An adaptive signal processing method for a low-quality bearing fault signal in an electric drive system, and a module and a fault diagnosis system employing

Abstract

A method of adaptive signal processing is provided for a low-quality bearing fault signal in an electric drive system, the method comprising importing a low-quality bearing fault signal, and construct

Description

[0005] In order to achieve effective and intuitive fault diagnosis for a electric drive system, the present invention proposes an adaptive signal processing method for a low-quality bearing fault signal in an electric drive system. The method is applied to a signal processing and visualization analysis module, comprises the following steps:

[0006] Step 1: Importing a low-quality bearing fault signal.

[0007] Step 2: Constructing multiple parallel filters with random center frequencies and bandwidths smaller than or equal to one-eighth of a sampling rate.

[0008] Step 3: Utilizing multiple parallel filters to initially filtering the low-quality bearing fault signal at a random center frequency to obtain an initial filtered signal.

[0009] Step 4: Calculating an envelope signal of the initial filtered signal, represented as:

[0010] \h[n] = xlh[n\ + j• Hilbertjx, Jn]}

[0011] Wherein: xt h [n] represents the initial filtered signal within a frequency band [ / , h], xt Afn] represents the envelope signal of the initial filtered signal.

[0012] To obtain a squared envelope spectrum of the initial filtered signal, Fourier transform is performed on the envelope signal, expressed as:

[0013] SES,>| = / J<q|x. Jw||2|

[0014] Step 5: Performing adaptive signal processing based on a Gini index on the initial filtered signal to obtain a fault frequency band. To update the filter center frequency, the particle swarm optimization algorithm is employed, wherein the Gini index of the squared envelope spectrum within frequency band serves as a fitness function.

[0015] The particle swarm optimisation algorithm updating the Gini index x^ based on the principles of global optimum and local optimum, is represented as:

[0017] Wherein: k represents an iteration number; x1^ and represent the Gini index and velocity at A' th iteration; q and c2 represent the acceleration factor; and r2 represent two independent random numbers within a range [0,1]; p^ , and p^ g represent a local best value and a global best value at the A' th iteration respectively; represents an inertia weight.

[0018] The Gini index of the squared envelope spectrum is represented as:

[0019] , x SEr,An\ x , =1-22---

[0020] Wherein: N represents a length of the original signal; sorts in an ascending order to become and represents a LI norm of SEj h[n], expressed as:

[0021] ^^]¾^^^

[0021] Adding 1 to a number of iterations for those less than a maximum number of iterations, and return to step 3.

[0023] Obtaining a frequency band with a highest Gini index in the last generation of the Gini index as a fault frequency band at the Gini index iteration count reaching the maximum particle iteration count.

[0024] Step 6: Filtering the initial low-quality bearing fault signal at the fault frequency band to obtain a fault frequency band signal.

[0025] The present invention further discloses a signal processing and visualization analysis module employing the aforementioned adaptive signal processing method for the low-quality bearing fault signal in the electric drive system, the module comprises: a signal processing program, and a GUI visualization interface.

[0026] The signal processing program implements the adaptive signal processing method via a computer program to obtain a fault frequency band signal;

[0027] The GUI visualization interface is employed to visually display the time domain, frequency domain, and envelope spectrum of the fault frequency band signal.

[0028] The bearing fault diagnosis system comprises the signal processing and visualization analysis module.

[0029] The bearing fault diagnosis system comprises: a signal acquisition module, a signal storage and transport module, a signal processing and visualization analysis module, a fault discrimination module.

[0030] The signal acquisition module employs an integrated low-sensitivity sensor to collect the low-quality signal from a rolling bearing with a sampling rate greater than twice the analysis frequency.

[0031] The signal storage and transport module includes a set of data transmission lines and a data storage unit connected with a computer. The low-quality bearing fault signal acquired by the signal acquisition module is inputted to the signal storage and transport module. To transfer the low-quality signal to the storage unit connected to the computer, a set of data transmission lines is employed. The memory of the data storage unit should be larger than the memory' occupied by the obtained low-quality signal.

[0032] The signal processing and visualization analysis module comprises a signal processing program and a GUI visualization interface. The low-quality' bearing fault signal from the signal storage and transport module is inputted to the signal processing and visualization analysis module. The signal processing program applies adaptive filtering to the low-quality signal to obtain the fault frequency band signal. The GUI interface visually displays fault impact features of the fault frequency band signal.

