CN115096581B - Complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics - Google Patents
Complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics Download PDFInfo
- Publication number
- CN115096581B CN115096581B CN202210719789.0A CN202210719789A CN115096581B CN 115096581 B CN115096581 B CN 115096581B CN 202210719789 A CN202210719789 A CN 202210719789A CN 115096581 B CN115096581 B CN 115096581B
- Authority
- CN
- China
- Prior art keywords
- frequency
- data
- samples
- transmission device
- tracing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/021—Gearings
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/028—Acoustic or vibration analysis
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention provides a complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics, comprising the following steps: step S1: acquiring sample data; step S2: preprocessing a training sample and a test sample; step S3: constructing a minimum hypersphere model for fault diagnosis; step S4: and performing anomaly tracing by using a recursive feature elimination method. The invention provides a complex transmission device fault diagnosis tracing method based on time-frequency domain characteristics, which solves the transmission device fault diagnosis problem of small data samples and a large amount of fault data deficiency based on the ideas of support vector data description and characteristic recursion elimination. The invention uses time domain index to represent the whole information of the transmission device, uses frequency domain index to represent the shaft and gear information, constructs the minimum hypersphere description and the whole health state of the characterization model machine, judges the fault state, eliminates the recursion characteristic, and searches the cause of the fault from the input characteristic layer.
Description
Technical Field
The invention relates to the technical field of device fault diagnosis, in particular to a fault diagnosis method and system of a transmission device, and more particularly relates to a fault diagnosis tracing method and system of a complex transmission device based on time-frequency domain characteristics.
Background
The vehicle transmission device is a complex electromechanical-electro-hydraulic compound gear shifting speed change system, can integrally realize important functions of steering, gear shifting, speed change and the like of a complex vehicle, and is a core power component of a modern complex vehicle. The complex transmission device is characterized by complex structure, is used under various working conditions of rotating speed, gear and load, and internally comprises a large number of transmission elements such as gears, shafts, bearings, friction plates and the like. Under severe working conditions, mechanical parts are easy to degrade, so that faults are caused, and the operation safety of the transmission device and even vehicles is threatened. Therefore, the method has great significance and value for fault diagnosis of mechanical parts in the complex transmission device.
The vibration signal can reflect rich information of the operating state of the rotating machine, thus generating a rotating machine state signature based on the vibration signal. Chinese patent document No. CN112161807a publication No. 20210101 discloses a fault diagnosis method, apparatus and storage medium for a speed change gear box, in which a certain characteristic frequency band in a vibration signal is extracted by an adaptive filter, and 18 time domain indexes including root mean square and kurtosis and 3 frequency domain indexes including frequency center are extracted as detection features. In order to facilitate the positioning and replacement of the faulty component by a service person, the fault diagnosis also needs to be positioned to a specific faulty component. The publication No. CN114252261A discloses 20220329, which discloses a fault diagnosis method and system for a steering system of a comprehensive transmission device, wherein voltage signals of different positions in the steering system of the comprehensive transmission device under different working conditions are obtained, a hidden Markov model is trained by utilizing a data set under different working conditions, the trained hidden Markov model is optimized by using a particle swarm algorithm, finally, the voltage signals to be identified are identified based on the optimized hidden Markov model, and the fault positions are determined according to the positions corresponding to the voltage signals to be identified.
Numerous studies have shown that when a rotating machine component fails, the vibration signal excited by its periodic motion can convey rich information that can be accurately represented in time and frequency domain features. The components such as gears, shafts and bearings in the transmission device can generate targeted time domain and frequency domain characteristics during faults, so that the faults are traced and positioned by means of the characteristics generated by each component of the transmission device, but a related complex transmission device fault diagnosis tracing method based on the time domain and frequency domain characteristics is lacking at present.
Existing methods rely heavily on large data, requiring large amounts of data to be collected for training of classification models for different fault types, but in practical application scenarios, acquisition of fault data is extremely difficult. The transmission device has various fault modes of typical elements such as bearings, gears, brackets, shafts and the like, and the faults are various, but the actual conditions limit the acquisition of data of multiple faults and compound faults, so that a classification model based on different types of fault data cannot be constructed, and further fault tracing cannot be performed.
Patent document CN111780971B (application number: CN 202010523440.0) discloses a system and a method for diagnosing faults of a multi-shaft transmission device based on a rotation speed sensor, wherein the system comprises a plurality of rotation speed sensors, a wireless communication module and an edge calculation gateway, the plurality of rotation speed sensors are respectively arranged on the multi-shaft transmission system, the rotation speed sensors are used for collecting rotation speed signals, the plurality of rotation speed sensors are connected with the edge calculation gateway through the wireless communication module, the edge calculation gateway comprises a signal receiving module, a data analysis module and a fault diagnosis module, and the method comprises the following steps: step S1: the rotating speed sensor acquires a rotating speed signal; step S2: performing time domain analysis and frequency domain analysis on the rotating speed signal; step S3: setting a rotating speed threshold range, and comparing the rotating speed threshold range with the threshold range according to an analysis result. But the invention does not describe and characterize the overall health status of the prototype by constructing a minimum hypersphere.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a fault diagnosis and tracing method and system for a complex transmission device based on time-frequency domain characteristics.
The invention provides a complex transmission device fault diagnosis tracing method based on time-frequency domain characteristics, which comprises the following steps:
step S1: collecting vibration signals to obtain sample data;
step S2: training samples according to the acquired signals, and preprocessing test samples;
step S3: constructing a minimum hypersphere model, and inputting a test sample for fault diagnosis;
step S4: and performing anomaly tracing by using a recursive feature elimination method to perform transmission fault diagnosis.
Preferably, in said step S1:
testing at a specified rotating speed, installing and arranging a vibration sensor on a transmission device for measuring vibration signals and converting the vibration signals into electric signals for input, and acquiring signals s of a vehicle-mounted CAN bus and a three-way acceleration vibration sensor through multichannel data i The acquisition device converts the vibration electric signals into digital signals and transmits the digital signals to the computer.
