CN117574114B - Remote reconstruction and jump disturbance detection method for running data of rotary machine - Google Patents
Remote reconstruction and jump disturbance detection method for running data of rotary machine Download PDFInfo
- Publication number
- CN117574114B CN117574114B CN202410054151.9A CN202410054151A CN117574114B CN 117574114 B CN117574114 B CN 117574114B CN 202410054151 A CN202410054151 A CN 202410054151A CN 117574114 B CN117574114 B CN 117574114B
- Authority
- CN
- China
- Prior art keywords
- data
- signal
- generator
- training
- vibration signal
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 44
- 230000000739 chaotic effect Effects 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 28
- 230000035772 mutation Effects 0.000 claims abstract description 24
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000036541 health Effects 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 7
- 230000009191 jumping Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims 1
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 claims 1
- 238000012423 maintenance Methods 0.000 abstract description 19
- 238000002474 experimental method Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 15
- 238000013135 deep learning Methods 0.000 description 11
- 238000003745 diagnosis Methods 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 8
- 230000007547 defect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000013016 damping Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000005312 nonlinear dynamic Methods 0.000 description 2
- 238000012552 review Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 239000003643 water by type Substances 0.000 description 1
Abstract
The invention is suitable for the field of remote intelligent operation and maintenance and information testing of rotary mechanical equipment operation, in particular to a method for remote reconstruction and jump disturbance detection of rotary mechanical operation data, which comprises the following steps: preprocessing the training set data, detecting mutation points in the data, and obtaining preprocessed equipment operation data; constructing conditions to generate an countermeasure network model, defining a network structure of the generator and the discriminator, and carrying out iterative updating on the model to obtain a reconstructed vibration signal; determining the reconstruction performance of the generated vibration signal by using a chaotic attractor method; performing jump mutation disturbance detection on the obtained generated vibration signal; and predicting and tracking the running state of the equipment according to the predicted output result, and finally realizing the predictive maintenance of the mechanical equipment. The invention can primarily solve the problems of less fault disclosure data, high acquisition cost and high experiment difficulty without considering the influence of working conditions and mechanical equipment structures.
Description
Technical Field
The invention belongs to the technical field of remote intelligent operation and maintenance and information testing of rotary mechanical equipment operation, and particularly relates to a method for remote reconstruction and jump disturbance detection of rotary mechanical operation data.
Background
Mechanical fault diagnosis and intelligent operation and maintenance are key technologies for ensuring long-term efficient, stable and reliable operation of important mechanical systems and equipment.
In an intelligent factory, the equipment which needs to be monitored, diagnosed and intelligent operation and maintenance at any time has large population scale, various equipment measuring points and high measuring point sampling frequency, meanwhile, the time for collecting data from the start of service to the end of service life of important mechanical equipment is long, the data scale is large and mass storage is needed.
However, the equipment has various fault types and short fault burst time, the instantaneous fault signal just becomes important resource and information for revealing the mechanical fault evolution process and fault propagation path, and in the fields of certain extreme service scenes such as nuclear radiation, strong corrosion, high-voltage electric field and the like, site engineers have difficulty in arranging sensors, so that the big data analysis, information mining, fault diagnosis and intelligent operation and maintenance of important mechanical equipment are difficult and heavy; meanwhile, the service life operation data in the existing extreme service scene has less public data, large experiment difficulty and high acquisition cost.
Therefore, how to explore faster and more universal self-adaptive generation of equipment service fault data becomes an important direction in the technical field of remote intelligent operation and maintenance and information testing of mechanical equipment operation.
In recent years, the mechanical system fault diagnosis method driven by big data reliability evaluation, deep learning and digital twin model can effectively reveal complex mapping relations between different fault types and signal characteristics, and provide theoretical basis for mechanical equipment fault diagnosis and intelligent operation and maintenance. Based on the method, how to utilize the deep learning theory to research health care technologies such as big data reliability evaluation, multi-source data fusion intelligent prediction, intelligent decision and the like is a problem faced in the field of fault diagnosis and intelligent operation and maintenance research of current mechanical equipment.
Disclosure of Invention
The invention aims to provide a remote reconstruction and jump disturbance detection method for running data of rotary machinery, and aims to solve the technical problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions.