[0033] The fault discrimination module comprises a fault identification program and an envelope analysis program. The fault frequency band signal from the signal processing and visualization analy sis module is inputted to the fault discrimination module. The fault identification program conducts a qualitative assessment of the existence of fault through an evaluation index. The evaluation index employed is kurtosis.

[0034] The bearing fault type is determined through an envelope analysis program, compares a theoretical fault frequency with the peak value of the envelope spectrum of the fault frequency band signal to determine a bearing fault type.

[0035] Beneficial effects: Due to the adoption of the above technical solution, the present invention has the following advantages:

[0036] The present invention proposes a fault diagnosis method for low-quality signals in an electric drive system. By utilizing the particle swarm algorithm to adaptively select the frequency band with the most prominent fault characteristics, it significantly reduces the impact of sensor and environmental noise on fault detection, enables more sensitive identification of subtle fault frequency bands, and exhibits strong robustness and stability.

[0037] The signal processing and visual analysis module of the present invention comprises two parts: a signal handler and a GUI visual interface. Compared with the original method of writing program code, it has better readability and interpretability, making it easier for users to judge the type of fault, with higher efficiency and better stability.

[0038] The invention further discloses a low-quality signal bearing fault state diagnosis system for an electric drive system, which has good stability and robustness, can effectively reduce the dependence of the electric drive system on signal quality, and helps to improve the reliability and production efficiency of equipment operation. It is great significance to expand the diagnosis application scenarios of electric drive systems and reduce monitoring costs. Brief Description of the Drawings

[0039] Figure 1 is a flow chart of adaptive processing method for a low-quality bearing fault signal in an electric drive system.

[0040] Figure 2 is a flow chart of an bearing fault diagnosis system for a low-quality' signal in an electric drive system.

[0041] Figure 3 is a GUI interface of the time domain diagram after filtering an innerring low-quality fault signal.

[0042] Figure 4 is a GUI interface of the time domain diagram after filtering an outerring low-quality fault signal.

[0043] Figure 5 is a GUI interface for an inner ring fault diagnosis result.

[0044] Figure 6 is a GUI interface for an outer ring fault diagnosis result. Embodiments

[0045] In order to make the purpose, features, and implementation of the present invention more clear and understandable, the present invention provides a brief introduction for the specific implementation case. It should be understood that the specific implementation case described are only used to explain the present invention and are not intended to limit the present invention.

[0046] The present invention proposes an adaptive signal processing method for a low-quality bearing fault signal in an electric drive system. Figure 1 is a flow chart of the low-quality signal adaptive processing method. The method is applied to a signal processing and visualization analysis module, and comprises the following steps:

[0047] Step 1: Importing a low-quality bearing signal.

[0048] Step 2: Constructing multiple parallel filters with random center frequencies and bandwidths smaller than or equal to one-eighth of a sampling rate. As an embodiment, constructing 10 parallel filters with random center frequencies and bandwidths smaller than or equal to one-eighth of a sampling rate.

[0049] Step 3: Utilizing multiple parallel filters to initially filtering the low-quality bearing fault signal at the random center frequencies, obtaining an initial filtered signal.

[0050] Step 4: Calculating an envelope signal of the initial filtered signal, represented as:

[0051] = M"]+J’KlbertK^

[0052] wherein: xt h\n\ representing the initial filtered signal within a frequency band [ / , h ], Xj h[n} representing the envelope signal of the initial filtered signal.

[0053] To obtain a squared envelope spectrum of the initial filtered signal, Fourier transform is performed on tlie envelope signal, expressed as:

[0054] SES^a] = FFT[\x,>||2|

[0055] Step 5: Performing adaptive signal processing based on a Gini index on the initial filtered signal to obtain a fault frequency band. To update the filter center frequency, the particle swarm optimization algorithm is employed, wherein the Gini index of the squared envelope spectrum within frequency band serves as a fitness function.

[0056] The particle swarm optimisation algorithm updating the Gini index x1^ based on the principles of global optimum and local optimum, represented as:

[0057] I= ® V‘ h + ^'h + ^2 ~ ) V+i=xt +vw (^l.h xl.h

[0058] Wherein: k represents an iteration number; and represent the Gini index and velocity at Hh iteration; and c2 represent the acceleration factor; and r2 represent two independent random numbers within a range [0,1]; p^ , and p^ represent a local best value and global best value at the £th iteration respectively; ® represents an inertia weight.

[0059] The Gini index of the squared envelope spectrum represented as:

[0060] , SE^An] ^ = 1-22---

[0061] Wherein: N represents a length of the original signal; sorts SElfl [n] in an ascending order to become SE^n], and represents a LI norm of SEt h\n\, expressed as:

[0062]

[0063] Adding 1 to a number of iterations for those less than 10, and return to step 3.

[0064] Obtaining a frequency band with a highest Gini index in the tenth generation of the Gini index as a fault frequency band at the Gini index iteration count reaching the maximum particle iteration count.

[0065] Step 6: Filtering the low-quality bearing fault signal at the fault frequency band to obtain a fault frequency band signal.

[0066] The present invention further discloses a signal processing and visualization analysis module employing the aforementioned adaptive signal processing method for the low-quality bearing fault signal in the electric drive system, the module comprises: a signal processing program, and a GUI visualization interface.

[0067] The signal processing program implements the adaptive signal processing method via a computer program to obtain a fault frequency band signal.

[0068] The GUI visualization interface is employed to visually display the time domain, frequency domain, and envelope spectrum of the fault frequency band signal.

[0069] Figure 2 is a flow chart of a bearing fault diagnosis system for a low-quality signal in an electric drive system.

[0070] The bearing fault diagnosis system comprises the signal processing and visualization analysis module.

[0071] The bearing fault diagnosis system comprises: a signal acquisition module 201, a signal storage and transport module 202, a signal processing and visualization analysis module 203, a fault discrimination module 204. The sampling rate is 20K and the sampling time is 1 second.

[0072] The signal acquisition module 201 employs an integrated low-sensitivity sensor to collect die low-quality signal from a rolling bearing with a sampling rate greater than twice the analysis frequency.

[0073] The signal storage and transport module 202 includes a set of data transmission lines and a data storage unit connected with a computer. The low -quality bearing fault signal acquired by die signal acquisition module 201 is inputted to the signal storage and transport module 202. To transfer the low-quality signal to the storage unit connected to the computer, a set of data transmission lines is employed. The memory of the data storage unit should be larger than the memory occupied by the obtained low-quality signal.

[0074] The signal processing and visualization analysis module 203 comprises a signal processing program and a GUI visualization interface. The low-quality bearing fault signal from the signal storage and transport module 202 is inputted to the signal processing and visualization analysis module 203. The signal processing program applies adaptive fdtering to the low-quality signal to obtain the fault frequency band signal. The GUI interface visually displays fault impact features of the fault frequency band signal.

[0075] The fault discrimination module 204 comprises a fault identification program and an envelope analysis program. The fault frequency band signal from the signal processing and visualization analysis module 203 is inputted to the fault discrimination module 204. The fault identification program conducts a qualitative assessment of the existence of fault through an evaluation index. The evaluation index employed is kurtosis, expressed as:

[0076] kurtosis(x) = --3 EhW")-^2]2

[0077] Wherein: represents the mean value of the signal sequence *; if the kurtosis index is greater than 3, it is considered a faulty bearing, and if the kurtosis index is less than or equal to 3, it is considered a normal bearing. In Figure 3, the kurtosis is 4.5787, indicating a faulty bearing. In Figure 4, the kurtosis of is 3.7812 , indicating a faulty bearing.

[0078] The bearing fault type is determined by envelope analysis program, comparing the theoretical fault frequency with the peak value of the envelope spectrum of the filtered signal to determine the bearing fault type. A theoretical fault frequency of an inner ring of the rolling bearing expressed as:

[0079] / 0 = 0.5z / ^1+-^-cos

[0080] A theoretical fault frequency of an outer ring of the rolling bearing express as:

[0081] / 0 = 0.5z / ^l—^cosa

[0082] Wherein: d represents the diameter of the rolling elements, D represents the pitch diameter of the bearing, a represents the contact angle of the bearing, z represents the number of rolling elements, f represents the rotational frequency.

[0083] In the embodiment, the operating speed is 3000rpm. The fault frequency of the inner ring is 246.94Hz, the fault frequency of the outer ring is 153.06Hz. In Figure 5 and Figure 6, the horizontal axis represents frequency, the vertical axis represents tlie amplitude of the frequency. At point a, the frequency is 247Hz, and the amplitude is 2.2, which is close to the theoretical fault frequency for the inner ring, hence Figure 5 is judged to have an inner ring fault. At point b, the frequency is 153Hz, and the amplitude is 1.9, which is close to the theoretical fault frequency for the outer ring, hence judged as an outer ring fault in Figure 6.

[0084] The above embodiment is explanations of specific implementation methods of the present invention, rather than limitations of the present invention. Technical personnel in the relevant technical field may make various changes and variations within the spirit and scope of the present invention. Therefore, all equivalent technical solutions should be included within the scope of protection of the patent for the present invention.

Claims

1. A method of adaptive signal processing for a low-quality bearing fault signal in an electric drive system, characterized in that the method comprises the following steps:step 1: importing a low-quality bearing fault signal;step 2: constructing multiple parallel filters with random center frequencies and bandwidths smaller than or equal to one-eighth of a sampling rate;step 3: utilizing multiple parallel filters to initially filtering the low-quality bearing fault signal at a random center frequency to obtain an initial filtered signal;step 4: calculating an envelope signal of the initial filtered signal, represented as:= xlh [n] + j • Hilbert{xz Jm]}wherein: xZA[n] representing the initial filtered signal within a frequency band [ / , / ?], x, J / z] representing the envelope signal of the initial filtered signal;performing Fourier transform on the envelope signal to obtain a squared envelope spectrum of the initial filtered signal, expressed as:SESz>] = FFr[|xz^step 5: performing adaptive signal processing based on a Gini index on the initial filtered signal to obtain a fault frequency band; employing the particle swarm optimisation algorithm to update the filter centre frequency, wherein the Gini index of the squared envelope spectrum within frequency band serving as a fitness function;the particle swarm optimisation algorithm updating the Gini index x^ based on the principles of global optimum and local optimum, represented as:< V1M = CJ Vl.h + (PL, ~ Xl,h ) + Ct' ~ Xh )fc+1 _ k . k+1XLh Xl.h + Vl,hwherein: k representing an iteration number; x*h and representing the Gini index and velocity at k th iteration; c, and c2 representing the acceleration factor; and r2 representing two independent random numbers within a range [0,1]; p^z and p*Ag representing a local best value and a global best value at the k th iteration respectively; © representing an inertia weight;the Gini index of the squared envelope spectrum being represented as:k 7wherein: N representing a length of the original signal; sorting SElh\n\ in an ascendingorder to become 5E^a[m], and □ SElh[n} representing a LI norm of SEl h\n\, expressed as:adding 1 to a number of iterations for those less than a maximum number of iterations, and return to step 3;obtaining a frequency band with a highest Gini index in the last generation of the Gini index as a fault frequency band at the Gini index iteration count reaching the maximum particle iteration count;step 6: filtering the initial low-quality bearing fault signal at the fault frequency band to obtain a fault frequency band signal.

2. A signal processing and visualization analysis module employing the method of adaptive signal processing for low-quality bearing fault signal in the electric drive system of claim 1, characterized in that the module comprises:a signal processing program implementing the method described in claim 1 via a computer program to obtain a fault frequency band signal;a GUI visualization interface employed to visually display the time domain, frequency domain, and envelope spectrum of the fault frequency band signal.

3. A bearing fault diagnosis system for the low-quality bearing fault signal in the electric drive system of claim 2, characterized in that the system comprises the signal processing and visualization analysis module of claim 2 and further comprises:a signal acquisition module, a signal storage and transport module, a signal processing and visualization analysis module, a fault discrimination module;the signal acquisition module employs an integrated low-sensitivity sensor to collect the low-quality signal from a rolling bearing with a sampling rate greater than twice the analysis frequency;the signal storage and transport module includes a set of data transmission lines and a data storage unit connected to a computer; inputting the low-quality bearing fault signal acquired by the signal acquisition module to the signal storage and transport module; employing a set of data transmission lines to transfer the low-quality signal to the storage unit connected with the computer; the memory of the data storage unit should be larger than the memory occupied by the obtained low-quality signal;the signal processing and visualization analysis module comprises a signal processing program and a GUI visualization interface; inputting the low-quality bearing fault signal from the signal storage and transport module to the signal processing and visualization analysis module; the signal processing program applying adaptive filtering to the low-quality signal, and obtain the fault frequency band signal; the GUI interface visually displaying fault impact features of the fault frequency band signal;the fault discrimination module comprises a fault identification program and an envelope analysis program; inputting the fault frequency band signal from the signal processing and visualization analysis module to the fault discrimination module; the fault identification program conducts a qualitative assessment of the existence of fault; the bearing fault type is determined by means of an envelope analysis program.

Images