Preferably, in said step S2:
calculating the signal s i Time and frequency domain features of (a): the time domain features include: root mean square value Z rms Kurtosis Z kurt And margin factor Z e The method comprises the steps of carrying out a first treatment on the surface of the The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C ；
Step S2.1: normalizing the training sample and the test sample, and performing noise reduction on the data;
step S2.2: calculating time domain features, root mean square values of signalsKurtosis-> Margin factor->
Wherein the discrete signal sequence s= [ s ] 1 ，s 2 ，...，s N ]N is the number of sampling points, s k For the data of the k-th one,as the mean value of the vibration signal, Z pp Is the signal peak;
step S2.3: calculating frequency domain index of signal, constructing arithmetic mean value X of frequency conversion and high-order harmonic energy of frequency conversion for each axis s ：
Wherein E is so Energy representing the o-th order of the conversion, o=1, 2, … H s ，H s Representing the highest order of the frequency conversion;
constructing a geometric mean C of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
For each gear pair, an arithmetic mean value X of the meshing frequency and the higher harmonic energy of the meshing frequency is defined g ：
Wherein E is gp Energy representing the p-th order of the meshing frequency, p=1, 2, … H g ，H g Representing the highest order of the meshing frequency;
for each gear pair, defining geometric mean C of meshing frequency and higher harmonic energy of the meshing frequency g ：
For each gear pair, an arithmetic mean M of the modulation intensity of each order of meshing frequency is defined X And geometric mean M C ：
Wherein E is bqp Represents the energy of the q-th side band of the p-th order meshing frequency, q=1, 2, … H b ，H b Indicating the number of side bands.
Preferably, in said step S3:
step S3.1: training data
Wherein the training setOne data x of (2) i N is the number of samples, m is the feature dimension;
first by a nonlinear transformation function Φ: the data is mapped from an original space to a feature space by x-F, F is a mapped high-dimensional feature space, and then a hypersphere with the minimum volume is constructed in the feature space, so that the following optimization problem is solved:
s.t.||Φ(x i )-C|| 2 ≤R 2 +ξ i ，ξ i ≥0
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error;
step S3.2: using the lagrangian multiplier method, solve the dual problem of the above problem:
wherein a is i For sample x i Corresponding Lagrangian coefficient, a j For sample x j Corresponding Lagrangian coefficients, K (x i ，x j ) Is a kernel function, equal to the inner product of the samples in feature space, K (x i ，x j )＝<Φ(x i )，Φ(x j )>By K ij Represents K (x) i ，x j ) The method comprises the steps of carrying out a first treatment on the surface of the Obtaining Lagrange coefficients corresponding to all samples by solving the dual problem, wherein the Lagrange coefficients are more than or equal to 0 and less than or equal to a in all training samples i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
in the method, in the process of the invention,is a support vector sample in the data, K vv ＝K(x v ，x v )，K vi ＝K(x v ，x i )；
Step S3.3: and (3) fault judgment:
the signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When in dataWhen only a normal sample exists, the distance d from the normal sample to the sphere center of the super sphere is calculated as follows:
wherein K is tt ＝K(x t ，x t )，K ti ＝K(x t ，x i )；
If d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to the normal sample, otherwise, the test sample belongs to the abnormal sample, and the radius R of the hypersphere is the health threshold under the given working condition.
Preferably, in said step S4:
performing anomaly tracing by using a recursive feature elimination method, wherein the number of training samples is n,
using all support vectors x v Square R of the resulting radius R 2 (x v ) As a criterion for measuring the magnitude of the boundary change, the average value J r Is defined as
J r ＝∑R 2 (x v )/t
Where t is the number of support vectors, and after introducing a linear kernel function, the criterion function can be expressed as:
let J r (-F) is the radius of the hypersphere obtained by removing features other than F, the feature that contributes most to anomalies is such that J r Features corresponding to maximum values of (-F) and, after removal of the feature F, the criterion function is applied by DJ r (F)＝J r -J r (-F) represents the feature F having the highest degree of contribution to abnormality * Is such that DJ r (F) Having features corresponding to the minimum, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) final fault positioning and tracing are carried out.
The invention provides a complex transmission device fault diagnosis tracing system based on time-frequency domain characteristics, which comprises the following components:
module M1: collecting vibration signals to obtain sample data;
module M2: training samples according to the acquired signals, and preprocessing test samples;
module M3: constructing a minimum hypersphere model, and inputting a test sample for fault diagnosis;
module M4: and performing anomaly tracing by using a recursive feature elimination method to perform transmission fault diagnosis.
Preferably, in said module M1:
testing at a specified rotating speed, installing and arranging a vibration sensor on a transmission device for measuring vibration signals and converting the vibration signals into electric signals for input, and acquiring signals s of a vehicle-mounted CAN bus and a three-way acceleration vibration sensor through multichannel data i The acquisition device converts the vibration electric signals into digital signals and transmits the digital signals to the computer.
Preferably, in said module M2:
calculating the signal s i Time and frequency domain features of (a): the time domain features include: root mean square value Z rm s 、 Kurtosis Z kurt And margin factor Z e The method comprises the steps of carrying out a first treatment on the surface of the The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C ；
Module M2.1: normalizing the training sample and the test sample, and performing noise reduction on the data;
module M2.2: calculating time domain features, root mean square values of signalsKurtosis-> Margin factor->
Wherein the discrete signal sequence s= [ s ] 1 ，s 2 ，...，s N ]N is the number of sampling points, s k For the data of the k-th one,as the mean value of the vibration signal, Z pp Is the signal peak;
module M2.3: calculating frequency domain index of signal, constructing arithmetic mean value X of frequency conversion and high-order harmonic energy of frequency conversion for each axis s ：
Wherein E is so Energy representing the o-th order of the conversion, o=1, 2, … H s ，H s Representing the highest order of the frequency conversion;
constructing a geometric mean C of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
For each gear pair, an arithmetic mean value X of the meshing frequency and the higher harmonic energy of the meshing frequency is defined g ：
Wherein E is gp Energy representing the p-th order of the meshing frequency, p=1, 2, … H g ，H g Representing the highest order of the meshing frequency;
for each gear pair, defining geometric mean C of meshing frequency and higher harmonic energy of the meshing frequency g ：
For each gear pair, an arithmetic mean M of the modulation intensity of each order of meshing frequency is defined X And geometric mean M C ：
Wherein E is bqp Represents the energy of the q-th side band of the p-th order meshing frequency, q=1, 2, … H b ，H b Indicating the number of side bands.