A method for remote reconstruction and jump disturbance detection of rotary machine operation data, the method comprising the steps of:
collecting a running data sample of the rotary mechanical equipment, regarding the running data sample as real data, and dividing the real data into training set data and test set data;
Preprocessing the training set data, detecting mutation points in the data, and obtaining preprocessed equipment operation data;
constructing a condition generation countermeasure network model, classifying and marking health signals and fault signals of collected running data samples of the rotary mechanical equipment, and defining a network structure of a generator and a discriminator;
generating vibration signals through signal generators in a condition generation countermeasure network, adding the generated vibration signals into an offline data training set, calculating model gradients, carrying out iterative updating on the model, training to obtain a new generation countermeasure network model, and generating the vibration signals through the updated signal generators;
determining the reconstruction performance of the generated vibration signal by using a chaotic attractor method;
performing jump mutation disturbance detection on the obtained generated vibration signal;
and predicting and tracking the running state of the equipment according to the predicted output result, so as to realize the predictive maintenance of the mechanical equipment.
As a further scheme of the invention, a condition generation countermeasure network model is constructed, a rotating mechanical equipment operation data sample is collected to carry out classification marking of health signals and fault signals, and the steps of defining the network structure of a generator and a discriminator are as follows:
Constructing a generator network and a discriminator network structure, and generating an countermeasure network model according to the composition conditions: assuming an actual dataset Independent co-distributed sampling at probability distribution/>Sample set generated by the generator/>The probability distribution of satisfaction is/>Training the generator, training samples are/>The probability distribution of satisfaction is/>The target function form is as follows:
;
wherein G is a minimisation maximum valuation function; generating a generator function of the countermeasure network model for the condition; /(I) Generating a minimum discriminant function of the countermeasure network model for the condition; pdata (x) real data distribution; PZ (z) is noise data distribution; /(I)Representing that the data x obeys a given real data distribution Pdata; /(I)Indicating that the data z obeys a noise data distribution PZ (z);
Marking a preprocessing signal of a mechanical equipment state, wherein a normal operation signal is marked as 1, a fault operation signal is marked as 2, inputting signal marking information into a condition generation countermeasure network model, and defining a generator and a discriminator network;
Inputting the given tag information and the running signal into a generator network to generate a signal of the same structure as training data corresponding to the same tag; training by alternately training a discriminator network and a generator network, namely training the discriminator for K times and training the generator for 1 time until the objective function converges or Nash equilibrium is reached; meanwhile, the authenticity of the training data and the data generated by the generator is judged by utilizing a discriminator network; and judging the training performance by comparing whether the game and the loss function of the generator and the arbiter are converged or not.
As a further aspect of the present invention, the step of determining the reconstruction performance of the generated vibration signal by using the chaotic attractor method specifically includes:
constructing a Duffing chaotic attractor model:
wherein, For input signal to be detected,/>Is the first order derivative of the input signal to be detected,/>For second order differentiation of the input signal to be detected, parameter/>,/>,/>,/>And/>Is a constant, wherein the parameter/>Damping amount is controlled,/>Controlling the linear stiffness,/>Controlling the amount of non-linearity if/>=0, The Duffing oscillator will reduce to a simple harmonic oscillator, parameter/>And/>Amplitude and angular frequency for the input signal;
inputting an original real signal as a signal to be detected into a Duffing chaotic attractor model to obtain a Duffing chaotic attractor graph and a poincare section graph;
Inputting the generated signal as a signal to be detected into a Duffing chaotic attractor model to obtain a Duffing chaotic attractor graph and a poincare section graph;
And comparing the Duffing chaotic attractor graph generated in the two steps with the Poincare section graph, and generating a vibration signal to perform abrupt disturbance detection.
As a further scheme of the invention, the step of carrying out jump mutation disturbance detection on the obtained generated vibration signal comprises the following steps:
Calculating a test statistical variable for generating a vibration signal For detecting the presence >Whether or not there is a generated vibration signal slave/>To/>Jump mutations of (a), namely:
wherein, For/>Generating vibration signal value at moment, the time variance of random fluctuation rate isStatistical variable/>Gradually obeying normal distribution with the average value of 0 and the standard deviation of pi/2; k is window size and satisfies/>,/>At the level of significance,/>Representation of arbitrary/>There is a finite constant/>So that/>,/>;
Selecting a level of significance;
Order theSo that it meets/>Wherein/>Is a cumulative distribution function, satisfies/>;
Calculation of,/>Wherein/>N is a sample number value;
If it is Then at/>Generating a vibration signal in a time period, wherein the jump mutation condition exists, otherwise, the jump mutation condition does not exist.
