cox sparse measure, Expressed as the weight matrix of the original vibration data, X H It is expressed as the conjugate transpose of the vector form corresponding to the original vibration data.

[0049] As a preferred solution of the fault diagnosis method of simulation-driven BCSM deconvolution described in the present invention, the filtering of the original vibration data using the initial filter includes, according to the generalized Rayleigh quotient form of BCSM, maximizing the BCSM in the square envelope spectrum by iteration, and finding the maximum eigenvalue and the maximum eigenvector by the eigenvalue decomposition method:

[0050]

[0051] The original vibration data is filtered using the optimal filter to obtain the filtered signal y f , expressed as:

[0052] y f =y*f o .

[0053] As a preferred solution of the simulation-driven BCSM deconvolution fault diagnosis system described in the present invention, it includes: a data acquisition module, a filter design module, an objective function iteration module, a filter application module, and a fault identification module;

[0054] The data acquisition module includes acquiring original vibration data from the wind turbine generator set through an acceleration sensor and a signal acquisition instrument;

[0055] The filter design module includes a bandpass filter f0 designed according to finite element simulation to filter the original signal and serve as an initial filter for BCSM deconvolution;

[0056] The objective function iteration module includes, then taking BCSM as the objective function and iterating out the optimal filter based on the theory of generalized Rayleigh entropy, and then filtering the original vibration data;

[0057] The filter application module includes using the designed bandpass filter as a BCSM deconvolution initial filter and executing the BCSM deconvolution method to obtain a final filtered signal;

[0058] The fault identification module includes, finally, performing envelope analysis on the filtered signal to identify fault characteristics.

[0059] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of any one of the methods of fault diagnosis of a simulation-driven BCSM deconvolution are implemented.

[0060] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any one of the methods of a simulation-driven BCSM deconvolution fault diagnosis method are implemented.

[0061] Beneficial effects of the invention: The invention overcomes the inability of wind turbine generator sets to accurately identify faults of their components due to the complex working environment. It has strong robustness in practical applications and can accurately judge the operating status of wind turbine generator sets. It also has fast operating speed and simple operation, and is highly convenient in practical applications, and has certain industrial value. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0063] Figure 1 A specific flow chart of a wind turbine fault diagnosis method of a simulation-driven BCSM deconvolution fault diagnosis method provided by an embodiment of the present invention.

[0064] Figure 2 A test bearing outer ring fault signal of a simulation-driven BCSM deconvolution fault diagnosis method provided in an embodiment of the present invention: (a) time domain and (b) envelope spectrum.

[0065] Figure 3 A simulation-driven BCSM deconvolution fault diagnosis method is provided in an embodiment of the present invention to test the first bending mode vibration shape of the outer ring of a bearing.

[0066] Figure 4 (a) Time domain and (b) frequency domain response curves of an initial filter suitable for a bearing inner ring designed by finite element simulation for a simulation-driven BCSM deconvolution fault diagnosis method provided in an embodiment of the present invention.

[0067] Figure 5 The filter signal (b) envelope of the optimal filter after BCSM deconvolution iteration of a simulation-driven BCSM deconvolution fault diagnosis method provided by an embodiment of the present invention.

[0068] Figure 6 A test bearing inner race fault signal of a simulation-driven BCSM deconvolution fault diagnosis method provided in an embodiment of the present invention: (a) time domain (b) envelope.

[0069] Figure 7 A simulation-driven BCSM deconvolution fault diagnosis method is provided in an embodiment of the present invention to test the first bending frequency modal vibration type of the inner ring of a bearing.

[0070] Figure 8 (a) Time domain and (b) frequency domain response curves of an initial filter suitable for a bearing outer ring designed by finite element simulation for a simulation-driven BCSM deconvolution fault diagnosis method provided in an embodiment of the present invention.

[0071] Fig. 9 A simulation-driven BCSM deconvolution fault diagnosis method provided in an embodiment of the present invention is a filtered signal of an optimal filter after BCSM deconvolution iteration: (a) time domain and (b) envelope. DETAILED DESCRIPTION

[0072] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0073] Example 1, reference Figure 1 This is the first embodiment of the present invention, which provides a simulation-driven BCSM deconvolution fault diagnosis method, comprising:

[0074] First, the monitoring parts are simulated by finite element method, and their bending frequency is determined by modal analysis. Then, the corresponding filter is designed according to its bending frequency and used as the initial filter of BCSM deconvolution. Then, the monitoring signal is processed by optimized BCSM deconvolution to obtain the required filtered signal, and then the filtered signal is demodulated by envelope spectrum, so as to realize the fault diagnosis of key components of wind turbine. Please refer to Figure 1 What is provided is a flow chart of the wind turbine fault diagnosis method. The details of the specific algorithm are as follows:

[0075] Since the pulses caused by the fault cannot be accurately identified due to interferences such as the transmission path, random interference, and environmental noise when collecting the original vibration data of the wind turbine, a mathematical model of the original vibration data is constructed to clearly describe the components of the original vibration data. The acquisition of the original vibration data is expressed as:

[0076] x=s m *p m +d (1)

[0077] d=s h *p h +s r *p r +s n *p n (2)

[0078] Among them, s m It represents the pulse caused by mechanical failure, d represents the influence of interference, and the influence of interference includes harmonic interference s h , random shocks m and ambient noise n , original vibration data x,p m 、p h 、p r 、p n They are m,s h ,s r ,s n The purpose of deconvolution is to maximize the original vibration data x by continuously iterating the initial filter f. i To obtain the optimal filter f o , and its deconvolution process is expressed as:

[0079] y=x*f i =(s m *p m +d)*f i ≈s m (3)

[0080] Where y represents the deconvolution signal, f i Denote as the initial filter;. Since both the transfer function and the fault-induced pulses are unknown, the separation of the fault pulses from the measured signal is blind.

[0081] Equation (3) can be converted into vector form as follows:

[0082] y=Xf i (4)

[0083]

[0084] Among them, since the characteristic decomposition in the deconvolution process has direction, the original vibration data is a scalar without direction, so the scalar needs to be converted into a vector form, and the academic standard expression of the vector form is boldface, so the boldface in the present invention is the vector form corresponding to the scalar. X is the vector form corresponding to the original vibration data, f i It is represented as the vector form of the initial filter, y0 represents the first value of the deconvolution signal, x0 represents the first value of the original vibration data, and x L-1 Represented as the L-1th value of the original vibration data, y L-1 Represented as the L-1th value of the deconvolved signal, x N-1 Represented as the N-1th value of the original vibration data, x L+N-1 It is represented as the L+N-1th value of the original vibration data, f0 is represented as the first value of the initial filter, and f L-1 It represents the L-1th value of the initial filter, N represents the length of the initial filter, and L represents the length of the original vibration data.

[0085] The problem of solving the optimal filter can be transformed into a maximization problem with special criteria, so the deconvolution process can be summarized as:

[0086]

[0087] Among them, O(y) represents the objective function of deconvolution, f o represents the filter obtained after optimization, Represents the maximization objective function.

[0088] Since the setting of the initial filter will affect its iteration effect, a suitable initial filter is designed based on finite element simulation. The designed initial filter is expressed as:

[0089]

[0090] Among them, ω p is the passing frequency, ω c is the cut-off frequency, is the ripple factor, P z is the z-order Chebyshev polynomial, jω represents the complex frequency, R z Expressed as the complex form of the filter, f z (ω) represents a digital z-order Chebyshev type I bandpass filter;

[0091] Since the Chebyshev I-type filter has strong robustness, the fault uses a digital z-order Chebyshev I-type bandpass filter to find the lowest order z and input ω p The left boundary of w p1 and the right boundary w p2 It is expressed as,

[0092]

[0093] Among them, f s and f c denote the sampling frequency and the center frequency respectively, BW is the bandwidth of the digital z-order Chebyshev type I bandpass filter, and the center frequency is determined based on the finite element simulation.

[0094] At the same time, in order to better eliminate the interference of the transmission path and enhance the periodic characteristics, the objective function of deconvolution is expanded from the time domain index to the sparse measure of the square envelope spectrum. The expression of the square envelope can be defined as:

[0095] ∈ x =|y| 2 =|Xf i | 2 (10)

[0096] Among them, ∈ x Expressed as the squared envelope of the raw vibration data;

[0097] According to formula (10), the square envelope is converted into vector form:

[0098]

[0099] Among them, y H is represented as the conjugate transposed matrix of the deconvolved signal, and diag is represented as a diagonal matrix.

[0100] Perform Fourier transform on the square envelope to obtain the vector form of the square envelope spectrum E x It is expressed as:

[0101] E x =F H ∈ x =F H diag(y H )Xf i (12)

[0102] in,

[0103]

[0104] Where F is the Fourier matrix, n = 0, ..., LN-1; k = 0, ..., K-1; k is the index of the maximum frequency of interest, F H Expressed as the conjugate transpose of the Fourier matrix.

[0105] In order to better measure the sparsity of the square envelope spectrum, the sparse measure of the envelope spectrum is calculated with the help of Box-Cox transformation according to the kurtosis and entropy of the square envelope spectrum and expressed as:

[0106]

[0107] Where λ represents the transformation parameter of BCSM. The generalized Rayleigh quotient of BCSM is as follows:

[0108]

[0109] Among them, f i H Expressed as the conjugate transpose of the initial filter;

[0110] The maximum eigenvalue ω and its corresponding eigenvector are equivalent to the maximum value of the generalized Rayleigh quotient of the filter coefficient f. The expression for maximizing the eigenvalue is defined as:

[0111]

[0112] in, Expressed as a weight matrix of the Box-cox sparse measure, Expressed as the weight matrix of the original vibration data, X H It is expressed as the conjugate transpose of the vector form corresponding to the original vibration data.

[0113] According to formula (16), the BCSM in its envelope spectrum is maximized to obtain the corresponding optimal filter:

[0114]

[0115] The specific process of BCSM deconvolution is as follows:

[0116] 1. Input a faulty vibration signal.

[0117] 2. Construct an initial filter to filter the original signal and start iteration.

[0118] 3. Calculate using formulas (17)-(18) and

[0119] 4. Calculate formula (16) according to the eigenvalue decomposition formula to obtain the maximum characteristic ω max , and obtain the optimal filter f according to formula (19) o .

[0120] Filter designed by FEM simulation: BCSM deconvolution using the simulated driven filter as the initial filter. In order to obtain the center frequency of the initial filter, FEM simulation calculations are performed on the modal section. Based on the modal vibration shape, the first bending frequency can be regarded as the carrier frequency when the bearing fails. Specifically, the design process of the initial filter can be divided into two parts. In the first part, FEM analysis is performed by simulating a given bearing in order to determine the center frequency of the bandpass filter. In the second part, the initial filter can then be designed with a fixed filter length.

[0121] Based on the above description, the detailed flow chart of the present invention is as follows Figure 1 The detailed description is as follows:

[0122] Step 1: Acquire raw vibration data from the wind turbine.

[0123] Step 2: Create a finite element model for the bearing and perform modal analysis using the finite element method.

[0124] Step 3: Based on the mode shape, select the potential carrier frequency as the first bending frequency. Then use the obtained carrier frequency as the center frequency of the designed bandpass filter.

[0125] Step 4: Use the designed bandpass filter as the BCSM deconvolution initial filter and perform the BCSM deconvolution method to obtain the final filtered signal.

[0126] Step 5: Perform envelope analysis on the filtered signal to identify fault characteristics.

[0127] The above is a schematic scheme of a simulation-driven BCSM deconvolution fault diagnosis method of this embodiment. It should be noted that the technical scheme of the system of the simulation-driven BCSM deconvolution fault diagnosis method and the technical scheme of the above-mentioned simulation-driven BCSM deconvolution fault diagnosis method belong to the same concept, and the details of the technical scheme of the system of the simulation-driven BCSM deconvolution fault diagnosis method in this embodiment that are not described in detail can all be referred to the description of the technical scheme of the above-mentioned simulation-driven BCSM deconvolution fault diagnosis method.

[0128] In actual industrial applications, damage to key components of wind turbines cannot be warned in a timely manner, and common fault diagnosis methods are unable to diagnose the health status of key components of wind turbines under complex working conditions. The simulation-driven BCSM deconvolution fault diagnosis system developed in the present invention can effectively and real-time identify the health status of key components of wind turbines, thereby realizing real-time operation and maintenance of wind turbines.

[0129] In this embodiment, a simulation-driven BCSM deconvolution fault diagnosis system includes:

[0130] Data acquisition module, filter design module, objective function iteration module, filter application module, fault identification module;

[0131] The data acquisition module includes acquiring original vibration data from the wind turbine generator set;

[0132] The filter design module includes a bandpass filter f0 designed according to finite element simulation to filter the original signal and serve as an initial filter for BCSM deconvolution;

[0133] The objective function iteration module includes, then taking BCSM as the objective function and iterating out the optimal filter based on the theory of generalized Rayleigh entropy, and then filtering the original vibration data;

[0134] The filter application module includes using the designed bandpass filter as the BCSM deconvolution initial filter and executing the BCSM deconvolution method to obtain the final filtered signal;

[0135] The fault identification module includes, finally, performing envelope analysis on the filtered signal to identify fault characteristics.

[0136] This embodiment further provides a computing device, applicable to a simulation-driven BCSM deconvolution fault diagnosis method, including:

[0137] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0138] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0139] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0140] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0141] Example 2, reference Figure 2-Figure 5 , which is the second embodiment of the present invention, provides a simulation-driven BCSM deconvolution fault diagnosis method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0142] Take the fault signal of the outer ring of the real wind turbine bearing. The running speed of the outer ring fault bearing is 500rpm, which represents 8.34Hz (ie 500 / 60Hz), of which the outer ring fault is 43.15Hz (5.178×8.34). Figure 2 (a) and (b) are the original signal and its envelope spectrum, where Amplitude and Frequency are amplitude and frequency respectively. Figure 2 As shown in (b), the demodulated frequency is 50 Hz (approximate value of 6×) and its harmonics, representing the periodic interference component. The pulse caused by the outer ring fault is completely submerged by the periodic pulse and its harmonics caused by the shaft period. In order to avoid the periodic pulse interference caused by the shaft movement, BCSM deconvolution with a simulation design filter is used to eliminate the periodic interference. Through the initial step of the present invention, a modal analysis is performed to obtain the first bending frequency, that is, the potential carrier frequency. A finite element model is established according to the geometric dimensions and material parameters. It is worth mentioning that the contact surface between the outer ring and the roller can be described as 9 spring-damper models (corresponding to the 9 rollers in the bearing), in which Plane42 is selected as the outer ring plane unit and Combine14 is selected as the spring-damper unit. The damping ratio c and stiffness k of the 4 spring-damper units are 0.01 and 1224.6×106N / m, respectively. Modal analysis is performed to obtain the natural frequencies of the first seven orders. As shown Figure 3 As shown, the first bending mode vibration mode (its natural frequency is 3803Hz) can be found in the fourth-order mode shape, where Model solution is modal analysis, SUB is the natural frequency of the first order, FREQ is the natural frequency value, DMX is the maximum displacement, SMX and SMN are the maximum and minimum values ​​of the ITME solution respectively. Then 3803Hz is selected as the carrier frequency, as the center frequency of the bandpass filter, and a fixed bandwidth of 100Hz is added. Figure 4 (a) and (b) show the time waveform and frequency response curve of the filter designed by finite element simulation, respectively, where Amplitude and Frequency are amplitude and frequency, respectively. The designed filter is used as the initial filter of the BCSM deconvolution method. The transformation parameter of BCSM deconvolution (0≤λ≤1) is more effective than λ>1 in monitoring the health of bearings. Therefore, three transformation parameters (λ=0, 0.5, 1) are used to evaluate the performance of the proposed method. The results of the proposed method are shown in Figure 5 As shown, Amplitude and Frequency are amplitude and frequency respectively. The periodic characteristics of the outer ring fault can be clearly seen in the time domain waveforms of the three transformation parameter results, and the characteristic frequencies of the outer ring fault (43.1Hz, 85.9Hz and 171.9Hz) can be easily found in the corresponding envelope spectrum. In summary, this invention can effectively extract the fault characteristics of the bearing. And after analyzing the actual vibration signal, it can be proved that the invention has good robustness and accuracy and is simple and convenient to operate. Therefore, the invention provides a theoretical basis for solving the fault diagnosis of wind turbines, and at the same time has good results when combined with engineering applications in practice.

[0143] Example 3, reference Figure 6-Figure 9 , which is the third embodiment of the present invention, provides a simulation-driven BCSM deconvolution fault diagnosis method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0144] Take a real wind turbine bearing inner ring failure case, the test bearing running speed is 700rpm, the indicated shaft frequency is 11.6Hz (i.e. 700 / 60Hz). Therefore, the bearing inner ring failure frequency is calculated to be 90.7Hz (i.e. 7.822×11.6)a). Figure 6 (a) and (b) show the original signal and its envelope spectrum, where Amplitude and Frequency are amplitude and frequency respectively. Figure 6 (b) It can be seen that the demodulation frequencies 11.3Hz and 23.4Hz accurately match the rotation frequency (about 11.6Hz) and its harmonics (about 23.2Hz). The stiffness k and damping ratio c are selected as 0.05 and 6.02×108N / m respectively, and then the modal analysis is performed based on the above parameters. Figure 7is the first-order bending mode shape of the inner ring (corresponding to the fourth-order mode shape), where Modelsolution is the modal analysis, SUB is the order natural frequency, FREQ is the natural frequency value, DMX is the maximum displacement, SMX and SMN are the maximum and minimum values ​​of the ITME solution respectively. Therefore, the fourth-order natural frequency can be selected as the first-order bending frequency of 5700Hz. The center frequency of 5700Hz is used as the initial filter with a fixed bandwidth of 100Hz. The principle is the same as described in the previous section. However, the frequency value of BPFI=90.7Hz cannot be perceived in ES. Similar to the outer ring, the initial filter is designed in the FEM simulation. The time wave and frequency response curves of the designed filter are shown in Figure 8 As shown, Amplitude and Frequency are amplitude and frequency respectively. Fig. 9 (a) and (b) show the time waveform and ES of the FEM-BCSM deconvolution filtering result, respectively, where Amplitude and Frequency are the amplitude and frequency, respectively. Fig. 9 As shown in (a), the periodic interference component (rotating shaft frequency) is significantly mitigated. In addition, the demodulated frequencies 91 Hz, 181.4 Hz, and 271.8 Hz are easily matched with BPFI (90.7 Hz) and its harmonics (181.4 Hz and 272.1 Hz). The periodic pulses caused by the rotating shaft motion are completely eliminated, so the proposed method successfully identifies the presence of the inner race fault of the wind turbine bearing.

[0145] The implementation cases should not be regarded as limitations of the invention, but any improvements made based on the spirit of the invention should be within the scope of protection of the invention.

[0146] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A simulation-driven BCSM deconvolution fault diagnosis method, characterized in that: include, Using acceleration sensors and data acquisition instruments to obtain raw vibration data from wind turbines; Based on the finite element simulation, a bandpass initial filter is designed to filter the original vibration data and serve as the initial filter for Box-Cox sparse measure deconvolution; Taking BCSM as the objective function, the optimal filter is iterated based on the theory of generalized Rayleigh entropy to filter the original vibration data; The designed bandpass filter is used as the initial filter of BCSM deconvolution, and the BCSM deconvolution method is performed to obtain the final filtered signal, and the envelope analysis of the filtered signal is performed to identify the fault characteristics; The bandpass initial filter designed according to the finite element simulation is expressed as: Among them, ω p is the passing frequency, ω c is the cut-off frequency, is the ripple factor, P z is the z-order Chebyshev polynomial, jω represents the complex frequency, R z Expressed as the complex form of the filter, f z (ω) represents a digital z-order Chebyshev type I bandpass filter, f i Denoted as the initial filter; According to the digital z-order Chebyshev I-type bandpass filter, find the lowest order z and input ω p The left boundary of w p1 and the right boundary w p2 It is expressed as: Among them, f s and f c denote the sampling frequency and the center frequency respectively, BW is the bandwidth of the digital z-order Chebyshev type I bandpass filter, and the center frequency is determined based on the finite element simulation.

2. A simulation-driven BCSM deconvolution fault diagnosis method as claimed in claim 1, characterized in that: The acquisition of raw vibration data is expressed as: x=s m *p m +d d=s h *p h +s r *p r +s n *p n Among them, S m It represents the impact signal caused by mechanical failure, and d represents the influence of interference, which includes harmonic interference s h , random shocks m and ambient noise n , original vibration data x,p m 、p h 、p r 、p n They are m ,s h ,s r ,s n The transfer path function, * is the convolution operator, and x represents the original vibration data; The deconvolution process is expressed as: y=x*fi=(s m *p m +d)*fi≈s m Where y represents the deconvolution signal; The deconvolved signal is converted into a vector form and expressed as: y=Xf i Where X is the vector form corresponding to the original vibration data, f i Represented as the vector form of the initial filter, y0 represents the first value of the deconvolution signal, x0 represents the first value of the original vibration data, X L-1 Represented as the L-1th value of the original vibration data, y L-1 Represented as the L-1th value of the deconvolved signal, X N-1 Represented as the N-1th value of the original vibration data, X L+N-1 It is represented as the L+N-1th value of the original vibration data, f0 is represented as the first value of the initial filter, and f L-1 It represents the L-1th value of the initial filter, N represents the length of the initial filter, and L represents the length of the original vibration data; The problem of solving the optimal filter is transformed into the maximization problem of the corresponding objective function through the theory of generalized Rayleigh entropy, and the deconvolution process is expressed as: Among them, O(y) represents the objective function of deconvolution, f o is represented as the optimal filter, Represents the maximization objective function.

3. A simulation-driven BCSM deconvolution fault diagnosis method as claimed in claim 2, characterized in that: The Box-Cox sparse measure deconvolution includes expanding the deconvolution objective function from a time domain indicator to a sparse measure of the square envelope spectrum, and introducing the concept of square envelope. The square envelope is expressed as: ∈ x =|y| 2 =|Xf i | 2 Among them, ∈ x Expressed as a squared envelope; The squared envelope spectrum is transformed into the following form: Among them, y H It is represented as the conjugate transposed matrix of the deconvolved signal, diag is represented as a diagonal matrix, and the vector form of the square envelope spectrum is E x It is expressed as: E x =F H ∈ x =F H diag(y H )Xf i where F is the Fourier matrix, n = 0, ..., LN-1; k = 0, ..., K-1; k is the index of the maximum frequency of interest, F H Expressed as the conjugate transpose of the Fourier matrix.

4. A fault diagnosis method of simulation-driven BCSM deconvolution as claimed in claim 3, characterized in that: The iterative optimal filter includes calculating the sparse measure based on the kurtosis and entropy of the square envelope spectrum by means of Box-Cox transformation, which is expressed as: Among them, λ represents the transformation parameter of BCSM, and the vector form of the generalized Rayleigh quotient of BCSM is expressed as: Among them, f i H Expressed as the conjugate transpose of the initial filter; The eigenvalue ω and the corresponding eigenvector are equivalent to the maximum value of the generalized Rayleigh entropy of fo. The maximized eigenvalue is expressed as: in, Expressed as a weight matrix of the Box-cox sparse measure, Expressed as the weight matrix of the original vibration data, X H It is expressed as the conjugate transpose of the vector form corresponding to the original vibration data. According to the generalized Rayleigh entropy of BCSM, the BCSM in the square envelope spectrum is maximized through iteration, and the maximum eigenvalue and maximum eigenvector are obtained by the eigenvalue decomposition method: Among them, the maximum eigenvector corresponds to the vector form of the optimal filter f o .

5. A simulation-driven BCSM deconvolution fault diagnosis method as claimed in claim 4, characterized in that: The obtaining of the final filtered signal includes using the optimal filter to filter the original vibration data to obtain the filtered signal y f , expressed as: y f =y*f o 。 6. A system based on a simulation-driven BCSM deconvolution fault diagnosis method according to any one of claims 1 to 5, characterized in that: Including data acquisition module, filter design module, objective function iteration module, filter application module, fault identification module; The data acquisition module includes obtaining original vibration data from the wind turbine generator set using an acceleration sensor and a signal acquisition instrument; The filter design module includes a bandpass filter designed according to finite element simulation to filter the original signal and serve as an initial filter for BCSM deconvolution; The objective function iteration module includes taking BCSM as the objective function and iterating the optimal filter based on the theory of generalized Rayleigh entropy to filter the original vibration data; The filter application module includes using the designed bandpass filter as the BCSM deconvolution initial filter, and executing the BCSM deconvolution method to iteratively update the filter to obtain the final filter; The fault identification module includes performing envelope analysis on the filtered signal to identify fault characteristics.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of a simulation-driven BCSM deconvolution fault diagnosis method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a simulation-driven BCSM deconvolution fault diagnosis method according to any one of claims 1 to 5 are implemented.

Images