Preferably, in said module M3:
module M3.1: training data
Wherein the training setOne data x of (2) i N is the number of samples, m is the feature dimension;
first by a nonlinear transformation function Φ: the data is mapped from an original space to a feature space by x-F, F is a mapped high-dimensional feature space, and then a hypersphere with the minimum volume is constructed in the feature space, so that the following optimization problem is solved:
s.t.||Φ(x i )-C|| 2 ≤R 2 +ξ i ，ξ i ≥0
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error;
module M3.2: using the lagrangian multiplier method, solve the dual problem of the above problem:
Wherein a is i For sample x i Corresponding Lagrangian coefficient, a j For sample x j Corresponding Lagrangian coefficients, K (x i ，x j ) Is a kernel function, equal to the inner product of the samples in feature space, K (x i ，x j )＝<Φ(x i )，Φ(x j )>By K ij Represents K (x) i ，x j ) The method comprises the steps of carrying out a first treatment on the surface of the Obtaining Lagrange coefficients corresponding to all samples by solving the dual problem, wherein the Lagrange coefficients are more than or equal to 0 and less than or equal to a in all training samples i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
in the method, in the process of the invention,is a support vector sample in the data, K vv ＝K(x v ，x v )，K vi ＝K(x v ，x i )；
Module M3.3: and (3) fault judgment:
the signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the data, the distance d from the data to the sphere center of the super sphere is calculated as follows:
wherein K is tt ＝K(x t ，x t )，K ti ＝K(x t ，x i )；
If d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to the normal sample, otherwise, the test sample belongs to the abnormal sample, and the radius R of the hypersphere is the health threshold under the given working condition.
Preferably, in said module M4:
performing anomaly tracing by using a recursive feature elimination method, wherein the number of training samples is n,
Using all support vectors x v Square R of the resulting radius R 2 (x v ) As a criterion for measuring the magnitude of the boundary change, the average value J r Is defined as
J r ＝∑R 2 (x v )/t
Where t is the number of support vectors, and after introducing a linear kernel function, the criterion function can be expressed as:
let J r (-F) is the radius of the hypersphere obtained by removing features other than F, the feature that contributes most to anomalies is such that J r Features corresponding to maximum values of (-F) and, after removal of the feature F, the criterion function is applied by DJ r (F)＝J r -J r (-F) represents the feature F having the highest degree of contribution to abnormality * Is such that DJ r (F) Having features corresponding to the minimum, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) final fault positioning and tracing are carried out.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention provides a complex transmission device fault diagnosis tracing method based on time-frequency domain characteristics, which solves the transmission device fault diagnosis problem of small data samples and a large amount of fault data deficiency based on the ideas of support vector data description and characteristic recursion elimination;
2. the invention uses time domain index to represent the whole information of the transmission device, uses frequency domain index to represent the shaft and gear information, constructs the minimum hypersphere description and the whole health state of the characterization model machine, judges the fault state, eliminates the recursion characteristic, and searches the cause of the fault from the input characteristic layer.
Drawings
Other features, objects and advantages of the present invention will become more apparent upon reading of the detailed description of non-limiting embodiments, given with reference to the accompanying drawings in which:
FIG. 1 is a general flow chart of a fault diagnosis method of a complex transmission device based on time-frequency characteristics.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the present invention, but are not intended to limit the invention in any way. It should be noted that variations and modifications could be made by those skilled in the art without departing from the inventive concept. These are all within the scope of the present invention.
Example 1:
according to the complex transmission device fault diagnosis tracing method based on the time-frequency domain characteristics, as shown in fig. 1, the method comprises the following steps:
step S1: collecting vibration signals to obtain sample data;
specifically, in the step S1:
testing at a specified rotating speed, installing and arranging a vibration sensor on a transmission device for measuring vibration signals and converting the vibration signals into electric signals for input, and acquiring signals s of a vehicle-mounted CAN bus and a three-way acceleration vibration sensor through multichannel data i The acquisition device converts the vibration electric signals into digital signals and transmits the digital signals to the computer.
Step S2: training samples according to the acquired signals, and preprocessing test samples;
specifically, in the step S2:
calculating the signal s i Time and frequency domain features of (a): the time domain features include: root mean square value Z rm s 、 Kurtosis Z kurt And margin factor Z e The method comprises the steps of carrying out a first treatment on the surface of the The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C ；
Step S2.1: normalizing the training sample and the test sample, and performing noise reduction on the data;
step S2.2: calculating time domain features, root mean square values of signalsKurtosis-> Margin factor->
Wherein the discrete signal sequence s= [ s ] 1 ，s 2 ，...，s N ]N is the number of sampling points, s k For the data of the k-th one,as the mean value of the vibration signal, Z pp Is the signal peak;
step S2.3: calculating frequency domain index of signal, constructing arithmetic mean value X of frequency conversion and high-order harmonic energy of frequency conversion for each axis s ：
Wherein E is so Energy representing the o-th order of the conversion, o=1, 2, … H s ，H s Representing the highest order of the frequency conversion;
constructing a geometric mean C of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
For each gear pair, an arithmetic mean value Xx of the meshing frequency and the higher harmonic energy of the meshing frequency is defined:
wherein E is gp Energy representing the p-th order of the meshing frequency, p=1, 2, … H g ，H g Representing the highest order of the meshing frequency;
for each gear pair, defining geometric mean C of meshing frequency and higher harmonic energy of the meshing frequency g ：
For each gear pair, an arithmetic mean M of the modulation intensity of each order of meshing frequency is defined X And geometric mean M C ：
Wherein E is bqp Represents the energy of the q-th side band of the p-th order meshing frequency, q=1, 2, … H b ，H b Indicating the number of side bands.
Step S3: constructing a minimum hypersphere model, and inputting a test sample for fault diagnosis;
specifically, in the step S3:
step S3.1: training data
Wherein the training setOne data x of (2) i N is the number of samples, m is the feature dimension;
first by a nonlinear transformation function Φ: the data is mapped from an original space to a feature space by x-F, F is a mapped high-dimensional feature space, and then a hypersphere with the minimum volume is constructed in the feature space, so that the following optimization problem is solved:
s.t.||Φ(x i )-C|| 2 ≤R 2 +ξ i ，ξ i ≥0
Wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error;
step S3.2: using the lagrangian multiplier method, solve the dual problem of the above problem:
wherein a is i For sample x i Corresponding Lagrangian coefficient, a j For sample x j Corresponding Lagrangian coefficients, K (x i ，x j ) Is a kernel function, equal to the inner product of the samples in feature space, K (x i ，x j )＝<Φ(x i )，Φ(x j )>By K ij Represents K (x) i ，x j ) The method comprises the steps of carrying out a first treatment on the surface of the Obtaining Lagrange coefficients corresponding to all samples by solving the dual problem, wherein the Lagrange coefficients are more than or equal to 0 and less than or equal to a in all training samples i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
in the method, in the process of the invention,is a support vector sample in the data, K vv ＝K(x v ，x v )，K vi ＝K(x v ，x i )；
Step S3.3: and (3) fault judgment:
the signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the data, the distance d from the data to the sphere center of the super sphere is calculated as follows:
wherein K is tt ＝K(x t ，x t )，K ti ＝K(x t ，x i )；
If d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to the normal sample, otherwise, the test sample belongs to the abnormal sample, and the radius R of the hypersphere is the health threshold under the given working condition.
Step S4: and performing anomaly tracing by using a recursive feature elimination method to perform transmission fault diagnosis.
Specifically, in the step S4:
performing anomaly tracing by using a recursive feature elimination method, wherein the number of training samples is n,
using all support vectors x v Square R of the resulting radius R 2 (x v ) As a criterion for measuring the magnitude of the boundary change, the average value J r Is defined as
J r ＝∑R 2 (x v )/t
Where t is the number of support vectors, and after introducing a linear kernel function, the criterion function can be expressed as:
let J r (-F) is the radius of the hypersphere obtained by removing features other than F, the feature that contributes most to anomalies is such that J r Features corresponding to maximum values of (-F) and, after removal of the feature F, the criterion function is applied by DJ r (F)＝J r -J r (-F) represents the feature F having the highest degree of contribution to abnormality * Is such that DJ r (F) Having features corresponding to the minimum, i.e.
/>
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) final fault positioning and tracing are carried out.
Example 2:
example 2 is a preferable example of example 1 to more specifically explain the present invention.
The time-frequency domain feature-based complex transmission device fault diagnosis tracing method provided by the invention can be understood by a person skilled in the art as a specific implementation mode of the time-frequency domain feature-based complex transmission device fault diagnosis tracing system, namely, the time-frequency domain feature-based complex transmission device fault diagnosis tracing system can be realized by executing the step flow of the time-frequency domain feature-based complex transmission device fault diagnosis tracing method.
The invention provides a complex transmission device fault diagnosis tracing system based on time-frequency domain characteristics, which comprises the following components:
module M1: collecting vibration signals to obtain sample data;
specifically, in the module M1:
testing at a specified rotation speed, and installing the vibration sensor on the clothThe signal s is arranged on the transmission device and used for measuring vibration signals and converting the vibration signals into electric signals for input, and the signals s are input into the multichannel data acquisition vehicle-mounted CAN bus and the three-way acceleration vibration sensor i The acquisition device converts the vibration electric signals into digital signals and transmits the digital signals to the computer.
Module M2: training samples according to the acquired signals, and preprocessing test samples;
specifically, in the module M2:
calculating the signal s i Time and frequency domain features of (a): the time domain features include: root mean square value Z rm s 、 Kurtosis Z kurt And margin factor Z e The method comprises the steps of carrying out a first treatment on the surface of the The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C ；
Module M2.1: normalizing the training sample and the test sample, and performing noise reduction on the data;
Module M2.2: calculating time domain features, root mean square values of signalsKurtosis-> Margin factor->
Wherein the discrete signal sequence s= [ s ] 1 ，s 2 ，...，s N ]N is the number of sampling points, s k For the data of the k-th one,as the mean value of the vibration signal, Z pp Is the signal peak;
module M2.3: calculating frequency domain index of signal, constructing arithmetic mean value X of frequency conversion and high-order harmonic energy of frequency conversion for each axis s ：
Wherein E is so Energy representing the o-th order of the conversion, o=1, 2, … H s ，H s Representing the highest order of the frequency conversion;
constructing a geometric mean C of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
/>
For each gear pair, an arithmetic mean value X of the meshing frequency and the higher harmonic energy of the meshing frequency is defined g ：
Wherein E is gp Energy representing the p-th order of the meshing frequency, p=1, 2, … H g ，H g Representing the highest order of the meshing frequency;
for each gear pair, defining geometric mean C of meshing frequency and higher harmonic energy of the meshing frequency g ：
For each gear pair, an arithmetic mean M of the modulation intensity of each order of meshing frequency is defined X And geometric mean M C ：
Wherein E is bqp Represents the energy of the q-th side band of the p-th order meshing frequency, q=1, 2, … H b ，H b Indicating the number of side bands.
Module M3: constructing a minimum hypersphere model, and inputting a test sample for fault diagnosis;
Specifically, in the module M3:
module M3.1: training data
Wherein the training setOne data x of (2) i N is the number of samples, m is the feature dimension;
first by a nonlinear transformation function Φ: the data is mapped from an original space to a feature space by x-F, F is a mapped high-dimensional feature space, and then a hypersphere with the minimum volume is constructed in the feature space, so that the following optimization problem is solved:
s.t.||Φ(x i )-C|| 2 ≤R 2 +ξ i ，ξ i ≥0
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error;
module M3.2: using the lagrangian multiplier method, solve the dual problem of the above problem:
wherein a is i For sample x i Corresponding Lagrangian coefficient, a j For sample x j Corresponding Lagrangian coefficients, K (x i ，x j ) Is a kernel function, equal to the inner product of the samples in feature space, K (x i ，x j )＝<Φ(x i )，Φ(x j )>By K ij Represents K (x) i ，x j ) The method comprises the steps of carrying out a first treatment on the surface of the Obtaining Lagrange coefficients corresponding to all samples by solving the dual problem, wherein the Lagrange coefficients are more than or equal to 0 and less than or equal to a in all training samples i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
In the method, in the process of the invention,is a support vector sample in the data, K vv ＝K(x v ，x v )，K vi ＝K(x v ，x i )；
Module M3.3: and (3) fault judgment:
the signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the dataWhen the distance d from the spherical center of the super-sphere to the spherical center is calculated as follows:
wherein K is tt ＝K(x t ，x t )，K ti ＝K(x t ，x i )；
If d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to the normal sample, otherwise, the test sample belongs to the abnormal sample, and the radius R of the hypersphere is the health threshold under the given working condition.
Module M4: and performing anomaly tracing by using a recursive feature elimination method to perform transmission fault diagnosis.