Compared with the prior art, the remote reconstruction and jump disturbance detection method for the running data of the rotary machine has the beneficial effects that:
firstly, the invention combines the deep learning and chaos monitoring technology, can primarily solve the problems of less fault disclosure data, high acquisition cost and great experimental difficulty, and researchers can generate the required simulation fault data more quickly with lower cost, and the algorithm of the invention has better generalization;
secondly, the invention can realize the on-line intelligent recognition of the state and the on-way state early warning without considering the influence of the working condition and the mechanical equipment structure, and achieves the functions of real-time monitoring and the like, and has strong instantaneity and high accuracy;
Thirdly, the defects existing in the existing manual inspection are overcome, the hidden fault discovery capability and the early warning capability of the early fault generation stage are improved, and theoretical support can be provided for intelligent operation and maintenance of the complex electromechanical system.
In summary, the invention provides a method for remote reconstruction and jump disturbance detection of running data of a rotary machine by combining a deep learning technology, which overcomes the defects that a traditional fault signal is difficult to acquire and the data value density is low, can detect whether the acquired signal has the phenomena of inaccurate reconstruction and jump disturbance, utilizes the deep learning to carry out data enhancement on small sample data, solves the problem of jump disturbance of the running fault signal of the whole service life of the machine, and has important significance for improving the fault diagnosis and intelligent operation and maintenance of the machine.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following description will briefly introduce the drawings that are needed in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the present invention.
FIG. 1 is a logic schematic diagram of a method for remote reconstruction and jump disturbance detection of rotary machine operation data according to the present invention;
FIG. 2 is a flow chart of an implementation of a method for remote reconstruction and jump disturbance detection of rotary machine operation data according to the present invention;
FIG. 3 is accelerated life vibration data for 4 channels of a bearing according to the present invention;
FIG. 4 is a set 5 raw accelerated life vibration signal for bearing channel #1 of the present invention;
FIG. 5 is a cross-sectional view of (a) a chaotic attractor graph and (b) Poincare of the original accelerated life vibration signal of channel #1 of the present invention;
FIG. 6 is a reconstructed vibration signal of bearing channel #1 group 5 of the present invention;
FIG. 7 is a cross-sectional view of (a) a chaotic attractor graph and (b) Poincare of a channel #1 reconstructed vibration signal of the present invention;
FIG. 8 is a waveform diagram of the original channel #1 group 5 signal of the present invention and a reconstructed fault signal diagram of the present invention within 0.04-0.06 seconds.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
At present, the mechanical system fault diagnosis method driven by big data reliability evaluation, deep learning and digital twin model can effectively reveal complex mapping relations between different fault types and signal characteristics, and provide theoretical basis for mechanical equipment fault diagnosis and intelligent operation and maintenance. Based on the method, how to utilize the deep learning theory to research health care technologies such as big data reliability evaluation, multi-source data fusion intelligent prediction, intelligent decision and the like is a problem faced in the field of fault diagnosis and intelligent operation and maintenance research of current mechanical equipment.
In order to solve the problems, the invention provides a remote reconstruction and jump disturbance detection method for running data of a rotary machine, which overcomes the defects that the traditional fault signals are difficult to acquire and the data value density is low, can detect whether the acquired signals are inaccurate in reconstruction and jump disturbance, utilizes deep learning to carry out data enhancement on small sample data, solves the problem of jump disturbance of the fault signals of the whole-service running of the machine equipment, and has important significance for improving the fault diagnosis and intelligent operation and maintenance of the machine equipment.
Specific implementations of the invention are described in detail below in connection with specific embodiments.
As shown in fig. 1 and 2, in an embodiment of the present invention, a method for remote reconstruction and jump disturbance detection of operation data of a rotating machine includes the steps of:
step S101: collecting a running data sample of the rotary mechanical equipment, regarding the running data sample as real data, and dividing the real data into training set data and test set data;
In step S101 of the present invention, a sensor is used to collect a running data sample of the rotating machine; exemplary, the sensors include, but are not limited to, force sensors and acceleration sensors;
The usage data of the example is derived from accelerated life fault data { literature 1:Qing Li, Zhida Ren, A new fractional-order augmented quaternion-valued approach fordegradation prognostics of bearings using generalized Hamilton-real calculus, IEEE Transactions on Instrumentation and Measurement, 2022, 71, 3529711}, of the bearing in the university of east China and Hangzhou, as shown in figure 3, the accelerated life data of 4 channels of the bearing are acquired every 10min at the rotating speed of 4600 revolutions per minute, the sampling frequency is 12kHz, and the sampling time is 1s; it is known that, with time, the bearing goes from normal operation to failure, and the signal amplitude goes from stable amplitude to increasing amplitude.
The method for remotely reconstructing and jumping disturbance detection of the running data of the rotary machine further comprises the following step S102: preprocessing the training set data, detecting mutation points in the data, and obtaining preprocessed equipment operation data;
in step S102 of the invention, the 1 st channel bearing accelerated life data is selected as a research object, the data set is divided into 10 groups, each group is 0.1S (namely 1200 sampling points), the training set data is preprocessed, abrupt points (or wild points) in the data are detected, preprocessed equipment operation data are obtained, 1 group of data selected randomly are analyzed, as shown in FIG. 4, the preprocessed 1 st channel 5 th group bearing fault data are obtained, and the signal fault pulse interval is obvious;
The method for remotely reconstructing and jumping disturbance detection of the running data of the rotary machine further comprises the following step S103: constructing a condition generation countermeasure network model, classifying and marking health signals and fault signals of collected running data samples of the rotary mechanical equipment, and defining a network structure of a generator and a discriminator;
The method for remotely reconstructing and jumping disturbance detection of the running data of the rotary machine further comprises the following step S104: generating vibration signals through signal generators in a condition generation countermeasure network, adding the generated vibration signals into an offline data training set, calculating model gradients, carrying out iterative updating on the model, training to obtain a new generation countermeasure network model, and generating the vibration signals through the updated signal generators;
In step S104 of the present invention, a vibration signal is generated by a signal generator in a condition generation countermeasure network, the generated vibration signal is added into an offline data training set, a model gradient is calculated by using dlfeval and a function modelGradients, the model is iteratively updated, a new generation countermeasure network model is obtained by training, and the vibration signal is generated by an updated signal generator; fig. 5 shows a reconstructed vibration signal of the 5 th group of the bearing channel #1 generated by the countermeasure network, which is constructed by the method, and the waveform shape of the reconstructed vibration signal is similar to that of the original accelerated life vibration signal, and the fault pulse interval of the reconstructed signal is obvious.
The method for remotely reconstructing and jumping disturbance detection of the running data of the rotary machine further comprises the following steps:
Step S105: determining the reconstruction performance of the generated vibration signal by using a chaotic attractor method;
step S106: performing jump mutation disturbance detection on the obtained generated vibration signal;
Step S107: and predicting and tracking the running state of the equipment according to the predicted output result, so as to realize the predictive maintenance of the mechanical equipment.
In step S103 of the present invention, a condition generating countermeasure network model is constructed, and the collected running data samples of the rotating machinery device are marked with health signals and fault signals in a classification manner, so that the steps of defining the network structure of the generator and the discriminator are as follows:
Constructing a generator network and a discriminator network structure, and generating an countermeasure network model according to the composition conditions: assuming an actual dataset Independent co-distributed sampling at probability distribution/>Sample set generated by the generator/>The probability distribution of satisfaction is/>Training the generator, training samples are/>The probability distribution of satisfaction is/>The target function form is as follows:
;
wherein G is a minimisation maximum valuation function; generating a generator function of the countermeasure network model for the condition; /(I) Generating a minimum discriminant function of the countermeasure network model for the condition; pdata (x) real data distribution; PZ (z) is noise data distribution; /(I)Representing that the data x obeys a given real data distribution Pdata; /(I)Indicating that the data z obeys a noise data distribution PZ (z);
Preferably, in the formula (1), the iteration number is 100, the batch size is 256, the learning rate is 0.002, the model gradient attenuation coefficient is 0.5, and the square gradient attenuation coefficient is 0.999.
Marking a preprocessing signal of a mechanical equipment state, wherein a normal operation signal is marked as 1, a fault operation signal is marked as 2, inputting signal marking information into a condition generation countermeasure network model, and defining a generator and a discriminator network;
Inputting the given tag information and the running signal into a generator network to generate a signal of the same structure as training data corresponding to the same tag; training by alternately training a discriminator network and a generator network, namely training the discriminator for K times and training the generator for 1 time until the objective function converges or Nash equilibrium is reached; meanwhile, the authenticity of the training data and the data generated by the generator is judged by utilizing a discriminator network; and judging the training performance by comparing whether the game and the loss function of the generator and the arbiter are converged or not.
