CN105628425A - Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine - Google Patents
Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine Download PDFInfo
- Publication number
- CN105628425A CN105628425A CN201610002290.2A CN201610002290A CN105628425A CN 105628425 A CN105628425 A CN 105628425A CN 201610002290 A CN201610002290 A CN 201610002290A CN 105628425 A CN105628425 A CN 105628425A
- Authority
- CN
- China
- Prior art keywords
- support vector
- vector machine
- rotating machinery
- kernel
- population
- 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.)
- Pending
Links
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 65
- 238000000137 annealing Methods 0.000 title claims abstract description 43
- 238000003745 diagnosis Methods 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000005457 optimization Methods 0.000 title claims abstract description 10
- 230000002068 genetic effect Effects 0.000 claims abstract description 35
- 230000008901 benefit Effects 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 39
- 238000000354 decomposition reaction Methods 0.000 claims description 20
- 230000002123 temporal effect Effects 0.000 claims description 13
- 239000013598 vector Substances 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 11
- 230000035772 mutation Effects 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 6
- 230000002759 chromosomal effect Effects 0.000 claims description 6
- 230000014759 maintenance of location Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 210000000349 chromosome Anatomy 0.000 claims description 3
- 238000007405 data analysis Methods 0.000 claims description 3
- 230000010429 evolutionary process Effects 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 108090000623 proteins and genes Proteins 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000002922 simulated annealing Methods 0.000 abstract description 5
- 230000015556 catabolic process Effects 0.000 abstract 1
- 238000006731 degradation reaction Methods 0.000 abstract 1
- 238000000605 extraction Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 6
- 239000000203 mixture Substances 0.000 description 4
- 238000003672 processing method Methods 0.000 description 4
- 238000002405 diagnostic procedure Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000001276 controlling effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 210000004940 nucleus Anatomy 0.000 description 2
- 241000272525 Anas platyrhynchos Species 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007850 degeneration Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000005469 granulation Methods 0.000 description 1
- 230000003179 granulation Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003447 ipsilateral effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000003211 malignant effect Effects 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000005476 size effect Effects 0.000 description 1
- 235000014347 soups Nutrition 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
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
- G01M99/00—Subject matter not provided for in other groups of this subclass
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
一种专门针对旋转机械早期故障的智能化诊断方法，以旋转机械运行过程中的振动信号为基础，首先将旋转机械的振动信号进行时域、频域、时频域信号处理，以典型的局部核函数和全局核函数为基础，构造多核支持向量机作为新型的智能诊断模型；以遗传算法和模拟退火算法为基础，构造遗传退火算法；利用遗传退火算法优化多核支持向量机的模型参数，实行多参数的并行寻优.本发明充分发挥了混合域特征集在旋转机械性能退化早期故障渐变特征提取、遗传退火算法在参数的并行寻优以及多核支持向量机在早期故障智能诊断等方面的优势，可以有效对旋转机械设备的早期故障进行诊断辨识，并且具有较强的抗干扰能力和泛化推广能力。
An intelligent diagnosis method for the early failure of rotating machinery, based on the vibration signal during the operation of the rotating machinery, firstly, the vibration signal of the rotating machinery is processed in the time domain, frequency domain, and time-frequency domain. Based on kernel function and global kernel function, construct multi-core support vector machine as a new type of intelligent diagnosis model; based on genetic algorithm and simulated annealing algorithm, construct genetic annealing algorithm; use genetic annealing algorithm to optimize the model parameters of multi-core support vector machine, implement Parallel optimization of multi-parameters. The present invention fully utilizes the advantages of hybrid domain feature set in early fault gradient feature extraction of rotating machinery performance degradation, genetic annealing algorithm in parallel optimization of parameters, and multi-core support vector machine in early fault intelligent diagnosis, etc. , can effectively diagnose and identify early faults of rotating machinery, and has strong anti-interference ability and generalization ability.
Description
Technical field
The present invention relates to the integral intelligent diagnostic method of a kind of rotating machinery Incipient Fault Diagnosis, particularly relate to a kind of rotating machinery initial failure integral intelligent diagnostic method optimizing multi-kernel support vector machine based on genetic Annealing.
Background technology
The initial failure of rotating machinery is the subject matter threatening rotating machinery safe and reliable operation; if able to the initial failure of rotating machinery is carried out reliable diagnosis; grasp the running status of rotating machinery in advance; the initial failure gradual change gradually that can avoid rotating machinery develops into typical fault, ultimately results in the malignant events such as rotating machinery shutdown suddenly and occurs. In a broad sense, rotating machinery Incipient Fault Diagnosis is for the rotating machinery being currently in operation, the running status that identification rotating machinery is current, finds out its source of trouble and trouble location, analyzes its fault cause, and proposes rational solution. Its research contents relates generally to mechanical kinetics, materialogy, plant equipment measuring technology, Engineering Signal treatment technology, computer technology etc., its main method is based on the theory of mechanical kinetics and materialogy, analyze the failure mechanism of rotating machinery, sensor is adopted to measure the operating state signal that rotating machinery is current, signal is analyzed process, and construct intellectualized algorithm the running status that rotating machinery is current is carried out diagnosis identification.
Tradition is in rotating machinery Incipient Fault Diagnosis process, and technical staff is first with the failure mechanism of the knowledge analysis rotating machinery of mechanical kinetics and materialogy, and utilizes various kinds of sensors to gather the operating state signal of rotating machinery, it is analyzed processing to all kinds of operating state signals obtained subsequently, extracts the principal character index of status signal, artificial intelligence, computer technology is finally utilized to set up fault diagnosis model, the value size according to characteristic index, infer the running status that rotating machinery is current. for rotating machinery, vibration signal is the most direct to the reflection of its running status, therefore, rotating machinery is usually employing vibration signal be analyzed, but the vibration signal intensity of reflection rotating machinery initial failure is relatively small, although fault features information can be collected by modern high-precision sensor, but it is affected by the interference of the environmental factorss such as signal route of transmission, propagation medium, is difficult to accurately extract the vibration performance of rotating machinery initial failure. in the building process of fault diagnosis model, what application was maximum at present is two classification method, k nearest neighbor sorting algorithm, artificial neural network (ArtificialNeuralNetwork, and support vector machine (SupportVectorMachine ANN), SVM) etc., two classification method, k nearest neighbor sorting algorithm is simple, but reliability is low, the reliability of ANN makes moderate progress, but study easily occurred or owes study and be absorbed in local minimum, SVM performance among this is relatively optimum, but parameter selects and the structure of kernel function also has certain randomness, therefore rationally reliable initial failure intelligent diagnostics model also not a duck soup is set up.
Present invention is generally directed to the retrofit that rotating machinery carries out based on the intelligent diagnostics algorithm of support vector machine. Carry out in failure diagnostic process in conventionally employed support vector machine, it is usually the direct support vector machine that the fault features index of acquisition is input to and carries out classification identification, structure for support vector machine, what generally adopt is monokaryon support vector machine, and the parameter for support vector machine generally adopts trial and error procedure or simple optimized algorithm to obtain. In the present invention, structure for support vector machine, on the basis of conventional single-core SVM, a kind of new machine learning model formed by being weighted all kinds of individual core merging, i.e. multinuclear hybrid supporting vector machine (Multi-kernelSupportVectorMachine, MSVM), MSVM inherits extensive Generalization Ability and the self-learning capability of conventional single-core SVM. Rotate in the diagnosis process of machinery initial failure utilizing MSVM, the kernel functional parameter of each individual core and the rationally selection of weight factor thereof will directly affect the identification performance of MSVM, by complex inheritance algorithm (GeneticAlgorithm, it is called for short GA) and simulated annealing (SimulatedAnnealing, it is called for short SA) respective advantage, construct genetic annealing algorithms (GeneticSimulatedAnnealingAlgorithm, it is called for short GA-SA), and then achieve the best to MSVM control parameter preferentially. The rotating machinery Incipient Fault Diagnosis method optimizing multi-kernel support vector machine based on genetic Annealing being consequently formed, the method has good capacity of resisting disturbance, and extensive Generalization Ability is strong, and the initial failure intelligent diagnostics for rotating machinery provides a kind of new approach.
Summary of the invention
Large rotating machinery comprises numerous transmission parts, including each type gear, bearing, axle system etc., once there is typical fault, not only results in the shutdown suddenly of rotating machinery, more seriously the personal safety of line operator can be produced great threat. The typical fault of rotating machinery is a process incremental, that progressively develop. If the commitment that can produce at rotating machinery fault, accomplish discovery as early as possible, as early as possible diagnosis, repair as early as possible, then being avoided that risk factor is accumulative transfinites. It can thus be seen that the Precise Diagnosis of the initial failure of rotating machinery is crucial. And the fault features of rotating machinery is very faint, both difficulty was effectively caught, again difficult effectively identification. Faint in order to solve rotating machinery fault features, the non-linear problem being difficult to identification between fault signature and fault mode, a kind of rotating machinery Incipient Fault Diagnosis method optimizing multi-kernel support vector machine based on genetic Annealing is proposed, be the technical scheme is that and adopt time domain, frequency domain, time-frequency domain signal processing method to extract the time domain in vibration signal, frequency domain, time and frequency domain characteristics index, this time domain, frequency domain, time and frequency domain characteristics index comprehensive are got up, constructs hybrid domain feature set; Adopt gaussian kernel and polynomial kernel to build multi-kernel support vector machine, and complex inheritance algorithm and simulated annealing build genetic annealing algorithms; Optimize all kinds of parameters of multi-kernel support vector machine with genetic annealing algorithms, obtain all kinds of optimized parameters of multi-kernel support vector machine; The multi-kernel support vector machine that the hybrid domain feature set of acquisition is input to genetic annealing algorithms optimization rotates the accurate recognition of machinery initial failure pattern. Specifically comprise the following steps that
Step one, gathers the vibrational state information in rotating machinery running by ICP acceleration transducer, subsequently this vibration signal is carried out noise reduction, filter preprocessing.
Step 2, at rotating machinery each vibration signal pretreated, calculating its time domain, frequency domain, time and frequency domain characteristics index value, constructs hybrid domain feature set.
Step 3, based on gaussian kernel support vector machine and Polynomial SVM, by Gauss karyomerite KRBFWith multinomial overall situation core KpolyCombine, construct multi-kernel support vector machine:
Kmix(xi,xj)=�� KRBF(xi,xj)+(1-��)Kpoly(xi,xj)
In formula: xiAnd xjFor the characteristic vector of the input space, the kernel functional parameter of �� and d respectively gaussian kernel function and Polynomial kernel function, �� (0 < �� < 1) is weight factor.
Step 4, heredity-annealingization algorithm optimization multi-kernel support vector machine, form chromosomal real number gene vectors X with the weight factor �� of multinuclear SVM, penalty parameter c, kernel functional parameter �� and d:
X=[��, c, ��, d]
Utilize multinuclear SVM to the classification accuracy f of hybrid domain feature set training sample to evaluate each chromosomal quality, adopt Metropolis acceptance criterion to determine new chromosome pij' whether enter population of future generation, in Metropolis acceptance criterion, adopt dynamic temperature-regulating strategy, if ti+1=(1-1/i)m��ti, wherein tiMoving back temperature parameter when evolving for i & lt, m �� Z is for controlling the decrease speed of temperature, in order to prevent the optimal solution of current population from losing in the next generation, adopts elite retention strategy, compares the optimal solution x in the fitness f of new explanation x and population*Fitness f*If f is better than f*, then x is replaced respectively with x and f*And f*��
Step 5, is input to the multi-kernel support vector machine after optimization by hybrid domain feature set, utilizes optimum multi-kernel support vector machine automatically to pick out the fault mode of rotating machinery initial failure.
In described step 2, the vibration characteristic signals that rotating machinery produces in running is the Main Basis of reflection running state of rotating machine, by calculating the time domain of vibration signal, frequency domain character index value, can the appearance of quantitatively characterizing rotating machinery initial failure, time domain, frequency domain character index are a lot, at this data analysis by experiment, have chosen and bearing initial failure is characterized obvious 6 temporal signatures indexs and 2 frequency domain character indexs; Described temporal signatures index includes root amplitude, root-mean-square value, variance, average amplitude, kurtosis and maximum; Described frequency domain character index includes means frequency and tune spread;
In order to obtain more fault characteristic information, reflect the initial failure state of rotating machinery comprehensively, also adopt integrated empirical mode decomposition (ensembleempiricalmodedecomposition, EEMD) extracting the time and frequency domain characteristics index in original vibration signal, vibration signal decomposes, through EEMD, the final result obtained and is:
Wherein: cjT () is original vibration signal x (t) limited the Intrinsic Mode component (Intrinsicmodefunction, IMF) obtained from high frequency to low frequency adaptive decomposition, r (t) is the residual components after decomposing. When rotating machinery breaks down, the frequency band distribution of vibration signal and spectrum energy all can change, and at this, first the vibration signal of rotating machinery carried out EEMD decomposition, obtain n IMF component; Secondly the energy T={E of each IMF component is calculated1,E2,��,En, namely
The IMF component energy feature obtained is normalized, i.e. T '={ E1/E,E2/E,��,En/ E}, whereinBy being analyzed above it can be seen that contain 6 temporal signatures, 2 frequency domain characters and n IMF component energy feature in the hybrid domain feature set of rotating machinery Incipient Fault Diagnosis, wherein n takes the integer between 5��10.
Adopting karyomerite function and overall situation Kernel multi-kernel support vector machine in described step 3, gaussian kernel function is typical karyomerite function, describes as follows:
KRBF(xi,xj)=exp (-| | xi-xj||2/��2)
Polynomial kernel function is typical overall situation kernel function, describes as follows:
Karyomerite function learning ability is strong, and generalization ability is weak; Overall situation kernel function generalization ability is strong, learning capacity is weak, traditional SVM adopts single core function, can effectively solve the problem that the classification identification problem of simple data, but the rotating machinery Incipient Fault Diagnosis problem for multiple complex data sources but has some limitations, in order to make SVM have better learning capacity and Generalization Capability, it is proposed to by karyomerite KRBFWith overall situation core KpolyCombine, construct multi-kernel support vector machine: Kmix(xi,xj), multi-kernel support vector machine has had the advantage of each individual core concurrently, has better classification identification performance.
Optimizing multi-kernel support vector machine detailed process based on heredity-annealing algorithm in step 4 is, is first initialize to control parameter, and arranging population scale is S, and independent evolutionary generation is N, and maximum evolutionary generation is M, and the crossover probability of population is p1, mutation probability is p2, initial annealing temperature is T0, overall situation evolutionary generation variable l1=0, randomly generate initial population;
According to the MSVM test result to training sample, evaluate the fitness of each individuality of current population, based on elite retention strategy, currently most individuality is deposited in memory apparatus; If optimum individual meets the condition of convergence in current population, then evolutionary process terminates, and returns global optimum individual, otherwise initializes Local Evolution algebraic variable l2=0;
For determining annealing temperature TiUnder population, in order to produce a new population, it is necessary to randomly select individual x from current populationi,xj, by crossover probability p1Carry out the real-valued operation that intersects, produce two new individual xi',x'j, calculate fitness f (xi'),f(x'j), select according to Metropolis acceptance criterion new individual, update memory apparatus simultaneously; Individuality after intersecting is pressed mutation probability p2Carry out real-valued mutation operation, calculate the fitness that variation is individual, select according to Metropolis acceptance criterion new individual, update memory apparatus simultaneously; l1=l1+ 1, l2=l2+ 1, if l2< N, then continue to produce new population;
If l1< M, is now not reaching to maximum iteration time, then revise the annealing temperature T of populationi, start iteration again, otherwise compare current population optimal solution and the value in memory apparatus, export optimal solution.
Optimize the characteristic vector of multi-kernel support vector machine as genetic Annealing after extracting the hybrid domain feature set normalized characterizing rotating machinery initial failure state; Utilize genetic Annealing to optimize multi-kernel support vector machine training sample is trained, obtain the optimal control parameter vector of multi-kernel support vector machine; Test sample is carried out identification diagnosis by the multi-kernel support vector machine Reconstruction initial failure intelligent diagnostics model finally utilizing optimum, exports final diagnostic result.
The present invention, based on the rolling bearing performance degeneration fuzzy granulation Forecasting Methodology of comentropy, adopts technique scheme can obtain following technique effect:
1) integrated application time domain, frequency domain, time-frequency domain signal processing method, never ipsilateral characterizes the time domain of running state of rotating machine, frequency domain, time-frequency domain statistical nature, it is possible to more accurately, more fully get the initial failure Weak characteristic of rotating machinery.
2) multi-kernel support vector machine is applied to the Incipient Fault Diagnosis of rotating machinery as the Incipient Fault Diagnosis model of a kind of superior performance, and the various coupling faults of rotating machinery can be made more identification conclusion accurately by it.
3) genetic annealing algorithms overcomes the premature problem of traditional genetic algorithm, solves the select permeability of multi-kernel support vector machine initial parameter, has higher versatility, robustness and degree of accuracy.
In sum, the present invention optimizes the rotating machinery Incipient Fault Diagnosis method of multi-kernel support vector machine based on genetic Annealing, analyse in depth the Major Difficulties in traditional rotating machinery Incipient Fault Diagnosis method, combine time domain, frequency domain, time-frequency domain signal processing method fault signature in early days extracts, multi-kernel support vector machine intelligent fault diagnosis in early days and genetic annealing algorithms advantage in multiparameter parallel optimization etc., define the integrated rotating machinery Incipient Fault Diagnosis pattern built by " the multi-kernel support vector machine fault diagnosis that fault features is extracted �� optimized ", maintain again the high accuracy in engineer applied simultaneously, high efficiency, the requirements such as replicability, a kind of new approach is provided for rotating machinery initial failure intelligent diagnostics.
Accompanying drawing explanation
Accompanying drawing 1 is the operation principle schematic diagram that in step 4 of the present invention, genetic Annealing optimizes multi-kernel support vector machine.
Accompanying drawing 2 is the operation principle schematic diagram that the present invention optimizes the rotating machinery Incipient Fault Diagnosis method of multi-kernel support vector machine based on genetic Annealing.
Detailed description of the invention
Below in conjunction with accompanying drawing 2, the embodiment of the rotating machinery initial failure intelligent diagnostics that the present invention optimizes multi-kernel support vector machine based on genetic Annealing is elaborated. The main purpose of the present embodiment is to be extracted by the Weak characteristic of initial failure in characteristic of rotating machines vibration signal by time domain, frequency domain, time-frequency domain signal processing method, multi-kernel support vector machine model is constructed based on Radial basis kernel function and Polynomial kernel function, genetic annealing algorithms is built based on genetic algorithm and simulated annealing, optimize multi-kernel support vector machine with genetic annealing algorithms, and then realize the Intelligence Diagnosis of rotating machinery initial failure. Embodiment comprises the following specific steps that:
Step one, gathers the vibrational state information in rotating machinery running by ICP acceleration transducer, subsequently this vibration signal is carried out noise reduction, filter preprocessing.
Described ICP sensor is the piezoelectric acceleration sensor of built-in miniature IC amplifier, traditional piezoelectric acceleration sensor and charge amplifier are integrated in one by it, directly can connect with record, display and collecting device, simplify test system, improve measuring accuracy and reliability. It is widely used in the fields such as nuclear explosion, Aero-Space, railway, bridge, building, car and boat, machinery, water conservancy, electric power, oil, chemical industry, environmental protection, earthquake.
Step 2, to at rotating machinery each vibration signal pretreated, calculate its temporal signatures index (subordinate list 1), frequency domain character index (subordinate list 2), time and frequency domain characteristics index respectively, wherein time and frequency domain characteristics index adopts integrated empirical mode decomposition technology (EEMD) to build by calculating the energy of mode function (IMF) component in each layer, gets up to build hybrid domain feature set by temporal signatures index (subordinate list 1), frequency domain character index (subordinate list 2), time and frequency domain characteristics index comprehensive;
Further, the vibration characteristic signals that rotating machinery produces in running is the Main Basis of reflection running state of rotating machine, and vibration signal can produce different characteristic parameter values to characterize the wave character of vibration signal at the statistical analysis of time domain and frequency domain. Temporal signatures value therein is the original foundation of rotary machinery fault diagnosis; Frequency domain character value is by carrying out spectrum transformation to original vibration signal, it is possible to extract more characteristic information. Therefore, by calculating the time domain of vibration signal, frequency domain character index value, it is possible to the appearance of quantitatively characterizing rotating machinery initial failure. Time domain, frequency domain character index are a lot, at this data analysis by experiment, have chosen and rotating machinery initial failure is characterized obvious 6 temporal signatures indexs and 2 frequency domain character indexs, as shown in Table 1 and Table 2, in practical engineering application, it is possible to add time domain, frequency domain character index flexibly according to practical situation.
Table 1 temporal signatures index
Table 2 frequency domain character index
Temporal signatures index in table 1 mainly reflects vibration signal amplitude energy size and seasonal effect in time series distribution situation; Frequency domain character index in table 2 mainly reflects the change of vibration signal main band position and the dispersion intensity of frequency spectrum. In order to obtain more fault characteristic information, reflect the initial failure state of rotating machinery comprehensively, integrated empirical mode decomposition (ensembleempiricalmodedecomposition, EEMD) is also adopted to extract the time and frequency domain characteristics index in original vibration signal.
Further, described integrated empirical mode decomposition (EEMD) is a kind of signal decomposition method based on data-driven, it is at empirical mode decomposition (empiricalmodedecomposition, EMD) on basis, by superposition white Gaussian noise, to reduce the modal overlap situation in signal decomposition process, the decomposition step of EEMD is as follows:
(1) in primary signal x (t), superposition average is 0, and standard deviation is the white Gaussian noise n of constantiT (), obtains signal x to be decomposedi(t):
xi(t)=x (t)+ni(t)
Wherein: xiT () is the signal after adding i & lt white Gaussian noise. The white noise n addediT () is too big or too little all can not effectively solve mode mixing problem compared with original vibration signal x (t), the too little change that will not cause EMD decomposition result of white noise ratio, then can make to decompose the harmonic components permeating some falsenesses in each IMF component obtained too greatly, under normal circumstances, white noise takes 0.1��0.4 times that standard deviation is primary signal standard deviation.
(2) to xiT () carries out empirical mode decomposition, obtain several IMF components | cij(t) and remainder ri(t), wherein cijT () decomposes, after representing that i & lt adds white Gaussian noise, the i-th IMF obtained.
(3) repeating M step (1), step (2), obtained IMF component carries out ensemble average, eliminate and repeatedly add the white noise impact on decomposition result, obtaining the EEMD IMF result finally decomposed is:
In formula: perform the EMD number of times M decomposed more big, be more conducive to the raising of Decomposition Accuracy under normal circumstances, but when M big to a certain extent time, the improvement of result is also extremely limited, at this through rotating machinery vibrating status signal is analysed in depth, take M=100.
(4) final result that vibration signal obtains through EEMD decomposition is:
Wherein: cjT () is original vibration signal x (t) limited the Intrinsic Mode component (Intrinsicmodefunction, IMF) obtained from high frequency to low frequency adaptive decomposition, r (t) is the residual components after decomposing.
When initial failure occurs in rotating machinery, the frequency band distribution of vibration signal and spectrum energy all can change, and at this, on the EEMD basis decomposed, first the vibration signal of rotating machinery carried out EEMD decomposition, obtain n IMF component; Secondly the energy T={E of each IMF component is calculated1,E2,��,En, namely
The IMF component energy feature obtained is normalized, i.e. T '={ E1/E,E2/E,��,En/ E}, whereinBy being analyzed above it can be seen that contain 6 temporal signatures, 2 frequency domain characters and n IMF component energy feature in the hybrid domain feature set of rotating machinery Incipient Fault Diagnosis, wherein n generally takes the integer between 5��10.
Step 3, based on gaussian kernel support vector machine and Polynomial SVM, by Gauss karyomerite KRBFWith multinomial overall situation core KpolyCombine, construct multi-kernel support vector machine:
Kmix(xi,xj)=�� KRBF(xi,xj)+(1-��)Kpoly(xi,xj)
In formula: xiAnd xjFor the characteristic vector of the input space, the kernel functional parameter of �� and d respectively gaussian kernel function and Polynomial kernel function, �� (0 < �� < 1) is weight factor.
Further, utilizing SVM to carry out in fault identification process, the selection of kernel function is most important, the discriminant function that different kernel functions is corresponding different, and then directly affects the identification precision of SVM. The kernel function of SVM mainly includes local kernels and global kernels, and gaussian kernel function is typical karyomerite function, describes as follows:
KRBF(xi,xj)=exp (-| | xi-xj||2/��2)
Polynomial kernel function is typical overall situation kernel function, describes as follows:
Karyomerite function learning ability is strong, and generalization ability is weak; Overall situation kernel function generalization ability is strong, and learning capacity is weak. Traditional SVM adopts single core function, it is possible to effectively solve the classification identification problem of simple data, but the rotating machinery Incipient Fault Diagnosis problem for multiple complex data sources but has some limitations. In order to make SVM have better learning capacity and Generalization Capability, it is proposed to by karyomerite KRBFWith overall situation core KpolyCombine, construct mixed nucleus SVM:
Kmix(xi,xj)=�� KRBF(xi,xj)+(1-��)Kpoly(xi,xj)
In formula: xiAnd xjFor the characteristic vector of the input space, the kernel functional parameter of �� and d respectively gaussian kernel function and Polynomial kernel function, �� (0 < �� < 1) is for regulating parameter. Mixed nucleus SVM has had the advantage of each individual core concurrently, has better classification identification performance.
Step 4, genetic Annealing algorithm optimization multi-kernel support vector machine, form chromosomal real number gene vectors X with the weight factor �� of multinuclear SVM, penalty parameter c, kernel functional parameter �� and d:
X=[��, c, ��, d]
Utilize multinuclear SVM to the classification accuracy f of hybrid domain feature set training sample to evaluate each chromosomal quality, adopt Metropolis acceptance criterion to determine new chromosome pij' whether enter population of future generation, in Metropolis acceptance criterion, adopt dynamic temperature-regulating strategy, if ti+1=(1-1/i)m��ti, wherein tiMoving back temperature parameter when evolving for i & lt, m �� Z is for controlling the decrease speed of temperature, in order to prevent the optimal solution of current population from losing in the next generation, adopts elite retention strategy, compares the optimal solution x in the fitness f of new explanation x and population*Fitness f*If f is better than f*, then x is replaced respectively with x and f*And f*��
Further, MSVM parameter algorithm flow process is optimized as shown in Figure 1 based on genetic annealing algorithms:
Specifically comprise the following steps that
(1) control parameter is initialized: arranging population scale is S, and independent evolutionary generation is N, and maximum evolutionary generation is M, and the crossover probability of population is p1, mutation probability is p2, initial annealing temperature is T0, overall situation evolutionary generation variable l1=0, randomly generate initial population;
(2) according to the MSVM test result to training sample, evaluate the fitness of each individuality of current population, based on elite retention strategy, currently most individuality is deposited in memory apparatus;
(3) if optimum individual meets the condition of convergence in current population, then evolutionary process terminates, and returns global optimum individual, otherwise initializes Local Evolution algebraic variable l2=0;
(4) for determining annealing temperature TiUnder population implement the following, to produce a new population:
I. from current population, randomly select individual xi,xj, by crossover probability p1Carry out the real-valued operation that intersects, produce two new individual xi',x'j, calculate fitness f (xi'),f(x'j), select according to Metropolis acceptance criterion new individual, update memory apparatus simultaneously;
Ii. the individuality after intersecting is pressed mutation probability p2Carry out real-valued mutation operation, calculate the fitness that variation is individual, select according to Metropolis acceptance criterion new individual, update memory apparatus simultaneously.
(5)l1=l1+ 1, l2=l2+ 1, if l2< N, then go to step (4), otherwise continue;
(6) if l1< M then revises the annealing temperature T of populationi, go to step (2), otherwise compare current population optimal solution and the value in memory apparatus, export optimal solution.
So far, the overall process of rotating machinery Incipient Fault Diagnosis is namely completed.
The present invention is not restricted to the described embodiments, and every utilizes principles of the invention and mode, through the technical scheme that conversion and replacement are formed, all in protection scope of the present invention.
Claims (5)
1. the rotating machinery Incipient Fault Diagnosis method optimizing multi-kernel support vector machine based on genetic Annealing, it is characterised in that: comprise the following specific steps that:
Step one, gathers the vibrational state information in rotating machinery running by ICP acceleration transducer, subsequently this vibration signal is carried out noise reduction, filter preprocessing;
Step 2, at rotating machinery each vibration signal pretreated, calculating its time domain, frequency domain, time and frequency domain characteristics index value, constructs hybrid domain feature set;
Step 3, based on gaussian kernel support vector machine and Polynomial SVM, by Gauss karyomerite KRBFWith multinomial overall situation core KpolyCombine, construct multi-kernel support vector machine:
Kmix(xi,xj)=�� KRBF(xi,xj)+(1-��)Kpoly(xi,xj)
In formula: xiAnd xjFor the characteristic vector of the input space, the kernel functional parameter of �� and d respectively gaussian kernel function and Polynomial kernel function, �� (0 < �� < 1) is weight factor;
Step 4, genetic Annealing algorithm optimization multi-kernel support vector machine, form chromosomal real number gene vectors X with the weight factor �� of multinuclear SVM, penalty parameter c, kernel functional parameter �� and d:
X=[��, c, ��, d]
Utilize multinuclear SVM to the classification accuracy f of hybrid domain feature set training sample to evaluate each chromosomal quality, adopt Metropolis acceptance criterion to determine new chromosome pij' whether enter population of future generation, in Metropolis acceptance criterion, adopt dynamic temperature-regulating strategy, if ti+1=(1-1/i)m��ti, wherein tiMoving back temperature parameter when evolving for i & lt, m �� Z is for controlling the decrease speed of temperature, in order to prevent the optimal solution of current population from losing in the next generation, adopts elite retention strategy, compares the optimal solution x in the fitness f of new explanation x and population*Fitness f*If f is better than f*, then x is replaced respectively with x and f*And f*;
Step 5, is input to the multi-kernel support vector machine after optimization by hybrid domain feature set, utilizes optimum multi-kernel support vector machine automatically to pick out the fault mode of rotating machinery initial failure.
2. optimize the rotating machinery Incipient Fault Diagnosis method of multi-kernel support vector machine based on genetic Annealing according to claim 1, it is characterized in that: in described step 2, the vibration characteristic signals that rotating machinery produces in running is the Main Basis of reflection running state of rotating machine, by calculating the time domain of vibration signal, frequency domain character index value, can the appearance of quantitatively characterizing rotating machinery initial failure, time domain, frequency domain character index is a lot, at this data analysis by experiment, have chosen and bearing initial failure is characterized obvious 6 temporal signatures indexs and 2 frequency domain character indexs, described temporal signatures index includes root amplitude, root-mean-square value, variance, average amplitude, kurtosis and maximum, described frequency domain character index includes means frequency and tune spread,
In order to obtain more fault characteristic information, reflect the initial failure state of rotating machinery comprehensively, also adopt integrated empirical mode decomposition (ensembleempiricalmodedecomposition, EEMD) extracting the time and frequency domain characteristics index in original vibration signal, vibration signal decomposes, through EEMD, the final result obtained and is:
Wherein: cjT () is original vibration signal x (t) limited the Intrinsic Mode component (Intrinsicmodefunction, IMF) obtained from high frequency to low frequency adaptive decomposition, r (t) is the residual components after decomposing. When rotating machinery breaks down, the frequency band distribution of vibration signal and spectrum energy all can change, and at this, first the vibration signal of rotating machinery carried out EEMD decomposition, obtain n IMF component; Secondly the energy T={E of each IMF component is calculated1, E2..., En, namely
The IMF component energy feature obtained is normalized, i.e. T '={ E1/ E, E2/ E ..., En/ E}, whereinBy being analyzed above it can be seen that contain 6 temporal signatures, 2 frequency domain characters and n IMF component energy feature in the hybrid domain feature set of rotating machinery Incipient Fault Diagnosis, wherein n takes the integer between 5��10.
3. the rotating machinery Incipient Fault Diagnosis method optimizing multi-kernel support vector machine based on genetic Annealing according to claim 1, it is characterized in that: described step 3 adopts karyomerite function and overall situation Kernel multi-kernel support vector machine, gaussian kernel function is typical karyomerite function, describes as follows:
KRBF(xi,xj)=exp (-| | xi-xj||2/��2)
Polynomial kernel function is typical overall situation kernel function, describes as follows:
Karyomerite function learning ability is strong, and generalization ability is weak; Overall situation kernel function generalization ability is strong, learning capacity is weak, traditional SVM adopts single core function, can effectively solve the problem that the classification identification problem of simple data, but the rotating machinery Incipient Fault Diagnosis problem for multiple complex data sources but has some limitations, in order to make SVM have better learning capacity and Generalization Capability, it is proposed to by karyomerite KRBFWith overall situation core KpolyCombine, construct multi-kernel support vector machine: Kmix(xi,xj), multi-kernel support vector machine has had the advantage of each individual core concurrently, has better classification identification performance.
4. the rotating machinery Incipient Fault Diagnosis method optimizing multi-kernel support vector machine based on genetic Annealing according to claim 1, it is characterized in that: optimizing multi-kernel support vector machine detailed process based on genetic annealing algorithms in step 4 is, first it is initialize to control parameter, arranging population scale is S, independent evolutionary generation is N, maximum evolutionary generation is M, and the crossover probability of population is p1, mutation probability is p2, initial annealing temperature is T0, overall situation evolutionary generation variable l1=0, randomly generate initial population;
According to the MSVM test result to training sample, evaluate the fitness of each individuality of current population, based on elite retention strategy, currently most individuality is deposited in memory apparatus; If optimum individual meets the condition of convergence in current population, then evolutionary process terminates, and returns global optimum individual, otherwise initializes Local Evolution algebraic variable l2=0;
For determining annealing temperature TiUnder population, in order to produce a new population, it is necessary to randomly select individual x from current populationi,xj, by crossover probability p1Carry out the real-valued operation that intersects, produce two new individual x 'i,x��j, calculate fitness f (x 'i),f(x��j), select according to Metropolis acceptance criterion new individual, update memory apparatus simultaneously; Individuality after intersecting is pressed mutation probability p2Carry out real-valued mutation operation, calculate the fitness that variation is individual, select according to Metropolis acceptance criterion new individual, update memory apparatus simultaneously; l1=l1+ 1, l2=l2+ 1, if l2< N, then continue to produce new population;
If l1< M, is now not reaching to maximum iteration time, then revise the annealing temperature T of populationi, start iteration again, otherwise compare current population optimal solution and the value in memory apparatus, export optimal solution.
5. the rotating machinery Incipient Fault Diagnosis method optimizing multi-kernel support vector machine based on genetic Annealing according to claim 1, it is characterised in that: the characteristic vector of multi-kernel support vector machine is optimized after extracting the hybrid domain feature set normalized characterizing rotating machinery initial failure state as genetic Annealing; Utilize genetic Annealing to optimize multi-kernel support vector machine training sample is trained, obtain the optimal control parameter vector of multi-kernel support vector machine; Test sample is carried out identification diagnosis by the multi-kernel support vector machine Reconstruction initial failure intelligent diagnostics model finally utilizing optimum, exports final diagnostic result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610002290.2A CN105628425A (en) | 2016-01-01 | 2016-01-01 | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610002290.2A CN105628425A (en) | 2016-01-01 | 2016-01-01 | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105628425A true CN105628425A (en) | 2016-06-01 |
Family
ID=56043565
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610002290.2A Pending CN105628425A (en) | 2016-01-01 | 2016-01-01 | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105628425A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106897703A (en) * | 2017-02-27 | 2017-06-27 | 辽宁工程技术大学 | Remote Image Classification based on AGA PKF SVM |
| CN107247931A (en) * | 2017-05-27 | 2017-10-13 | 哈尔滨理工大学 | A kind of De-noising Technique For Vibration Signals of Rotating Machinery |
| CN107300588A (en) * | 2017-06-23 | 2017-10-27 | 陕西师范大学 | A kind of PSO SVM optimization methods that the fusion of acoustical signal multiple domain is collided based on iblet |
| CN107300856A (en) * | 2017-06-30 | 2017-10-27 | 哈尔滨理工大学 | A kind of rotating machinery method for predicting residual useful life based on FDA and SVDD |
| CN107315996A (en) * | 2017-05-19 | 2017-11-03 | 燕山大学 | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA |
| CN107977679A (en) * | 2017-12-06 | 2018-05-01 | 中北大学 | A Method for Early Fault Diagnosis of Complex Devices Based on Frequency Response Function and Operational Response Characteristics |
| CN108062564A (en) * | 2017-12-12 | 2018-05-22 | 内蒙古科技大学 | A kind of optimization multinuclear multiple features fusion support vector machines knows method for distinguishing for bearing fault |
| CN108446629A (en) * | 2018-03-19 | 2018-08-24 | 河北工业大学 | Rolling Bearing Fault Character extracting method based on set empirical mode decomposition and modulation double-spectrum analysis |
| CN109029993A (en) * | 2018-06-20 | 2018-12-18 | 中国计量大学 | In conjunction with the bearing fault detection algorithm of genetic algorithm optimization parameter and machine vision |
| CN110032981A (en) * | 2019-04-19 | 2019-07-19 | 电子科技大学 | Based on the rotating machinery fault recognition methods for improving support vector machines |
| CN110767018A (en) * | 2019-11-05 | 2020-02-07 | 中电科仪器仪表有限公司 | Microwave teaching system for automatically configuring parameters and working method thereof |
| CN110966100A (en) * | 2018-09-30 | 2020-04-07 | 中国航发商用航空发动机有限责任公司 | Combustion oscillation monitoring device and method |
| CN111721835A (en) * | 2020-06-28 | 2020-09-29 | 上海理工大学 | An intelligent monitoring method for the condition of the grinding wheel of hollow drill |
| CN112037289A (en) * | 2020-09-16 | 2020-12-04 | 安徽意欧斯物流机器人有限公司 | Off-line parameter calibration method based on genetic algorithm |
| CN112785003A (en) * | 2021-01-28 | 2021-05-11 | 武汉市博畅软件开发有限公司 | Method and device for controlling electrically tunable filter |
| CN114553671A (en) * | 2022-02-28 | 2022-05-27 | 国家电网有限公司 | Diagnosis method for power communication network fault alarm |
| CN114622954A (en) * | 2022-05-16 | 2022-06-14 | 合肥派光感知信息技术有限公司 | Illumination and monitoring integrated tunnel safety monitoring system and method |
| CN118670661A (en) * | 2024-08-13 | 2024-09-20 | 南京易信同控制设备科技有限公司 | Fault diagnosis method and system for movable steel structure facility |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6408696B1 (en) * | 1999-12-06 | 2002-06-25 | Ai Signal Research, Inc. | Coherent phase line enhancer spectral analysis technique |
| CN101587017A (en) * | 2009-06-19 | 2009-11-25 | 湖南大学 | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum |
| CN102829974A (en) * | 2012-08-07 | 2012-12-19 | 北京交通大学 | LMD (local mean decomposition) and PCA (principal component analysis) based rolling bearing state identification method |
-
2016
- 2016-01-01 CN CN201610002290.2A patent/CN105628425A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6408696B1 (en) * | 1999-12-06 | 2002-06-25 | Ai Signal Research, Inc. | Coherent phase line enhancer spectral analysis technique |
| CN101587017A (en) * | 2009-06-19 | 2009-11-25 | 湖南大学 | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum |
| CN102829974A (en) * | 2012-08-07 | 2012-12-19 | 北京交通大学 | LMD (local mean decomposition) and PCA (principal component analysis) based rolling bearing state identification method |
Non-Patent Citations (3)
| Title |
|---|
| 戴豪民等: "基于WMRMR的滚动轴承混合域特征选择方法", 《振动与冲击》 * |
| 陈法法等: "基于遗传退火优化MSVM的齿轮箱故障诊断", 《振动、测试与诊断》 * |
| 高昌鑫等: "基于EEMD的齿轮箱故障诊断", 《煤炭技术》 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106897703A (en) * | 2017-02-27 | 2017-06-27 | 辽宁工程技术大学 | Remote Image Classification based on AGA PKF SVM |
| CN107315996A (en) * | 2017-05-19 | 2017-11-03 | 燕山大学 | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA |
| CN107247931A (en) * | 2017-05-27 | 2017-10-13 | 哈尔滨理工大学 | A kind of De-noising Technique For Vibration Signals of Rotating Machinery |
| CN107300588A (en) * | 2017-06-23 | 2017-10-27 | 陕西师范大学 | A kind of PSO SVM optimization methods that the fusion of acoustical signal multiple domain is collided based on iblet |
| CN107300856A (en) * | 2017-06-30 | 2017-10-27 | 哈尔滨理工大学 | A kind of rotating machinery method for predicting residual useful life based on FDA and SVDD |
| CN107977679A (en) * | 2017-12-06 | 2018-05-01 | 中北大学 | A Method for Early Fault Diagnosis of Complex Devices Based on Frequency Response Function and Operational Response Characteristics |
| CN107977679B (en) * | 2017-12-06 | 2021-07-30 | 中北大学 | A method for diagnosing early faults of complex devices based on frequency response function and operating response characteristics |
| CN108062564A (en) * | 2017-12-12 | 2018-05-22 | 内蒙古科技大学 | A kind of optimization multinuclear multiple features fusion support vector machines knows method for distinguishing for bearing fault |
| CN108062564B (en) * | 2017-12-12 | 2021-07-09 | 内蒙古科技大学 | A method of optimizing multi-kernel and multi-feature fusion support vector machine for bearing fault identification |
| CN108446629A (en) * | 2018-03-19 | 2018-08-24 | 河北工业大学 | Rolling Bearing Fault Character extracting method based on set empirical mode decomposition and modulation double-spectrum analysis |
| CN109029993A (en) * | 2018-06-20 | 2018-12-18 | 中国计量大学 | In conjunction with the bearing fault detection algorithm of genetic algorithm optimization parameter and machine vision |
| CN110966100A (en) * | 2018-09-30 | 2020-04-07 | 中国航发商用航空发动机有限责任公司 | Combustion oscillation monitoring device and method |
| CN110966100B (en) * | 2018-09-30 | 2021-03-05 | 中国航发商用航空发动机有限责任公司 | Combustion oscillation monitoring device and method |
| CN110032981A (en) * | 2019-04-19 | 2019-07-19 | 电子科技大学 | Based on the rotating machinery fault recognition methods for improving support vector machines |
| CN110032981B (en) * | 2019-04-19 | 2022-07-26 | 电子科技大学 | Rotary machine fault identification method based on improved support vector machine |
| CN110767018B (en) * | 2019-11-05 | 2021-10-26 | 中电科思仪科技股份有限公司 | Microwave teaching system for automatically configuring parameters and working method thereof |
| CN110767018A (en) * | 2019-11-05 | 2020-02-07 | 中电科仪器仪表有限公司 | Microwave teaching system for automatically configuring parameters and working method thereof |
| CN111721835A (en) * | 2020-06-28 | 2020-09-29 | 上海理工大学 | An intelligent monitoring method for the condition of the grinding wheel of hollow drill |
| CN111721835B (en) * | 2020-06-28 | 2023-03-31 | 上海理工大学 | Intelligent monitoring method for grinding wheel state of hollow drill |
| CN112037289A (en) * | 2020-09-16 | 2020-12-04 | 安徽意欧斯物流机器人有限公司 | Off-line parameter calibration method based on genetic algorithm |
| CN112785003A (en) * | 2021-01-28 | 2021-05-11 | 武汉市博畅软件开发有限公司 | Method and device for controlling electrically tunable filter |
| CN112785003B (en) * | 2021-01-28 | 2022-08-26 | 武汉市博畅软件开发有限公司 | Method and device for controlling electrically tunable filter |
| CN114553671A (en) * | 2022-02-28 | 2022-05-27 | 国家电网有限公司 | Diagnosis method for power communication network fault alarm |
| CN114622954A (en) * | 2022-05-16 | 2022-06-14 | 合肥派光感知信息技术有限公司 | Illumination and monitoring integrated tunnel safety monitoring system and method |
| CN114622954B (en) * | 2022-05-16 | 2022-08-16 | 合肥派光感知信息技术有限公司 | Illumination and monitoring integrated tunnel safety monitoring system and method |
| CN118670661A (en) * | 2024-08-13 | 2024-09-20 | 南京易信同控制设备科技有限公司 | Fault diagnosis method and system for movable steel structure facility |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105628425A (en) | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine | |
| Wu et al. | A hybrid classification autoencoder for semi-supervised fault diagnosis in rotating machinery | |
| Zhang et al. | An enhanced convolutional neural network for bearing fault diagnosis based on time–frequency image | |
| Saufi et al. | Gearbox fault diagnosis using a deep learning model with limited data sample | |
| Zhou et al. | A novel method based on nonlinear auto-regression neural network and convolutional neural network for imbalanced fault diagnosis of rotating machinery | |
| CN106980822B (en) | A kind of rotary machinery fault diagnosis method based on selective ensemble study | |
| CN105354587B (en) | A fault diagnosis method for wind turbine gearbox | |
| Alrifaey et al. | A novel deep learning framework based RNN-SAE for fault detection of electrical gas generator | |
| CN106197999B (en) | A kind of planetary gear method for diagnosing faults | |
| CN114778112A (en) | Audio Identification and Fault Diagnosis Method for Mechanical Faults of Wind Turbine Generators | |
| CN112016251B (en) | Nuclear power device fault diagnosis method and system | |
| Zhong et al. | Multi-fault rapid diagnosis for wind turbine gearbox using sparse Bayesian extreme learning machine | |
| CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
| CN114936532B (en) | A Fault Diagnosis Method for Offshore Wind Turbines | |
| CN109214356A (en) | A kind of fan transmission system intelligent fault diagnosis method based on DCNN model | |
| CN107560844A (en) | A kind of fault diagnosis method and system of gearbox of wind turbine | |
| CN103953490A (en) | Implementation method for monitoring status of hydraulic turbine set based on HLSNE | |
| CN103558042A (en) | Rapid unit failure diagnosis method based on full state information | |
| CN117276600B (en) | Fault diagnosis method of proton exchange membrane fuel cell system based on PSO-GWO-DELM | |
| Zhang et al. | Feature-level consistency regularized semi-supervised scheme with data augmentation for intelligent fault diagnosis under small samples | |
| Shang et al. | Fault diagnosis method of rolling bearing based on deep belief network | |
| CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
| Xiuli et al. | Fault diagnosis method of wind turbine gearbox based on deep belief network and vibration signal | |
| Du et al. | Data augmentation on fault diagnosis of wind turbine gearboxes with an enhanced flow-based generative model | |
| CN116792264A (en) | GRU-based wind driven generator multi-fault diagnosis method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160601 |
|
| RJ01 | Rejection of invention patent application after publication |