Specifically, in the module M4:
performing anomaly tracing by using a recursive feature elimination method, wherein the number of training samples is n,
using all support vectors x v Square R of the resulting radius R 2 (x v ) As a criterion for measuring the magnitude of the boundary change, the average value J r Is defined as
J r ＝∑R 2 (x v )/t
Where t is the number of support vectors, and after introducing a linear kernel function, the criterion function can be expressed as:
let J r (-F) is the radius of the hypersphere obtained by removing features other than F, the feature that contributes most to anomalies is such that J r Features corresponding to maximum values of (-F) and, after removal of the feature F, the criterion function is applied by DJ r (F)＝J r -J r (-F) represents the feature F having the highest degree of contribution to abnormality * Is such that DJ r (F) Having features corresponding to the minimum, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) final fault positioning and tracing are carried out.
Example 3:
example 3 is a preferable example of example 1 to more specifically explain the present invention.
The invention provides a complex transmission device fault diagnosis tracing method based on time-frequency domain characteristics, which can carry out fault diagnosis on a transmission device and position fault reasons under the condition that a fault data set is not rich.
The invention provides a method based on support vector data description and feature recursion elimination, which is used for extracting time domain and frequency domain features of a vibration signal of a transmission device, judging the working state of the transmission device by using the support vector data description method and finding out the most likely cause of system faults by using the feature recursion elimination method.
The invention provides a time-frequency domain feature-based complex transmission device fault diagnosis tracing method, which is characterized in that time domain and frequency domain features are extracted to carry out fault diagnosis according to priori knowledge and vibration signals of a complex transmission device, faults of the transmission device are traced back to the time domain and frequency domain feature levels, contribution degrees of all features are ordered by using a feature recursion elimination method, and main features causing faults are determined to realize fault tracing.
Aiming at the problems, the invention aims to provide a fault diagnosis and tracing method of a complex transmission device based on time-frequency domain characteristics for the whole complex transmission device, so as to realize fault judgment and tracing positioning of the transmission device. The flow chart of the method is shown in figure 1.
Step 1, testing at a specified rotating speed, installing and arranging a vibration sensor on a transmission device for measuring vibration signals and converting the vibration signals into electric signals for input, and acquiring signals s of a vehicle-mounted CAN bus and a three-way acceleration vibration sensor through multichannel data i The acquisition device converts the vibration electric signal intoThe digital signal is transmitted to a computer.
Step 2, the signal s needs to be calculated i Time domain and frequency domain features of (a). The time domain features include: root mean square value Z rms Kurtosis Z kurt And margin factor Z e . The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C 。
And 3, constructing a minimum hypersphere model for fault diagnosis.
Step 31, training data Where n is the number of samples and m is the feature dimension. First by a nonlinear transformation function Φ: the x-F maps the data from the original space to the characteristic space, and then constructs the hypersphere with the minimum volume in the characteristic space, and specifically solves the following optimization problems:
s.t.||Φ(x i )-C|| 2 ≤R 2 +ξ i ，ξ i ≥0
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error.
Step 32, solving the dual problem of the above problem using Lagrangian multiplier method
Wherein a is i For sample x i Corresponding Lagrangian coefficients, K (x i ，x j ) Is a kernel function, equivalent to the inner product of samples in feature space, i.e., K (x i ，x j )＝<Φ(x i )，Φ(x j )>For simplicity, K is used hereinafter ij And (3) representing. By solving the dual problem, lagrange coefficients corresponding to all samples can be obtained, and in all training samples, the Lagrange coefficients are satisfied with 0-a i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
wherein K is vv ＝K(x v ，x v )，K vi ＝K(x v ，x i )，
And step 33, fault judgment.
The signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the data, the distance d from the data to the sphere center of the super sphere is calculated as follows:
if d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to a normal sample, otherwise, belongs to an abnormal sample. The radius R of the hypersphere is the health threshold under the given working condition.
And 4, performing anomaly tracing by using a recursive feature elimination method. Training samples are n, R is obtained by using all support vectors 2 (x v ) As a criterion function of the magnitude of the boundary change, the average is defined as J r ＝∑R 2 (x v ) T, where t is the number of support vectors, and after introducing a linear kernel function, the criterion function is expressed as:
let J r (-F) is the radius of the hypersphere obtained by removing features other than F, the feature that contributes most to anomalies is such that J r Features corresponding to the maximum value of the value of (-F). DJ for criterion function after removal of feature F r (F)＝J r -J r (-F) represents. Feature F having highest contribution to abnormality * Is of minimum DJ r (F) Corresponding features, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) performing final fault location.
In step 1, a single axis acceleration sensor, a displacement sensor, or other vibration sensor may be used in addition to the three-way acceleration sensor.
In step 2, other time domain features or frequency domain features may also be selected by an optimization method.
In step 3, when an abnormal sample is fused, the super sphere radius R can be perfected by adding an abnormal sample, so that the phenomenon of overfitting is prevented.
Example 4:
example 4 is a preferable example of example 1 to more specifically explain the present invention.
The invention provides a fault diagnosis and tracing method of a complex transmission device based on time-frequency domain characteristics, which has great practical reference significance for fault diagnosis of the complex transmission device and comprises the following specific steps of.
And step 1, acquiring sample data. And arranging the vibration sensors according to the optimized positions on the transmission device comprehensive test bed, connecting the vehicle-mounted CAN bus and the vibration sensors with a signal acquisition device and a computer, and acquiring and recording data returned by the vehicle-mounted CAN bus and the vibration sensors by the computer system. When the integrated transmission device runs at the engine rotating speeds of 1200RPM, 1700RPM and 2300RPM, a three-way acceleration sensor is used for collecting vibration signals every 10 seconds, and 10 sections are collected. The discrete vibration signal s (t) is obtained through the pretreatment of the de-mean value and the de-trend item i ) I=1, 2,. -%, N wherein t i Representing the sampling instants.
And 2, preprocessing a training sample and a test sample.
And step 21, carrying out normalization processing on the training samples and the test samples, and carrying out noise reduction processing on the data.
In step 22 of the process of the present invention,
calculating time domain features, root mean square values of signalsKurtosis->Margin factor->
Step 23, calculating the frequency domain index of the signal.
Constructing an arithmetic mean value X of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s And geometric mean C s ：Wherein E is so Energy representing each order of the frequency conversion, o represents the order o=1, 2, … H of the o-th frequency conversion s ，H s Representing the highest order of the frequency conversion; for each gear pair, an arithmetic mean value X of the meshing frequency and the higher harmonic energy of the meshing frequency is defined g And geometric mean C g ：/> Wherein E is gp The energy of each order of the meshing frequency is represented, p represents the order p=1, 2, … H of the p-th meshing frequency g ，H g Representing the highest order of the meshing frequency; defining the arithmetic mean M of the modulation intensities of the meshing frequencies of the respective orders X And geometric mean M C ：/>Wherein E is bqp Represents the energy of each side band of the p-th order meshing frequency, q is the q-th side band q=1, 2, … H b ，H b Indicating the number of side bands.
Taking the narrow-band energy in the range of 1% of the actual order after resampling as the frequency energy to calculate the frequency domain index in consideration of the frequency dispersion caused by the rotation speed fluctuation which is necessarily existed in the actual engine.
And 3, constructing a minimum hypersphere model for fault diagnosis.
Step 31, training dataWhere n is the number of samples and m is the feature dimension. First by a nonlinear transformation function Φ: the x-F maps the data from the original space to the characteristic space F, and then the hypersphere with the minimum volume is constructed in the characteristic space, so that the following optimization problem is solved:
s.t.||Φ(x i )-C|| 2 ≤R 2 +ξ i ，ξ i ≥0
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error.
Step 32, solving the dual problem of the above problem using Lagrangian multiplier method
Wherein a is i For sample x i Corresponding Lagrangian coefficients, K (x i ，x j ) Is a kernel function, equivalent to the inner product of samples in feature space, i.e., K (x i ，x j )＝<Φ(x i )，Φ(x j )>For simplicity, K is used hereinafter ij And (3) representing. By solving the dual problem, lagrange coefficients corresponding to all samples can be obtained, and in all training samples, the Lagrange coefficients are satisfied with 0-a i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
wherein K is vv ＝K(x v ，x v )，K vi ＝K(x v ，x i )，
And step 33, fault judgment.
The signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the data, the distance d from the data to the sphere center of the super sphere is calculated as follows:
if d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to a normal sample, otherwise, belongs to an abnormal sample. The radius R of the hypersphere is the health threshold under the given working condition.
And 4, performing anomaly tracing by using a recursive feature elimination method. Training samples are n, R is obtained by using all support vectors 2 (x v ) As a criterion function of the magnitude of the boundary change, the average is defined as J r ＝∑R 2 (x v ) T, where t is the number of support vectors, and after introducing a linear kernel function, the criterion function is expressed as:
let J r (-F) is the radius of the hypersphere obtained by removing features other than F, the feature that contributes most to anomalies is such that J r Features corresponding to the maximum value of the value of (-F). DJ for criterion function after removal of feature F r (F)＝J r -J r (-F) represents. Feature F having highest contribution to abnormality * Is of minimum DJ r (F) Corresponding features, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) performing final fault location.
Those skilled in the art will appreciate that the systems, apparatus, and their respective modules provided herein may be implemented entirely by logic programming of method steps such that the systems, apparatus, and their respective modules are implemented as logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc., in addition to the systems, apparatus, and their respective modules being implemented as pure computer readable program code. Therefore, the system, the apparatus, and the respective modules thereof provided by the present invention may be regarded as one hardware component, and the modules included therein for implementing various programs may also be regarded as structures within the hardware component; modules for implementing various functions may also be regarded as being either software programs for implementing the methods or structures within hardware components.
The foregoing describes specific embodiments of the present invention. It is to be understood that the invention is not limited to the particular embodiments described above, and that various changes or modifications may be made by those skilled in the art within the scope of the appended claims without affecting the spirit of the invention. The embodiments of the present application and features in the embodiments may be combined with each other arbitrarily without conflict.
Claims (8)
1. The fault diagnosis and tracing method for the complex transmission device based on the time-frequency domain features is characterized by comprising the following steps of:
step S1: collecting vibration signals to obtain sample data;
step S2: training samples according to the acquired signals, and preprocessing test samples;
step S3: constructing a minimum hypersphere model, and inputting a test sample for fault diagnosis;
step S4: performing abnormal tracing by using a recursive feature elimination method to perform transmission device fault diagnosis;
in the step S2:
calculating the signal s i Time and frequency domain features of (a): the time domain features include: root mean square value Z rms Kurtosis Z kurt And margin factor Z e The method comprises the steps of carrying out a first treatment on the surface of the The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C ；
Step S2.1: normalizing the training sample and the test sample, and performing noise reduction on the data;
step S2.2: calculating time domain features, root mean square values of signalsKurtosis-> Margin factor->
Wherein the discrete signal sequence s= [ s ] 1 ,2,..., N ]N is the number of sampling points, s k For the data of the k-th one,as the mean value of the vibration signal, Z pp Is the signal peak;
step S2.3: constructing an arithmetic mean value X of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
Wherein E is so Energy representing the o-th order of the conversion, o=1, 2, … H s ，H s Representing the highest order of the frequency conversion;
constructing a geometric mean C of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
For each gear pair, an arithmetic mean value X of the meshing frequency and the higher harmonic energy of the meshing frequency is defined g ：
Wherein E is gp Energy representing the p-th order of the meshing frequency, p=1, 2, … H g ，H g Representing the highest order of the meshing frequency;
for each gear pair, defining geometric mean C of meshing frequency and higher harmonic energy of the meshing frequency g ：
For each gear pair, an arithmetic mean M of the modulation intensity of each order of meshing frequency is defined X And geometric mean M C ：
Wherein E is bqp Representing the p-th order of the meshing frequencyRate energy of the q-th sideband, q=1, 2, … H b ，H b Indicating the number of side bands.
2. The method for tracing fault diagnosis of complex transmission device based on time-frequency domain features as claimed in claim 1, wherein in said step S1:
testing at a specified rotating speed, installing and arranging a vibration sensor on a transmission device for measuring vibration signals and converting the vibration signals into electric signals for input, and acquiring signals s of a vehicle-mounted CAN bus and a three-way acceleration vibration sensor through multichannel data i The acquisition device converts the vibration electric signals into digital signals and transmits the digital signals to the computer.
3. The method for tracing fault diagnosis of complex transmission device based on time-frequency domain features as claimed in claim 1, wherein in said step S3:
step S3.1: training data
Wherein the training setOne data x of (2) i N is the number of samples, m is the feature dimension;
firstly, mapping data from an original space to a feature space through a nonlinear transformation function phi x-F, wherein F is a mapped high-dimensional feature space, and then constructing a hypersphere with the minimum volume in the feature space, so as to solve the following optimization problems:
s.t.‖Φ(x i )-C‖ 2 ≤R 2 +ξ i ,ξ i ≥0
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is pineA relaxation factor, lambda is a penalty parameter that balances the volume of the hypersphere and the fraction error;
step S3.2: using the lagrangian multiplier method, solve the dual problem of the above problem:
wherein a is i For sample x i Corresponding Lagrangian coefficient, a j For sample x j Corresponding Lagrangian coefficients, K (x i ,x j ) Is a kernel function, equal to the inner product of the samples in feature space, K (x i ,x j )＝<Φ(x i ),Φ(x j )>By K ij Represents K (x) i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Obtaining Lagrange coefficients corresponding to all samples by solving the dual problem, wherein the Lagrange coefficients are more than or equal to 0 and less than or equal to a in all training samples i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
in the method, in the process of the invention,is a number ofAccording to the support vector samples, K vv ＝K(x v ,x v )，K vi ＝K(x v ,x i )；
Step S3.3: and (3) fault judgment:
the signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the data, the distance d from the data to the sphere center of the super sphere is calculated as follows:
wherein K is tt ＝K(x t ,x t )，K ti ＝K(x t ,x i )；
If d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to the normal sample, otherwise, the test sample belongs to the abnormal sample, and the radius R of the hypersphere is the health threshold under the given working condition.
4. The method for tracing fault diagnosis of complex transmission device based on time-frequency domain features as claimed in claim 1, wherein in said step S4:
performing anomaly tracing by using a recursive feature elimination method, wherein the number of training samples is n,
using all support vectors x v Square R of the resulting radius R 2 (x v ) As a criterion for measuring the magnitude of the boundary change, the average value J r Is defined as
J r ＝ΣR 2 (x v )/t
Where t is the number of support vectors, and after introducing a linear kernel function, the criterion function can be expressed as:
Let J r (-) is a superobtained by removing features FSphere radius, the feature that contributes most to anomalies is to make J r Features corresponding to the maximum (-) value, after removal of feature F, use of DJ for criterion function r ()＝J r -J r (-) represents the feature F with the highest degree of contribution to abnormality * Is such that DJ r () Having features corresponding to the minimum, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * And (5) final fault positioning and tracing are carried out.
5. A complex transmission device fault diagnosis traceability system based on time-frequency domain features is characterized by comprising:
module M1: collecting vibration signals to obtain sample data;
module M2: training samples according to the acquired signals, and preprocessing test samples;
module M3: constructing a minimum hypersphere model, and inputting a test sample for fault diagnosis;
module M4: performing abnormal tracing by using a recursive feature elimination method to perform transmission device fault diagnosis;
in the module M2:
calculating the signal s i Time and frequency domain features of (a): the time domain features include: root mean square value Z rms Kurtosis Z kurt And margin factor Z e The method comprises the steps of carrying out a first treatment on the surface of the The frequency domain features include: arithmetic mean value X of the converted frequency and the converted frequency higher harmonic energy for each axis s And geometric mean C s The method comprises the steps of carrying out a first treatment on the surface of the Definition of the meshing frequency for a gear pair and arithmetic mean value X of the higher harmonic energy of the meshing frequency g And geometric mean C g Defining arithmetic mean M of modulation intensity of each order of meshing frequency X And geometric mean M C ；
Module M2.1: normalizing the training sample and the test sample, and performing noise reduction on the data;
module M2.2: calculating time domain features, root mean square values of signalsKurtosis-> Margin factor->
Wherein the discrete signal sequence s= [ s ] 1 ,2,..., N ]N is the number of sampling points, s k For the data of the k-th one,as the mean value of the vibration signal, Z pp Is the signal peak;
module M2.3: constructing an arithmetic mean value X of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
Wherein E is so Energy representing the o-th order of the conversion, o=1, 2, … H s ，H s Representing the highest order of the frequency conversion;
constructing a geometric mean C of the frequency conversion and the high-order harmonic energy of the frequency conversion for each axis s ：
For each gear pair, an arithmetic mean value X of the meshing frequency and the higher harmonic energy of the meshing frequency is defined g ：
Wherein E is gp Energy representing the p-th order of the meshing frequency, p=1, 2, … H g ，H g Representing the highest order of the meshing frequency;
for each gear pair, defining geometric mean C of meshing frequency and higher harmonic energy of the meshing frequency g ：
For each gear pair, an arithmetic mean M of the modulation intensity of each order of meshing frequency is defined X And geometric mean M C ：
Wherein E is bqp Represents the energy of the q-th side band of the p-th order meshing frequency, q=1, 2, … H b ，H b Indicating the number of side bands.
6. The complex transmission fault diagnosis and tracing system based on time-frequency domain characteristics according to claim 5, wherein in said module M1:
testing at a specified rotating speed, installing and arranging a vibration sensor on a transmission device for measuring vibration signals and converting the vibration signals into electric signals for input, and acquiring signals s of a vehicle-mounted CAN bus and a three-way acceleration vibration sensor through multichannel data i The acquisition device converts the vibration electric signals into digital signals and transmits the digital signals to the computer.
7. The complex transmission fault diagnosis and tracing system based on time-frequency domain characteristics according to claim 5, wherein in said module M3:
module M3.1: training data
Wherein the training setOne data x of (2) i N is the number of samples, m is the feature dimension;
firstly, mapping data from an original space to a feature space through a nonlinear transformation function phi x-F, wherein F is a mapped high-dimensional feature space, and then constructing a hypersphere with the minimum volume in the feature space, so as to solve the following optimization problems:
wherein R is the radius of the super sphere, C is the sphere center of the super sphere, and xi is the diameter of the super sphere i I=1, 2 … n is the relaxation factor, λ is the penalty parameter for balancing the hypersphere volume and the fraction error;
module M3.2: using the lagrangian multiplier method, solve the dual problem of the above problem:
wherein a is i For sample x i Corresponding Lagrangian coefficient, a j For sample x j Corresponding Lagrangian coefficients, K (x i ,x j ) Is a kernel function, equal to the inner product of the samples in feature space, K (x i ,x j )＝<Φ(x i ),Φ(x j )>By K ij Represents K (x) i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Obtaining Lagrange coefficients corresponding to all samples by solving the dual problem, wherein the Lagrange coefficients are more than or equal to 0 and less than or equal to a in all training samples i Samples of lambda are called support vectors, and the training data set is assumed to be for a set of samples belonging to the support vectorThe calculation formulas of the sphere center C and the radius R of the super sphere are respectively as follows:
in the method, in the process of the invention,is a support vector sample in the data, K vv ＝K(x v ,x v )，K vi ＝K(x v ,x i )；
Module M3.3: and (3) fault judgment:
the signal acquisition and data preprocessing method is carried out on the transmission device to be tested to obtain a test sample x t When only normal samples exist in the data, the distance d from the data to the sphere center of the super sphere is calculated as follows:
wherein K is tt ＝K(x t ,x t )，K ti ＝K(x t ,x i )；
If d is less than or equal to R, the test sample is on or in the hypersphere, and belongs to the normal sample, otherwise, the test sample belongs to the abnormal sample, and the radius R of the hypersphere is the health threshold under the given working condition.
8. The complex transmission fault diagnosis and tracing system based on time-frequency domain characteristics according to claim 5, wherein in said module M4:
performing anomaly tracing by using a recursive feature elimination method, wherein the number of training samples is n,
using all support vectors x v Square R of the resulting radius R 2 ( v ) As a criterion for measuring the magnitude of the boundary change, the average value J r Defined as J r ＝R 2 ( v )/t
Where t is the number of support vectors, and after introducing a linear kernel function, the criterion function can be expressed as:
let J r (-) is the radius of the hypersphere obtained by removing features F, the feature that contributes most to anomalies is such that J r Features corresponding to the maximum (-) value, after removal of feature F, use of DJ for criterion function r ()＝J r -J r (-) represents the feature F with the highest degree of contribution to abnormality * Is such that DJ r () Having features corresponding to the minimum, i.e.
Tracing faults to time domain and frequency domain characteristics F of specified gears or shafts * Final processFault location and tracing of (a).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210719789.0A CN115096581B (en) | 2022-06-23 | 2022-06-23 | Complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210719789.0A CN115096581B (en) | 2022-06-23 | 2022-06-23 | Complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115096581A CN115096581A (en) | 2022-09-23 |
| CN115096581B true CN115096581B (en) | 2023-07-28 |
Family
ID=83293402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210719789.0A Active CN115096581B (en) | 2022-06-23 | 2022-06-23 | Complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115096581B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117056849B (en) * | 2023-10-12 | 2024-02-02 | 智能制造龙城实验室 | Unsupervised method and system for monitoring abnormal state of complex mechanical equipment |
| CN117932230B (en) * | 2024-03-21 | 2024-05-28 | 北京航空航天大学 | Fault diagnosis method based on characteristic decoupling and sample recombination under small sample condition |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103488091A (en) * | 2013-09-27 | 2014-01-01 | 上海交通大学 | Data-driving control process monitoring method based on dynamic component analysis |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104657574B (en) * | 2014-06-13 | 2017-10-31 | 苏州大学 | The method for building up and device of a kind of medical diagnosismode |
| CN107393525B (en) * | 2017-07-24 | 2020-06-12 | 湖南大学 | Speech emotion recognition method integrating feature evaluation and multilayer perceptron |
| US20200311611A1 (en) * | 2019-03-26 | 2020-10-01 | Caseware International Inc. | Feature generation and feature selection for machine learning tool |
| CN110020638B (en) * | 2019-04-17 | 2023-05-12 | 唐晓颖 | Facial expression recognition method, device, equipment and medium |
| AU2020101854A4 (en) * | 2020-08-17 | 2020-09-24 | China Communications Construction Co., Ltd. | A method for predicting concrete durability based on data mining and artificial intelligence algorithm |
-
2022
- 2022-06-23 CN CN202210719789.0A patent/CN115096581B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103488091A (en) * | 2013-09-27 | 2014-01-01 | 上海交通大学 | Data-driving control process monitoring method based on dynamic component analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115096581A (en) | 2022-09-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115096581B (en) | Complex transmission device fault diagnosis tracing method and system based on time-frequency domain characteristics | |
| Zou et al. | Bearing fault diagnosis method based on EEMD and LSTM | |
| Shen et al. | Fault diagnosis of rotating machinery based on the statistical parameters of wavelet packet paving and a generic support vector regressive classifier | |
| CN102607845B (en) | Bearing fault characteristic extracting method for redundantly lifting wavelet transform based on self-adaptive fitting | |
| CN111679654B (en) | Electric drive system vibration signal diagnosis method and device based on neural network | |
| CN107702922B (en) | Rolling bearing fault diagnosis method based on LCD and stacking automatic encoder | |
| CN111089720B (en) | Regularization sparse filtering method suitable for gear fault diagnosis under variable rotating speed | |
| CN112906644A (en) | Mechanical fault intelligent diagnosis method based on deep migration learning | |
| CN110824281A (en) | Method and system for on-line monitoring and fault diagnosis of synchronous phase modulator | |
| Wang et al. | Gear fault diagnosis based on variational modal decomposition and wide+ narrow visual field neural networks | |
| Hu et al. | Vibration signal denoising method based on CEEMDAN and its application in brake disc unbalance detection | |
| CN112182490B (en) | Reactor state diagnosis method and system | |
| CN111398811A (en) | PMSM demagnetization fault diagnosis method based on terminal current cost sensitive learning | |
| Xu et al. | An adaptive spectrum segmentation method to optimize empirical wavelet transform for rolling bearings fault diagnosis | |
| CN113390631A (en) | Fault diagnosis method for gearbox of diesel engine | |
| CN114462508A (en) | Power transformer health state assessment method based on multi-mode neural network | |
| CN112577736A (en) | Wind turbine generator set planetary gearbox fault diagnosis method based on SANC and 1D-CNN-LSTM | |
| Cao et al. | Fault diagnosis of rolling bearing based on multiscale one-dimensional hybrid binary pattern | |
| CN112881015A (en) | Measuring point insensitive fault detection method based on bearing force identification | |
| CN112734001A (en) | Wind power transmission chain intelligent fault diagnosis method based on order spectrum migration | |
| Wu et al. | Discovering New Prognostic Features for the Harmonic Reducer in Remaining Useful Life Prediction | |
| Ma et al. | Compound Fault Diagnosis of Rolling Bearing Based on ALIF‐KELM | |
| Xu et al. | Residual life prediction of metro traction motor bearing based on convolutional neural network | |
| CN114235408A (en) | Bearing fault diagnosis method and system based on improved cascade variational modal decomposition | |
| CN115979636A (en) | Comprehensive transmission state abnormity intelligent detection method based on high-dimensional vibration signal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |