CN106778908A - A kind of novelty detection method and apparatus - Google Patents
A kind of novelty detection method and apparatus Download PDFInfo
- Publication number
- CN106778908A CN106778908A CN201710021030.4A CN201710021030A CN106778908A CN 106778908 A CN106778908 A CN 106778908A CN 201710021030 A CN201710021030 A CN 201710021030A CN 106778908 A CN106778908 A CN 106778908A
- Authority
- CN
- China
- Prior art keywords
- sample
- sum squares
- prediction sum
- prediction
- squares
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Include the invention discloses a kind of novelty detection method and apparatus:The sample under normal condition is collected as training sample and checking sample;Characteristic value is extracted from training sample, and according to training sample characteristic value physical training condition model;Characteristic value is extracted from checking sample, the sample of use state model treatment checking simultaneously obtains predicted characteristics value, one corresponding Prediction sum squares is determined according to each group of characteristic value and predicted characteristics value, and Prediction sum squares threshold value is calculated according to all Prediction sum squares;Whether the Prediction sum squares according to test sample belong to new foreign peoples with the multilevel iudge test sample of Prediction sum squares threshold value.The present invention being capable of high robust and the efficient novelty detection solved the problems, such as in mechanical fault diagnosis.
Description
Technical field
The present invention relates to mechanical test field, especially, it is related to a kind of novelty detection method and apparatus.
Background technology
Safety coefficient very high is considered when being designed due to rotating machinery, can only generally be collected substantial amounts of normal
State sample data.Furthermore, it is contemplated that the loss caused by failure, does not carry out typical fault implantation experiment typically, therefore, in rotation
In the diagnosis application of tool of making a connection, it is difficult to obtain typical fault sample and complete fault mode feature.Therefore, how to pass through
Learn normal condition sample data to recognize abnormal state of affairs (malfunction), the difficulty as rotary machinery fault diagnosis field
Topic.
Novelty detection technology can preferably solve this problem.In recent years, the method that scholars have studied many.They
Substantially it is summarised as three classes:Statistical method, neural net method and Support Vector data description (Support Vector Data
Description, SVDD) method.
Conventional statistical method has parametric method and nonparametric method.Parametric method is by estimating the probability density function of training sample
To judge whether new sample data belongs to known class, such as gauss hybrid models.Parametric method is needed to training data when modeling
Distribution make normal distribution it is assumed that then calculate distributed model parameter.This method amount of calculation is few, easily realizes, is adapted to
Inline diagnosis.But the measurement data of reality is frequently not normal distribution, so, Parameter Estimation Method practicality receives limit
System.Nonparametric method does not need the distribution of prior estimated data, therefore suffers from being widely applied.Such as K- neighbours (K-Nearest
Neighbor, KNN) method, Parzen methods etc. are all typical parametric methods, but the Detection results of these methods are not only selected parameter
Sensitivity is selected, and noiseproof feature is poor.
Novelty detection method based on neutral net has multi-layered perception neural networks (Multi-Layer
Perception, MLP), RBF neural, LVQ Networks (Learning Vector Quantization,
) and Self-organizing Maps (Self Organizing Map, SOM) neutral net LVQ.Relative to statistical method, neutral net is not required to
The probability distribution knowledge of priori is wanted, training can be reduced and calculated intensity, improve novelty detection Generalization Ability.But, nerve net
Network method is not only easily trapped into local minimum, and there is study and owed study, easily the problems such as appearance " dead neuron ".
Input sample is mapped to high-dimensional feature space by SVDD methods using kernel function, and one is then constructed in feature space
The individual hypersphere for covering the normal sample more than most probable, using this hypersphere as decision boundary.SVDD is pushing away for SVM methods
Extensively, small sample, Nonlinear Learning can be preferably solved the problems, such as, is widely used in mechanical fault diagnosis field.However, just
As SVM methods, SVDD methods recognition effect is equally influenceed parameter adjustment and optimization, it is necessary to strict by parameter.Tax and
Duin is proposed by leaving-one method checking come adjusting parameter, but leaving-one method amount of calculation is huge.The SVDD side of many SVDD Model Fusions
Method can reduce influence of the parameter to classification results, but, how to determine that model and the quantity of model become problem again.
For each treatment method parameter selection is sensitive, noiseproof feature is poor in the prior art, there is study and owe study, meter
Calculation amount is huge, the low problem of computational efficiency, and effective solution is there is no at present.
The content of the invention
In view of this, it is an object of the invention to propose a kind of novelty detection method and apparatus, can high robust with
It is efficient to solve the problems, such as that only substantial amounts of normal sample does not have the mechanical fault diagnosis of fault sample.
Based on above-mentioned purpose, the technical scheme that the present invention is provided is as follows:
A kind of novelty detection method is the embodiment of the invention provides, including:
The sample under normal condition is collected as training sample and checking sample;
Characteristic value is extracted from training sample, and according to training sample characteristic value physical training condition model;
Characteristic value is extracted from checking sample, while use state model treatment checking sample obtains predicted characteristics value, root
Determine a corresponding Prediction sum squares according to each group of characteristic value and predicted characteristics value, and according to all squared prediction errors
With calculating Prediction sum squares threshold value;
Prediction sum squares according to test sample are with the multilevel iudge test sample of Prediction sum squares threshold value
It is no to belong to new foreign peoples.
In some embodiments, the extraction step of the characteristic value includes:
Sample is carried out local feature Scale Decomposition (Local characteristic scale decomposition,
LCD), intrinsic scale component (Intrinsic scale component, ISC) set is obtained;
The intrinsic scale relevant with noise and decomposable process point is rejected from intrinsic scale component set with correlation coefficient process
Amount, obtains leading intrinsic scale component set;
The leading intrinsic scale component set of reconstruct obtains the vibration acceleration signal after noise reduction;
The Time-domain Statistics characteristic quantity of high correlation is extracted from the vibration acceleration signal after noise reduction as characteristic value.
In some embodiments, the sample collected under normal condition is to gather the original vibration letter under normal condition
Number.
In some embodiments, the characteristic value includes:Kurtosis, peak factor, nargin, the pulse factor, shape factor.
In some embodiments, the state model is one below:It is linear model, linear reciprocal model, pure secondary
Model or secondary interaction models.
In some embodiments, it is described to calculate Prediction sum squares threshold value bag according to all Prediction sum squares
Include:
The average of Prediction sum squares of all checking samples is calculated as desired value;
The variance of Prediction sum squares of all checking samples is calculated as standard deviation;
Prediction sum squares desired value, standard deviation and false drop rate according to all checking samples calculate squared prediction error
And threshold value.
In some embodiments, the Prediction sum squares according to test sample and Prediction sum squares threshold value
Multilevel iudge test sample whether belong to new foreign peoples and include:
The Prediction sum squares of one test sample and Prediction sum squares threshold value are carried out into size to compare;
When Prediction sum squares are more than Prediction sum squares threshold value, judge that the test sample is new foreign peoples's sample;
When Prediction sum squares are less than Prediction sum squares threshold value, judge that the test sample is normal sample.
The embodiment of the present invention additionally provides a kind of electronic equipment, including at least one processor;And with described at least one
The memory of individual processor communication connection;Wherein, have can be by the finger of at least one computing device for the memory storage
Order, it is described to instruct by least one computing device, so that at least one processor is able to carry out the above method.
The embodiment of the present invention additionally provides a kind of non-transient computer readable storage medium storing program for executing, and the non-transient computer is readable
Storage medium stores computer instruction, and the computer instruction is used to make the computer perform the above method.
The embodiment of the present invention additionally provides a kind of computer program product, and the computer program product includes storage non-
Calculation procedure in transitory computer readable storage medium, the computer program includes programmed instruction, when described program instruction
When being computer-executed, the computer is set to perform the above method.
In sum, the sample collected first under normal condition of the invention is as training sample and verifies sample, from training
Characteristic value is extracted in sample, and according to training sample characteristic value physical training condition model;Then characteristic value is extracted from checking sample,
The sample of use state model treatment checking simultaneously obtains predicted characteristics value, and one is determined according to each group of characteristic value and predicted characteristics value
Individual corresponding Prediction sum squares, and calculate Prediction sum squares threshold value according to all Prediction sum squares;Last root
Whether belong to strange with the multilevel iudge test sample of Prediction sum squares threshold value according to the Prediction sum squares of test sample
Class.
The present invention being capable of high robust and the efficient novelty detection solved the problems, such as in mechanical fault diagnosis.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to institute in embodiment
The accompanying drawing for needing to use is briefly described, it should be apparent that, drawings in the following description are only some implementations of the invention
Example, for those of ordinary skill in the art, on the premise of not paying creative work, can also obtain according to these accompanying drawings
Obtain other accompanying drawings.
Fig. 1 is a kind of flow chart of the novelty detection method according to the embodiment of the present invention;
During Fig. 2 is a kind of novelty detection according to the embodiment of the present invention, the flow chart of characteristic value is extracted from sample;
Fig. 3 is a kind of detail flowchart of one embodiment of the novelty detection method according to the embodiment of the present invention;
Fig. 4 is an implementation for performing a kind of electronic equipment of novelty detection method provided in an embodiment of the present invention
The hardware architecture diagram of example.
Specific embodiment
To make the object, technical solutions and advantages of the present invention become more apparent, below in conjunction with the embodiment of the present invention
Accompanying drawing, the technical scheme in the embodiment of the present invention is further carried out it is clear, complete, describe in detail, it is clear that it is described
Embodiment is only a part of embodiment of the invention, rather than whole embodiments.Based on the embodiment in the present invention, this area
The every other embodiment that those of ordinary skill is obtained, belongs to the scope of protection of the invention.
Based on above-mentioned purpose, a kind of one side of the embodiment of the present invention, it is proposed that of novelty detection method
Embodiment.
As shown in figure 1, the novelty detection method for providing according to embodiments of the present invention includes:
Step S101, collects the sample under normal condition as training sample and checking sample;
Step S103, extracts characteristic value from training sample, and according to training sample characteristic value physical training condition model;
Step S105, characteristic value is extracted from checking sample, while use state model treatment checking sample is predicted
Characteristic value, a corresponding Prediction sum squares are determined according to each group of characteristic value and predicted characteristics value, and according to all pre-
Survey error sum of squares and calculate Prediction sum squares threshold value;
Step S107, the Prediction sum squares according to test sample are surveyed with the multilevel iudge of Prediction sum squares threshold value
Whether sample originally belongs to new foreign peoples.
As shown in Fig. 2 in some embodiments, the extraction step of the characteristic value includes:
Step S201, local feature Scale Decomposition is carried out to sample, obtains intrinsic scale component set;
Step S203, is rejected in relevant with noise and decomposable process with correlation coefficient process from intrinsic scale component set
Scale component is reported, leading intrinsic scale component set is obtained;
Step S205, the leading intrinsic scale component set of reconstruct obtains the vibration acceleration signal after noise reduction;
Step S207, extracts the Time-domain Statistics characteristic quantity of high correlation as spy from the vibration acceleration signal after noise reduction
Value indicative.
In some embodiments, the sample collected under normal condition is to gather the original vibration letter under normal condition
Number.
In some embodiments, the characteristic value includes:Kurtosis, peak factor, nargin, the pulse factor, shape factor.
In some embodiments, the state model is one below:It is linear model, linear reciprocal model, pure secondary
Model or secondary interaction models.
In some embodiments, it is described to calculate Prediction sum squares threshold value bag according to all Prediction sum squares
Include:
The average of Prediction sum squares of all checking samples is calculated as desired value;
The variance of Prediction sum squares of all checking samples is calculated as standard deviation;
Prediction sum squares desired value, standard deviation and false drop rate according to all checking samples calculate squared prediction error
And threshold value.
In some embodiments, the Prediction sum squares according to test sample and Prediction sum squares threshold value
Multilevel iudge test sample whether belong to new foreign peoples and include:
The Prediction sum squares of one test sample and Prediction sum squares threshold value are carried out into size to compare;
When Prediction sum squares are more than Prediction sum squares threshold value, judge that the test sample is new foreign peoples's sample;
When Prediction sum squares are less than Prediction sum squares threshold value, judge that the test sample is normal sample.
In sum, the sample collected first under normal condition of the invention is as training sample and verifies sample, from training
Characteristic value is extracted in sample, and according to training sample characteristic value physical training condition model;Then characteristic value is extracted from checking sample,
The sample of use state model treatment checking simultaneously obtains predicted characteristics value, and one is determined according to each group of characteristic value and predicted characteristics value
Individual corresponding Prediction sum squares, and calculate Prediction sum squares threshold value according to all Prediction sum squares;Last root
Whether belong to strange with the multilevel iudge test sample of Prediction sum squares threshold value according to the Prediction sum squares of test sample
Class.
The present invention can high robust with it is efficient solve only have substantial amounts of normal sample there is no the machinery of fault sample
Troubleshooting issue.
Based on above-mentioned purpose, a kind of second aspect of the embodiment of the present invention, it is proposed that of novelty detection method
Embodiment.
Classifying identification method (Variable predictive model-based class based on variable prediction model
Discriminate, VPMCD) it is a kind of new mode identification method.The method is divided into two mistakes of model training and Classification and Identification
Journey.In model training process, VPMCD methods are using selection linear model, linear reciprocal model, pure secondary model and secondary interaction
One of four kinds of models, using regression analysis, the phase between system features variable are obtained with the minimum discriminant function of error
Mutual internal relation, so as to set up the variable prediction model (VPM) of reflection system nature feature.In Classification and Identification stage, VPMCD side
Method predicts the characteristic value of unknown sample using VPM, then realizes judging unknown with the minimum foundation of Prediction sum squares
The classification of sample, realizes Classification and Identification.
The embodiment of the present invention is based on VPMCD methods, it is proposed that a kind of novelty detection method (ND-VPMCD) method.It is first
First, set up forecast model using normal sample data and set Prediction sum squares threshold value, then for a certain test sample,
The characteristic value of test sample is predicted using the forecast model for training, calculates Prediction sum squares, finally with prediction
Whether error sum of squares is discriminant function less than threshold value, judges whether test sample is new foreign peoples.The specific steps of the method are such as
Under:
First, ND-VPMCD model trainings are carried out.
The sample under N number of normal condition is collected altogether, and sample is randomly divided into two groups, one group of NtrainIndividual sample, for training
Model, remaining N-NtrainIndividual sample is another group, for verifying that model and given threshold (can also be regardless of when number of samples is few
Group).Then, to NtrainIndividual normal sample extracts characteristic value, composition characteristic vector X=[X1,X2,…,Xp], and physical training condition mould
Type VPMnormal。
Then, statistical method determines Prediction sum squares threshold value.
Use VPMnormalN-N of the model to normal conditiontrainEach characteristic variable of individual sample is predicted, and asks for pre-
Survey error sum of squares vector SSEnormal.Normal condition SSEnormalShould be at it in threshold interval.If within the sampling period,
SSEnormalExceed threshold value, then need re -training model.According to Chebyshev inequality, for any real number ε > 0, have
P{|SSEnormal-u|≥ε}≤σ2/ε2 (1)
ε=n σ, n > 0 are made, then above-mentioned inequality can be transformed to:
P{|SSEnormal-u|≥nσ}≤1/n2 (2)
Wherein u is mathematic expectaion, and σ is standard deviation.For to regular inspection rate α=1/n2, then SSE can be obtainednormalIt is normal
Region is:C:[0,μ±nσ].
Assuming that SSEnormalBelong to normal distribution, it is contemplated that be always on the occasion of then SSEnormalBilateral threshold region be changed to
It is unilateral.For example, given false drop rate α=0.1,0.05,0.025, the corresponding threshold interval C of difference1:[0,μ+3σ]、C2:[0,μ+
4.5σ]、C3:[0, μ+6 σ], can use SSE in practical applicationnormalAverage replace mathematic expectaion u, replace standard with variance
Difference σ.Therefore, under different false drop rate requirements, corresponding maximum squared error and MSSE can be obtainednormal=u+n σ conducts
Threshold value differentiates whether test sample belongs to new foreign peoples.
Finally, ND-VPMCD classification is carried out.
Collecting test sample z, and its characteristic value is extracted, composition characteristic vector X=[X1,X2,…,Xp]。
For test sample z, using VPMnormalIt is predicted, the predicted value vector of all features is respectively obtained
Calculate the error sum of squares SSE of test sample z predicted valuesz.For test sample z, judge whether it belongs to strange
Class, will see SSEzWhether MSSE is more thannormal.If i.e.
SSEz> MSSEnormal (3)
Set up, then test sample z belongs to new foreign peoples, otherwise, belongs to normal class.
UCI databases iris data sets chosen below carry out simulation analysis, verify ND-VPMCD novelty detection methods
Validity.Iris data sets include 3 class samples, respectively Setosa (ST), Versicolor (VS) and Virginica
(VR), per class each 50 groups of data of sample, 150 groups of data are had, 4 property values of every group of data, i.e. feature value vector are X=[X1,
X2,X3,X4].ST classes are considered as normal class by this experiment, altogether 50 groups of ST class data, and class tag definition is+1;VS and VR are regarded
It is new foreign peoples (non-target class), altogether 100 groups of new heterogeneous datas that class tag definition is -1.It is random from 50 groups of ST class data sets
20 groups of data are extracted as training sample, with the LI models of r=2 as the variable prediction model of ST classes, be see the table below.
Meanwhile, calculate the error sum of squares SSE of ST class samplesnormal, and obtain MSSEnormal=0.5195.Then, use
The variable prediction model of ST classes is predicted to 30 groups of ST classes data and 100 groups of non-ST class testings samples, and calculates each
The Prediction sum squares SSE of test samplez.Finally, which kind of the discriminant function according to formula (3) is belonging respectively to come judgement sample
Not.
The performance indications of novelty detection device have:Verification and measurement ratio rt(Rate of True Alarm) and false alarm rate rf(Rate
of False Alarm).Verification and measurement ratio is the ratio that foreign peoples's sample is judged to exception class, then false dismissed rate 1-rt.False alarm rate is normal
Sample is judged to the ratio of exception class.Performance detector high should existing verification and measurement ratio higher have relatively low false alarm rate again.ND-
VPMCD methods be see the table below to the analysis of simulation experiment result of iris data, and ND-VPMCD methods are examined to the new foreign peoples of iris data
Survey rate is 100%, and false alarm rate is 0, can effectively detect new foreign peoples.
In order to verify the validity and practicality of the mechanical failure diagnostic method that the embodiment of the present invention is proposed, made from bearing
It is experimental subjects, on bearing fault experimental bench, rolling bearing 6307 is tested.By laser cutting respectively in the axis of rolling
6307 inner ring and the grooving of outer ring processing width 0.15mm, deep 0.13mm is held to set inner ring failure and outer ring failure.In experiment
200 samples of vibration acceleration signal of normal condition are gathered, and they are divided into two groups, every group of 100 samples.Collection
Used as analyze data, sample frequency is 4096Hz, axle to each 18 samples of vibration acceleration of inner ring failure and outer ring malfunction
Rotating speed is 680rpm, and sampled data length is 1024 points.Normal condition is defined as target class in experiment, class label is+1, will
Inner ring failure and outer ring malfunction are defined as non-target class, i.e. exception class, and class label is -1.
Due to sensor obtain signal contain stronger background noise, using extracted again after LCD method noise reductions feature to
Amount.LCD decomposition is carried out to original vibration acceleration signal first, some ISC components are obtained.Then rejected using correlation coefficient process
The pseudo- ISC component relevant with noise and decomposable process, obtains leading ISC components, then reconstructs these leading ISC components, obtains
Vibration acceleration signal after to noise reduction.Finally there are the Time-domain Statistics of preferable correlation to the signal extraction after these noise reductions
Characteristic quantity:Kurtosis, peak factor, nargin, the pulse factor, shape factor.
Fig. 3 is illustrated that the flow chart of the novelty detection method based on ND-VPMCD.First, from first group of normal sample
In randomly select NtrainIndividual sample is vectorial as ND- with foregoing 5 Time-domain Statistics characteristic quantities composition characteristic as training sample
The input vector training predictive variable model VPM of VPMCDtraining, then using VPMtrainingTo remaining 100-NtrainIt is individual just
Normal sample is predicted, and obtains Prediction sum squares, Prediction sum squares average and standard deviation is calculated, so as to obtain threshold value
MSSEnormal.In experiment, it is considered to which QI and LI includes that the normal condition number of samples in L models and Q model, and experiment is more, therefore
Using the LI models and the QI models of r=4 of r=3, N is chosen respectivelytrainTrained for 20,30,40 and 50 and set up VPMtraining。
Finally, using second group of 100 normal sample and 36 fault samples as test sample, using the predictive variable model for training
VPMtrainingEach characteristic variable of test sample is predicted, and obtains the Prediction sum squares of predicted value and each sample, with pre-
Survey whether error sum of squares is discriminant function less than threshold value, realize novelty detection.
The testing result of different number of training two kinds of models now see the table below, and the average detected time in experiment is
0.28s。
Visible detection rate is 100%, so as to demonstrate the reliability of ND-VPMCD methods.In addition we know, ND-VPMCD
The performance of detector is influenceed smaller by types of models and number of training.It is also noted that when number of training is 30, checking
When sample number is 70, false alarm rate is minimum, that is, when verifying that sample number is about 2~3 times of number of training, accuracy of detection is compared with Gao Erxu
Alert rate is relatively low.
In order to be analyzed, rolling bearing event is carried out using widely used SVDD methods and ND-VPMCD methods
Barrier novelty detection.For the fairness for comparing, 30 samples are equally randomly selected from first group and makees training sample, and by
Two groups of 100 normal samples and 36 fault samples do test sample.
SVDD methods are it needs to be determined that two parameters:The upper bound ν and nuclear parameter σ of training error rate.Following table shows different
SVDD testing results in the case of parameter setting, as seen from the table, set different parameters, and SVDD classifying qualities have larger difference
It is different.Therefore, the embodiment of the present invention uses network searching method, optimized parameter is found with the minimum object function of false alarm rate, so that
Obtaining optimal parameter is:ν=0.2, σ=2.0, now, SVDD classification results are:Verification and measurement ratio rt=100%, false alarm rate rf
=10.64%.
In sum, the sample collected first under normal condition of the invention is as training sample and verifies sample, from training
Characteristic value is extracted in sample, and according to training sample characteristic value physical training condition model;Then characteristic value is extracted from checking sample,
The sample of use state model treatment checking simultaneously obtains predicted characteristics value, and one is determined according to each group of characteristic value and predicted characteristics value
Individual corresponding Prediction sum squares, and calculate Prediction sum squares threshold value according to all Prediction sum squares;Last root
Whether belong to strange with the multilevel iudge test sample of Prediction sum squares threshold value according to the Prediction sum squares of test sample
Class.
The present invention can high robust with it is efficient solve only have substantial amounts of normal sample there is no the machinery of fault sample
Troubleshooting issue.
Based on above-mentioned purpose, the 3rd aspect of the embodiment of the present invention, it is proposed that one kind performs the novelty detection side
One embodiment of the electronic equipment of method.
The electronic equipment for performing the novelty detection method includes:
At least one processor;And,
The memory being connected with least one processor communication;Wherein,
The memory storage has can be by the instruction of at least one computing device, and the instruction is by described at least one
Individual computing device, so that at least one processor is able to carry out any one method as described above.
Fig. 4 is illustrated that one embodiment of the electronic equipment of the execution novelty detection method of present invention offer
Hardware architecture diagram.
By taking electronic equipment as shown in Figure 4 as an example, a processor 401 and a storage are included in the electronic equipment
Device 402, and can also include:Input unit 403 and output device 404.
Processor 401, memory 402, input unit 403 and output device 404 can be by bus or other modes
Connection, in Fig. 4 as a example by being connected by bus.
Memory 402 can be used to store non-volatile software journey as a kind of non-volatile computer readable storage medium storing program for executing
Sequence, non-volatile computer executable program and module, the novelty detection method correspondence such as in the embodiment of the present application
Programmed instruction/module.Processor 401 by run non-volatile software program of the storage in memory 402, instruction and
Module, so that the various function application of execute server and data processing, that is, realize the new foreign peoples inspection of above method embodiment
Survey method.
Memory 402 can include storing program area and storage data field, wherein, storing program area can store operation system
Application program required for system, at least one function;Storage data field can be stored to be created according to using for novelty detection device
Data built etc..Additionally, memory 402 can include high-speed random access memory, nonvolatile memory can also be included,
For example, at least one disk memory, flush memory device or other non-volatile solid state memory parts.In certain embodiments,
Memory 402 is optional including the memory remotely located relative to processor 401.The example of above-mentioned network is included but is not limited to mutually
Networking, intranet, LAN, mobile radio communication and combinations thereof.
Input unit 403 can receive the numeral or character information of input, and produce the user with novelty detection device
Set and the relevant key signals of function control are input into.Output device 404 may include the display devices such as display screen.
One or more of module storages, when being performed by the processor 401, are held in the memory 402
Novelty detection method in the above-mentioned any means embodiment of row.
Any one embodiment of the electronic equipment for performing the novelty detection method, can reach and correspond to therewith
The identical or similar effect of foregoing any means embodiment.
The embodiment of the present application provides a kind of non-transient computer storage medium, and the computer-readable storage medium is stored with meter
Calculation machine executable instruction, the computer executable instructions can perform the novelty detection method in above-mentioned any means embodiment.
The embodiment of the non-transient computer storage medium, can reach corresponding foregoing any means embodiment it is identical or
Similar effect.
One of ordinary skill in the art will appreciate that all or part of flow in realizing above-described embodiment method, can be
Related hardware is instructed to complete by computer program, described program can be stored in a computer read/write memory medium
In, the program is upon execution, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..The embodiment of the computer program, can reach corresponding foregoing any means embodiment identical
Or similar effect.
Additionally, typically, device, equipment described in the disclosure etc. can be various electric terminal equipments, such as mobile phone, individual
Digital assistants (PDA), panel computer (PAD), intelligent television etc., or large-scale terminal device, such as server, therefore this
Disclosed protection domain should not limit as certain certain types of device, equipment.Client described in the disclosure can be with electricity
The combining form of sub- hardware, computer software or both is applied in above-mentioned any one electric terminal equipment.
Additionally, the computer program for being also implemented as being performed by CPU according to disclosed method, the computer program
Can store in a computer-readable storage medium.When the computer program is performed by CPU, limit in disclosed method is performed
Fixed above-mentioned functions.
Additionally, above method step and system unit can also utilize controller and cause controller reality for storing
The computer-readable recording medium of the computer program of existing above-mentioned steps or Elementary Function is realized.
In addition, it should be appreciated that computer-readable recording medium (for example, memory) as herein described can be volatile
Property memory or nonvolatile memory, or both volatile memory and nonvolatile memory can be included.As example
Son and it is nonrestrictive, nonvolatile memory can include read-only storage (ROM), programming ROM (PROM), electrically programmable
ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory.Volatile memory can include arbitrary access
Memory (RAM), the RAM can serve as external cache.Nonrestrictive as an example, RAM can be with more
The form of kind is obtained, such as synchronous random access memory (DRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate SDRAM
(DDR SDRAM), enhancing SDRAM (ESDRAM), synchronization link DRAM (SLDRAM) and direct RambusRAM (DRRAM).Institute
The storage device of disclosed aspect is intended to the memory of including but not limited to these and other suitable type.
Those skilled in the art will also understand is that, the various illustrative logical blocks with reference to described by disclosure herein, mould
Block, circuit and algorithm steps may be implemented as the combination of electronic hardware, computer software or both.It is hard in order to clearly demonstrate
This interchangeability of part and software, the function with regard to various exemplary components, square, module, circuit and step it is entered
General description is gone.This function is implemented as software and is also implemented as hardware depending on concrete application and applying
To the design constraint of whole system.Those skilled in the art can in a variety of ways realize described for every kind of concrete application
Function, but this realize that decision should not be interpreted as causing a departure from the scope of the present disclosure.
Various illustrative logical blocks, module and circuit with reference to described by disclosure herein can be utilized and are designed to
The following part of function described here is performed to realize or perform:General processor, digital signal processor (DSP), special collection
Into circuit (ASIC), field programmable gate array (FPGA) or other PLDs, discrete gate or transistor logic, divide
Any combinations of vertical nextport hardware component NextPort or these parts.General processor can be microprocessor, but alternatively, treatment
Device can be any conventional processors, controller, microcontroller or state machine.Processor can also be implemented as computing device
Combination, for example, the combination of DSP and microprocessor, multi-microprocessor, one or more microprocessors combination DSP core or any
Other this configurations.
The step of method or algorithm with reference to described by disclosure herein can be directly contained in hardware in, held by processor
In capable software module or in combination of the two.Software module may reside within RAM memory, flash memory, ROM storages
Device, eprom memory, eeprom memory, register, hard disk, removable disk, CD-ROM or known in the art it is any its
In the storage medium of its form.Exemplary storage medium is coupled to processor so that processor can be from the storage medium
Middle reading information writes information to the storage medium.In an alternative, the storage medium can be with processor collection
Into together.Processor and storage medium may reside within ASIC.ASIC may reside within user terminal.In a replacement
In scheme, processor and storage medium can be resident in the user terminal as discrete assembly.
In one or more exemplary designs, the function can be real in hardware, software, firmware or its any combination
It is existing.If realized in software, can be stored the function as one or more instructions or code in computer-readable
Transmitted on medium or by computer-readable medium.Computer-readable medium includes computer-readable storage medium and communication media,
The communication media includes any medium for helping that computer program is sent to another position from position.Storage medium
It can be any usable medium that can be accessed by a general purpose or special purpose computer.It is nonrestrictive as an example, the computer
Computer-readable recording medium can include RAM, ROM, EEPROM, CD-ROM or other optical disc memory apparatus, disk storage equipment or other magnetic
Property storage device, or can be used for carrying or storage form program code and can for needed for instruction or data structure
Any other medium accessed by universal or special computer or universal or special processor.Additionally, any connection can
It is properly termed as computer-readable medium.If for example, using coaxial cable, optical fiber cable, twisted-pair feeder, digital subscriber line
(DSL) or such as infrared ray, radio and microwave wireless technology come from website, server or other remote sources send software,
Then the wireless technology of above-mentioned coaxial cable, optical fiber cable, twisted-pair feeder, DSL or such as infrared elder generations, radio and microwave is included in
The definition of medium.As used herein, disk and CD include compact disk (CD), laser disk, CD, digital versatile disc
(DVD) the usual magnetically reproduce data of, floppy disk, Blu-ray disc, wherein disk, and CD is using laser optics ground reproduce data.On
The combination for stating content should also be as being included in the range of computer-readable medium.
Above is exemplary embodiment disclosed by the invention, it should be noted that in the sheet limited without departing substantially from claim
On the premise of scope of disclosure, may be many modifications and change.According to the method right of open embodiment described herein
It is required that function, step and/or action be not required to any particular order perform.Although additionally, the element of the disclosure can be with individual
Body form is described or required, it is also contemplated that it is multiple, it is unless explicitly limited odd number.
It should be appreciated that it is used in the present context, unless context clearly supports exception, singulative "
It is individual " (" a ", " an ", " the ") be intended to also include plural form.It is to be further understood that "and/or" used herein is
Finger includes any of or more than one project listed in association and is possible to combine.
Above-mentioned embodiment of the present disclosure sequence number is for illustration only, and the quality of embodiment is not represented.
One of ordinary skill in the art will appreciate that realizing that all or part of step of above-described embodiment can be by hardware
To complete, it is also possible to instruct the hardware of correlation to complete by program, described program can be stored in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only storage, disk or CD etc..
Those of ordinary skill in the art should be understood:The discussion of any of the above embodiment is exemplary only, not
It is intended to imply that the scope of the present disclosure (including claim) is limited to these examples;Under the thinking of the embodiment of the present invention, the above
Can also be combined between technical characteristic in embodiment or different embodiments, and there is the present invention as described above and implemented
Many other changes of the different aspect of example, for simplicity, they are provided not in details.Therefore, it is all in the embodiment of the present invention
Spirit and principle within, any omission, modification, equivalent, improvement for being made etc. should be included in the embodiment of the present invention
Within protection domain.
Claims (10)
1. a kind of novelty detection method, it is characterised in that including:
The sample under normal condition is collected as training sample and checking sample;
Characteristic value is extracted from training sample, and according to training sample characteristic value physical training condition model;
Characteristic value is extracted from checking sample, while use state model treatment checking sample obtains predicted characteristics value, according to every
One group of characteristic value determines a corresponding Prediction sum squares with predicted characteristics value, and according to all Prediction sum squares meters
Calculate Prediction sum squares threshold value;
Whether the Prediction sum squares according to test sample belong to the multilevel iudge test sample of Prediction sum squares threshold value
In new foreign peoples.
2. method according to claim 1, it is characterised in that the extraction step of the characteristic value includes:
Local feature Scale Decomposition is carried out to sample, intrinsic scale component set is obtained;
The intrinsic scale component relevant with noise and decomposable process is rejected from intrinsic scale component set with correlation coefficient process, is obtained
Intrinsic scale component set must be dominated;
The leading intrinsic scale component set of reconstruct obtains the vibration acceleration signal after noise reduction;
The Time-domain Statistics characteristic quantity of high correlation is extracted from the vibration acceleration signal after noise reduction as characteristic value.
3. method according to claim 2, it is characterised in that the sample under the collection normal condition is the normal shape of collection
Original vibration signal under state.
4. method according to claim 2, it is characterised in that the characteristic value includes:Kurtosis, peak factor, nargin, arteries and veins
Rush the factor, shape factor.
5. method according to claim 2, it is characterised in that the state model is one below:It is linear model, linear
Interaction models, pure secondary model or secondary interaction models.
6. method according to claim 2, it is characterised in that described to calculate prediction according to all Prediction sum squares and miss
Difference quadratic sum threshold value includes:
The average of Prediction sum squares of all checking samples is calculated as desired value;
The variance of Prediction sum squares of all checking samples is calculated as standard deviation;
Prediction sum squares desired value, standard deviation and false drop rate according to all checking samples calculate Prediction sum squares threshold
Value.
7. method according to claim 6, it is characterised in that the Prediction sum squares according to test sample with it is pre-
Whether the multilevel iudge test sample of survey error sum of squares threshold value belongs to new foreign peoples includes:
The Prediction sum squares of one test sample and Prediction sum squares threshold value are carried out into size to compare;
When Prediction sum squares are more than Prediction sum squares threshold value, judge that the test sample is new foreign peoples's sample;
When Prediction sum squares are less than Prediction sum squares threshold value, judge that the test sample is normal sample.
8. a kind of electronic equipment, it is characterised in that including at least one processor;And with least one processor communication
The memory of connection;Wherein, have can be by the instruction of at least one computing device, the instruction quilt for the memory storage
At least one computing device, so that at least one processor is able to carry out such as claim 1-7 any one institute
The method stated.
9. a kind of non-transient computer readable storage medium storing program for executing, it is characterised in that the non-transient computer readable storage medium storing program for executing is deposited
Storage computer instruction, the computer instruction is used to make the method described in the computer perform claim requirement 1-7 any one.
10. a kind of computer program product, it is characterised in that the computer program product includes storage in non-transient computer
Calculation procedure on readable storage medium storing program for executing, the computer program includes programmed instruction, when described program instruction is held by computer
During row, make the method described in the computer perform claim requirement 1-7 any one.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710021030.4A CN106778908A (en) | 2017-01-11 | 2017-01-11 | A kind of novelty detection method and apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710021030.4A CN106778908A (en) | 2017-01-11 | 2017-01-11 | A kind of novelty detection method and apparatus |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106778908A true CN106778908A (en) | 2017-05-31 |
Family
ID=58947773
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710021030.4A Pending CN106778908A (en) | 2017-01-11 | 2017-01-11 | A kind of novelty detection method and apparatus |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106778908A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107480249A (en) * | 2017-08-08 | 2017-12-15 | 中国工商银行股份有限公司 | Ranking method, equipment and the system of magnanimity information |
CN108257670A (en) * | 2018-01-22 | 2018-07-06 | 北京颐圣智能科技有限公司 | The method for building up and device of medical explanation model |
CN109299738A (en) * | 2018-09-19 | 2019-02-01 | 语联网(武汉)信息技术有限公司 | Choosing method, device and the electronic equipment of contribution gene |
CN109427050A (en) * | 2017-08-23 | 2019-03-05 | 阿里巴巴集团控股有限公司 | Guide wheel quality determining method and equipment |
CN109909801A (en) * | 2019-03-13 | 2019-06-21 | 湖北文理学院 | Turntable error calibration method, device and electronic equipment |
CN110598797A (en) * | 2019-09-18 | 2019-12-20 | 北京明略软件***有限公司 | Fault detection method and device, storage medium and electronic device |
CN111209567A (en) * | 2019-12-30 | 2020-05-29 | 北京邮电大学 | Method and device for judging perceptibility of improving robustness of detection model |
CN111444471A (en) * | 2020-02-25 | 2020-07-24 | 国网河南省电力公司电力科学研究院 | Cable production quality abnormity detection method and system based on multivariate Gaussian distribution |
WO2020155756A1 (en) * | 2019-01-28 | 2020-08-06 | 平安科技(深圳)有限公司 | Method and device for optimizing abnormal point proportion based on clustering and sse |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101587017A (en) * | 2009-06-19 | 2009-11-25 | 湖南大学 | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum |
CN103033362A (en) * | 2012-12-31 | 2013-04-10 | 湖南大学 | Gear fault diagnosis method based on improving multivariable predictive models |
-
2017
- 2017-01-11 CN CN201710021030.4A patent/CN106778908A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101587017A (en) * | 2009-06-19 | 2009-11-25 | 湖南大学 | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum |
CN103033362A (en) * | 2012-12-31 | 2013-04-10 | 湖南大学 | Gear fault diagnosis method based on improving multivariable predictive models |
Non-Patent Citations (3)
Title |
---|
SONGRONG LUO ETC.: "《A Fault Diagnosis Model Based on LCD-SVD-ANN-MIV and VPMCD f or Rotating Machinery》", 《SHOCK AND VIBRATION》 * |
程军圣等: "《一种新的非平稳信号分析方法-— 局部特征尺度分解法》", 《振动工程学报》 * |
罗颂荣等: "基于变量预测模型模式识别的旋转机械故障诊断研究", 《万方学术》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107480249A (en) * | 2017-08-08 | 2017-12-15 | 中国工商银行股份有限公司 | Ranking method, equipment and the system of magnanimity information |
CN107480249B (en) * | 2017-08-08 | 2020-09-11 | 中国工商银行股份有限公司 | Mass information rating method, device and system |
CN109427050A (en) * | 2017-08-23 | 2019-03-05 | 阿里巴巴集团控股有限公司 | Guide wheel quality determining method and equipment |
CN108257670A (en) * | 2018-01-22 | 2018-07-06 | 北京颐圣智能科技有限公司 | The method for building up and device of medical explanation model |
CN108257670B (en) * | 2018-01-22 | 2021-06-22 | 北京颐圣智能科技有限公司 | Method and device for establishing medical interpretation model |
CN109299738A (en) * | 2018-09-19 | 2019-02-01 | 语联网(武汉)信息技术有限公司 | Choosing method, device and the electronic equipment of contribution gene |
CN109299738B (en) * | 2018-09-19 | 2021-10-26 | 语联网(武汉)信息技术有限公司 | Manuscript gene selection method and device and electronic equipment |
WO2020155756A1 (en) * | 2019-01-28 | 2020-08-06 | 平安科技(深圳)有限公司 | Method and device for optimizing abnormal point proportion based on clustering and sse |
CN109909801A (en) * | 2019-03-13 | 2019-06-21 | 湖北文理学院 | Turntable error calibration method, device and electronic equipment |
CN110598797A (en) * | 2019-09-18 | 2019-12-20 | 北京明略软件***有限公司 | Fault detection method and device, storage medium and electronic device |
CN110598797B (en) * | 2019-09-18 | 2022-07-08 | 北京明略软件***有限公司 | Fault detection method and device, storage medium and electronic device |
CN111209567A (en) * | 2019-12-30 | 2020-05-29 | 北京邮电大学 | Method and device for judging perceptibility of improving robustness of detection model |
CN111209567B (en) * | 2019-12-30 | 2022-05-03 | 北京邮电大学 | Method and device for judging perceptibility of improving robustness of detection model |
CN111444471A (en) * | 2020-02-25 | 2020-07-24 | 国网河南省电力公司电力科学研究院 | Cable production quality abnormity detection method and system based on multivariate Gaussian distribution |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106778908A (en) | A kind of novelty detection method and apparatus | |
CN111459700B (en) | Equipment fault diagnosis method, diagnosis device, diagnosis equipment and storage medium | |
CN109817339B (en) | Patient grouping method and device based on big data | |
CN108399248A (en) | A kind of time series data prediction technique, device and equipment | |
Hajian et al. | New automated power quality recognition system for online/offline monitoring | |
CN107101829B (en) | A kind of intelligent diagnosing method of aero-engine structure class failure | |
CN104680542A (en) | Online learning based detection method for change of remote-sensing image | |
CN104749334A (en) | Mode-recognition-based design method for biological abnormal water quality evaluation system | |
CN105335379B (en) | The method and apparatus to sort to the combination of mutation, test case, random seed in mutation test | |
Guan et al. | Rolling bearing fault diagnosis based on information fusion and parallel lightweight convolutional network | |
CN106792883A (en) | Sensor network abnormal deviation data examination method and system | |
CN116610998A (en) | Switch cabinet fault diagnosis method and system based on multi-mode data fusion | |
Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
CN109961030A (en) | Pavement patching information detecting method, device, equipment and storage medium | |
CN112200048A (en) | Regression model-based rotating equipment fault prediction method and system and readable storage medium | |
CN111753877B (en) | Product quality detection method based on deep neural network migration learning | |
CN116204768A (en) | Fault diagnosis method, device, computer equipment and storage medium | |
CN114237046A (en) | Partial discharge pattern recognition method based on SIFT data feature extraction algorithm and BP neural network model | |
Cao et al. | Seismic damage identification of high arch dams based on an unsupervised deep learning approach | |
CN113822336A (en) | Cloud hard disk fault prediction method, device and system and readable storage medium | |
Xie et al. | Internal defect inspection in magnetic tile by using acoustic resonance technology | |
CN104463912A (en) | Multi-scale target tracking method based on cluster similarity | |
CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion | |
Liu et al. | A novel incipient fault diagnosis method for analogue circuits based on an MLDLCN | |
CN114202224B (en) | Method, apparatus, medium for detecting weld quality in a production environment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170531 |