Further, in the embodiment of the present invention, the step of determining the reconstruction performance of the generated vibration signal by using the chaotic attractor method specifically includes:
constructing a Duffing chaotic attractor model:
;
wherein, For the input signal { document to be detected 2:J.M.T. Thompson, H.B. Stewart, Nonlinear dynamics and chaos, John Wiley&Sons, 2002. (2) R. Lifshitz, M.C. Cross, Nonlinear mechanics of nanomechanical and micromechanicalresonators, in reviews of nonlinear dynamics and complexity, Wiley, 2008. (3) M. J. Brennan, I. Kovacic, A. Carrella, T. P. Waters, On the jump-up andjump-down frequencies of the Duffing oscillator, Journal of Sound and Vibration, 318 (4-5) (2008) 1250-1261.},/>Is the first order derivative of the input signal to be detected,/>For second order differentiation of the input signal to be detected, parameter/>,/>,/>,/>And/>Is a constant, wherein the parameter/>Damping amount is controlled,/>Controlling the linear stiffness,/>Controlling the amount of non-linearity if/>=0, The Duffing oscillator will reduce to a simple harmonic oscillator, parameter/>And/>Amplitude and angular frequency for the input signal;
inputting an original real signal as a signal to be detected into a Duffing chaotic attractor model to obtain a Duffing chaotic attractor graph and a poincare section graph;
Inputting the generated signal as a signal to be detected into a Duffing chaotic attractor model to obtain a Duffing chaotic attractor graph and a poincare section graph;
And comparing the Duffing chaotic attractor graph generated in the two steps with the Poincare section graph, and generating a vibration signal to perform abrupt disturbance detection.
In the embodiment of the invention, input is an Input signal to be detected, and can be an original accelerated life vibration signal or a vibration signal after reconstruction; when any signal to be detected is used as an input signal, the Duffing chaotic attractor model can be used for chaotic attractor graph and Poincare section graph under the signal.
Specifically, as shown in fig. 7, which is a chaotic attractor graph and a poincare cross-sectional view of a channel #1 reconstructed vibration signal, comparing the Duffing chaotic attractor graph and the poincare cross-sectional view of fig. 6 and 7, it can be known that the Duffing chaotic attractor graph of the vibration signal reconstructed by the invention and the original channel #1 group 5 signal is similar to the poincare cross-sectional view, and the accuracy and the practicability of the reconstructed data of the invention are proved. Fig. 8 is a waveform diagram of the 5 th set of signals of the original channel #1 and a reconstructed fault signal diagram of the present invention in 0.04-0.06 seconds, and the accuracy of the reconstructed data of the present invention can be seen.
Further, in the embodiment of the present invention, the step of performing jump mutation disturbance detection on the obtained generated vibration signal includes:
Calculating a test statistical variable for generating a vibration signal For detecting the presence >Whether or not there is a generated vibration signal slave/>To/>Jump mutations of (a), namely:
;
wherein, For/>Generating vibration signal value at moment, the time variance of random fluctuation rate isStatistical variable/>Gradually obeying normal distribution with the average value of 0 and the standard deviation of pi/2; k is window size and satisfies/>,/>At the level of significance,/>Representation of arbitrary/>There is a finite constant/>So that/>,/>;
Selecting a level of significance;
Order theSo that it meets/>Wherein/>Is a cumulative distribution function, satisfies/>;
Calculation of,/>Wherein/>N is a sample number value;
If it is Then at/>Generating a vibration signal in a time period, wherein the jump mutation condition exists, otherwise, the jump mutation condition does not exist.
The present invention assumes, by way of example, a level of significance=0.01, { Document 3:S.S. Lee, P.A. Mykland, Jumps in financial markets: A new nonparametric test and jump dynamics, Review of Financial studies, 2008, 21(6):2535-2563};
Order theCalculation/>,/>Wherein/>N is a sample number value such that it satisfies:
;
i.e. threshold value ;
Judging if it isDescription of the reconstruction failure Signal at/>Generating a vibration signal in a time period, wherein the jump mutation condition exists, otherwise, the jump mutation condition does not exist.
In summary, according to the method for remotely reconstructing and jumping disturbance detection of mechanical equipment operation data, the sensor is used for collecting the rotating mechanical equipment operation data sample, and the rotating mechanical equipment operation data sample is regarded as real data and is divided into training set data and testing set data; preprocessing the training set data, detecting mutation points in the data, and obtaining preprocessed equipment operation data; constructing conditions to generate an countermeasure network model, defining a network structure of the generator and the discriminator, and carrying out iterative updating on the model to obtain a reconstructed vibration signal; determining the reconstruction performance of the generated vibration signal by using a chaotic attractor method; performing jump mutation disturbance detection on the obtained generated vibration signal; and predicting and tracking the running state of the equipment according to the predicted output result, and finally realizing the predictive maintenance of the mechanical equipment.
The invention can primarily solve the problems of less fault disclosure data, high acquisition cost and high experimental difficulty without considering the influence of working conditions and mechanical equipment structures, overcomes the defects of the existing manual inspection, can realize the on-line intelligent recognition of the state and the on-road state early warning, and has good algorithm instantaneity, high accuracy and strong generalization.
The remote reconstruction and jump disturbance detection method for the running data of the rotary machine has the advantages that:
firstly, the invention combines the deep learning and chaos monitoring technology, can primarily solve the problems of less fault disclosure data, high acquisition cost and great experimental difficulty, and researchers can generate the required simulation fault data more quickly with lower cost, so that the algorithm has better generalization;
secondly, the invention can realize the on-line intelligent recognition of the state and the on-way state early warning without considering the influence of the working condition and the mechanical equipment structure, and achieves the functions of real-time monitoring and the like, and has strong instantaneity and high accuracy;
Thirdly, the defects existing in the existing manual inspection are overcome, the hidden fault discovery capability and the early warning capability of the early fault generation stage are improved, and theoretical support can be provided for intelligent operation and maintenance of a complex electromechanical system;
Therefore, the invention provides a device operation data remote reconstruction and jump disturbance detection method by combining a deep learning technology, overcomes the defects that the traditional fault signals are difficult to acquire and the data value density is low, can detect whether the acquired signals are inaccurate in reconstruction and jump disturbance, utilizes the deep learning to carry out data enhancement on small sample data, solves the problem of jump disturbance of the whole-service operation fault signals of the mechanical equipment, and has important significance for improving the fault diagnosis and intelligent operation and maintenance of the mechanical equipment.
Although embodiments of the invention have been disclosed above, they are not limited to the use listed in the specification and embodiments. It can be applied to various fields suitable for the present invention. Additional modifications will readily occur to those skilled in the art. Therefore, the invention is not to be limited to the specific details and illustrations shown and described herein, without departing from the general concepts defined in the claims and their equivalents.
Claims (1)
1. The method for remotely reconstructing and jumping disturbance detection of the running data of the rotary machine is characterized by comprising the following steps:
collecting a running data sample of the rotary mechanical equipment, regarding the running data sample as real data, and dividing the real data into training set data and test set data;
Preprocessing the training set data, detecting mutation points in the data, and obtaining preprocessed equipment operation data;
constructing a condition generation countermeasure network model, classifying and marking health signals and fault signals of collected running data samples of the rotary mechanical equipment, and defining a network structure of a generator and a discriminator;
generating vibration signals through signal generators in a condition generation countermeasure network, adding the generated vibration signals into an offline data training set, calculating model gradients, carrying out iterative updating on the model, training to obtain a new generation countermeasure network model, and generating the vibration signals through the updated signal generators;
determining the reconstruction performance of the generated vibration signal by using a chaotic attractor method;
performing jump mutation disturbance detection on the obtained generated vibration signal;
carrying out prediction tracking on the running state of the equipment according to the prediction output result;
the construction condition generation countermeasure network model, the health signal and fault signal classification marking is carried out on the collected running data sample of the rotary mechanical equipment, and the steps of defining the network structure of the generator and the discriminator are as follows:
Constructing a generator network and a discriminator network structure, and generating an countermeasure network model according to the composition conditions: assuming an actual dataset Independent co-distributed sampling at probability distribution/>Sample set generated by the generator/>The probability distribution of satisfaction is/>Training the generator, training samples are/>The probability distribution of satisfaction isThe target function form is as follows:
;
wherein G is a minimisation maximum valuation function; generating a generator function of the countermeasure network model for the condition; /(I) Generating a minimum discriminant function of the countermeasure network model for the condition; pdata (x) real data distribution; PZ (z) is noise data distribution; /(I)Representing that the data x obeys a given real data distribution Pdata; /(I)Indicating that the data z obeys a noise data distribution PZ (z);
marking a preprocessing signal of a mechanical equipment state, wherein a normal operation signal is marked as 1, a fault operation signal is marked as 2, signal marking information is input into a condition generation countermeasure network model, and a generator and a discriminator network are defined at the same time;
Inputting the given tag information and the running signal into a generator network to generate a signal of the same structure as training data corresponding to the same tag; training by alternately training a discriminator network and a generator network, training the discriminator for K times, and training the generator for 1 time until the objective function converges or Nash equilibrium is reached; meanwhile, the authenticity of the training data and the data generated by the generator is judged by utilizing a discriminator network; whether the game and the loss function are converged or not through comparing the generator and the discriminator;
the step of determining the reconstruction performance of the generated vibration signal by using the chaotic attractor method specifically comprises the following steps:
constructing a Duffing chaotic attractor model:
;
wherein, For input signal to be detected,/>Is the first order derivative of the input signal to be detected,/>Second-order differentiation of the input signal to be detected;
inputting an original real signal as a signal to be detected into a Duffing chaotic attractor model to obtain a Duffing chaotic attractor graph and a poincare section graph;
Inputting the generated signal as a signal to be detected into a Duffing chaotic attractor model to obtain a Duffing chaotic attractor graph and a poincare section graph;
determining the reconstruction performance of the generated vibration signal by comparing the Duffing chaotic attractor graph with the Poincare section graph;
the step of performing jump mutation disturbance detection on the obtained generated vibration signal comprises the following steps:
Calculating a test statistical variable for generating a vibration signal For detecting the presence >Whether or not there is a vibration signal generation slave at any timeTo/>Is a jump mutation of (a):
wherein/> For/>Generating vibration signal value at moment, the time variance of random fluctuation rate isStatistical variable/>Gradually obeying normal distribution with the average value of 0 and the standard deviation of pi/2; k is window size and satisfies/>,/>In order to be a level of significance,Representation of arbitrary/>There is a finite constant/>So that/>,/>;
Selecting a level of significance; Find threshold/>So that it meets/>Wherein/>Is a cumulative distribution function, satisfies/>;
Calculation of,/>Wherein/>N is a sample number value;
If it is Then at/>Generating a vibration signal in a time period, wherein jump mutation exists; otherwise, no jump mutation condition exists.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410054151.9A CN117574114B (en) | 2024-01-15 | Remote reconstruction and jump disturbance detection method for running data of rotary machine |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410054151.9A CN117574114B (en) | 2024-01-15 | Remote reconstruction and jump disturbance detection method for running data of rotary machine |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117574114A CN117574114A (en) | 2024-02-20 |
CN117574114B true CN117574114B (en) | 2024-04-19 |
Family
ID=
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4287547A (en) * | 1978-08-29 | 1981-09-01 | Bbc Brown, Boveri & Co. Ltd. | Method and apparatus for fault and/or fault direction detection |
US8054717B1 (en) * | 2006-10-16 | 2011-11-08 | Marvell International Ltd. | Defect detection design |
CN108647786A (en) * | 2018-07-10 | 2018-10-12 | 电子科技大学 | The rotating machinery on-line fault monitoring method of neural network is fought based on depth convolution |
CN111914875A (en) * | 2020-06-05 | 2020-11-10 | 华南理工大学 | Fault early warning method of rotating machinery based on Bayesian LSTM model |
CN112417758A (en) * | 2020-11-18 | 2021-02-26 | 广东核电合营有限公司 | Diesel engine fault prediction method based on combined neural network |
CN113962381A (en) * | 2021-09-30 | 2022-01-21 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Weak signal data enhancement method |
CN115081484A (en) * | 2022-07-04 | 2022-09-20 | 南京航空航天大学 | Aircraft engine sensor fault diagnosis method based on CRJ-OSELM algorithm |
CN115345277A (en) * | 2022-08-05 | 2022-11-15 | 南京理工大学 | PMSM turn-to-turn short circuit diagnosis method based on time sequence generation countermeasure network |
CN116735206A (en) * | 2023-06-27 | 2023-09-12 | 大连海事大学 | Bearing friction state identification method based on attractor and convolutional neural network |
CN116910513A (en) * | 2023-07-14 | 2023-10-20 | 江苏大学 | Centrifugal pump rotor fault diagnosis method based on CWGAN-GP and double-flow CNN model |
CN116992386A (en) * | 2023-08-17 | 2023-11-03 | 武汉大学 | Multi-scale fusion and time-frequency countermeasure bearing fault signal generation method and system |
CN117290771A (en) * | 2023-09-22 | 2023-12-26 | 南京航空航天大学 | Rotary machine fault diagnosis method for generating countermeasure network based on improved auxiliary classification |
CN117332315A (en) * | 2023-09-28 | 2024-01-02 | 郑州轻工业大学 | Bearing fatigue period prediction method based on maximum entropy and generation countermeasure network |
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4287547A (en) * | 1978-08-29 | 1981-09-01 | Bbc Brown, Boveri & Co. Ltd. | Method and apparatus for fault and/or fault direction detection |
US8054717B1 (en) * | 2006-10-16 | 2011-11-08 | Marvell International Ltd. | Defect detection design |
CN108647786A (en) * | 2018-07-10 | 2018-10-12 | 电子科技大学 | The rotating machinery on-line fault monitoring method of neural network is fought based on depth convolution |
CN111914875A (en) * | 2020-06-05 | 2020-11-10 | 华南理工大学 | Fault early warning method of rotating machinery based on Bayesian LSTM model |
CN112417758A (en) * | 2020-11-18 | 2021-02-26 | 广东核电合营有限公司 | Diesel engine fault prediction method based on combined neural network |
CN113962381A (en) * | 2021-09-30 | 2022-01-21 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Weak signal data enhancement method |
CN115081484A (en) * | 2022-07-04 | 2022-09-20 | 南京航空航天大学 | Aircraft engine sensor fault diagnosis method based on CRJ-OSELM algorithm |
CN115345277A (en) * | 2022-08-05 | 2022-11-15 | 南京理工大学 | PMSM turn-to-turn short circuit diagnosis method based on time sequence generation countermeasure network |
CN116735206A (en) * | 2023-06-27 | 2023-09-12 | 大连海事大学 | Bearing friction state identification method based on attractor and convolutional neural network |
CN116910513A (en) * | 2023-07-14 | 2023-10-20 | 江苏大学 | Centrifugal pump rotor fault diagnosis method based on CWGAN-GP and double-flow CNN model |
CN116992386A (en) * | 2023-08-17 | 2023-11-03 | 武汉大学 | Multi-scale fusion and time-frequency countermeasure bearing fault signal generation method and system |
CN117290771A (en) * | 2023-09-22 | 2023-12-26 | 南京航空航天大学 | Rotary machine fault diagnosis method for generating countermeasure network based on improved auxiliary classification |
CN117332315A (en) * | 2023-09-28 | 2024-01-02 | 郑州轻工业大学 | Bearing fatigue period prediction method based on maximum entropy and generation countermeasure network |
Non-Patent Citations (3)
Title |
---|
A New Method for Long-Distance Data Reconstruction in Big Data Environment;Qing Li 等;《IEEE Xplore》;20191212;第1-12页 * |
Detectionofmicro-cracksusingnonlinearlambwavesbased on theDuffing-Holmessystem;Xiaofeng Liu 等;《ELSEVIER》;20171231;全文 * |
动作时序优化振动信号混沌吸引子特征的断 路器操动状态辨识方法;刘会兰;《高电压技术》;20231012;全文 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Saufi et al. | Low-speed bearing fault diagnosis based on ArSSAE model using acoustic emission and vibration signals | |
CN109858140B (en) | Fault diagnosis method for water chilling unit based on information entropy discrete Bayesian network | |
Sarmadi et al. | Probabilistic data self-clustering based on semi-parametric extreme value theory for structural health monitoring | |
CN107844067B (en) | A kind of gate of hydropower station on-line condition monitoring control method and monitoring system | |
Joshuva et al. | A comparative study of Bayes classifiers for blade fault diagnosis in wind turbines through vibration signals | |
Xu et al. | Quantile regression neural network‐based fault detection scheme for wind turbines with application to monitoring a bearing | |
CN113947017A (en) | Method for predicting residual service life of rolling bearing | |
CN115542099B (en) | Online GIS partial discharge detection method and device | |
CN111504647A (en) | AR-MSET-based performance degradation evaluation method for rolling bearing | |
Hartmann et al. | Coupling sensor-based structural health monitoring with finite element model updating for probabilistic lifetime estimation of wind energy converter structures | |
CN111881594B (en) | Non-stationary signal state monitoring method and system for nuclear power equipment | |
Márquez et al. | An overview of wind turbine maintenance management | |
Saidi et al. | Particle filter-based prognostic approach for high-speed shaft bearing wind turbine progressive degradations | |
Yan et al. | Advanced signal processing for structural health monitoring | |
CN117574114B (en) | Remote reconstruction and jump disturbance detection method for running data of rotary machine | |
Hu et al. | A new seismic‐based monitoring approach for wind turbines | |
CN117574114A (en) | Remote reconstruction and jump disturbance detection method for running data of rotary machine | |
US20230024104A1 (en) | Identification of false transformer humming using machine learning | |
Ye et al. | Evaluation of dynamic uncertainty of rolling bearing vibration performance | |
CN114547796B (en) | Ball mill feature fusion fault diagnosis method based on optimized BN network | |
CN116070368A (en) | Method for predicting residual life of high-speed bearing of offshore wind turbine | |
Zhou et al. | A fatigue life prediction method for the drive system of wind turbine using internet of things | |
CN114004059A (en) | Health portrait method for hydroelectric generating set | |
De Sitter et al. | Monitoring changes in the soil and foundation characteristics of an offshore wind turbine using automated operational modal analysis | |
CN117589444B (en) | Wind driven generator gear box fault diagnosis method based on federal learning |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant |