CN108267311A - A kind of mechanical multidimensional big data processing method based on tensor resolution - Google Patents
A kind of mechanical multidimensional big data processing method based on tensor resolution Download PDFInfo
- Publication number
- CN108267311A CN108267311A CN201810061289.6A CN201810061289A CN108267311A CN 108267311 A CN108267311 A CN 108267311A CN 201810061289 A CN201810061289 A CN 201810061289A CN 108267311 A CN108267311 A CN 108267311A
- Authority
- CN
- China
- Prior art keywords
- tensor
- signal
- big data
- truncation parameters
- processing method
- 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
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/028—Acoustic or vibration analysis
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The present invention provides a kind of mechanical multidimensional big data processing method based on tensor resolution, this method includes signal acquisition, frequency domain information structure, tensor model construction, Truncation Parameters are chosen, target tensor reconstructs.Signal modeling into tensor form, in higher dimensional space can be solved the problems, such as big data by tensor tool, by simple vibration measurement, using the processing method based on tensor resolution, efficiently and reliably carry out multidimensional processiug by the present invention.
Description
Technical field
The present invention relates to big data process field more particularly to a kind of mechanical multidimensional big data processing based on tensor resolution
Method.
Background technology
In recent years, as the mankind explore the expansion of range, data record scope is expanded rapidly, has accumulated the number of magnanimity
According to.In mechanical fault diagnosis field, since mechanized equipment is distributed, wide, measuring point is numerous, data sampling frequency is high, is on active service and lasts length
Etc. reasons, obtain the diagnostic data of magnanimity multidimensional.It is to carry out fault diagnosis that useful information ingredient is excavated from big data
It is crucial.
With the continuous high speed development of multidimensional data, future, mass data inevitably will be at higher dimensional space
Reason.This patent proposes that will acquire multi-channel Vibration Signal is modeled as tensor form, passes through tensor resolution tool in higher dimensional space
Solve big data process problem.Signal processing mode based on tensor has the advantages of intrinsic and feature so that it is mechanical big
Data processing field has far-reaching researching value and application potential.The present invention proposes the signal processing side based on tensor analysis
Method, it will provided for the processing of other field multidimensional big data using stage.
Big data processing coverage is very extensive, and high dimensional data processing always is hot spot.High dimensional data it is openness,
Lead in data processing method and the lower dimensional space in higher dimensional space that there are significant differences.Many maturations in traditional lower dimensional space
Algorithm can not attain the results expected or even can not run in higher dimensional space.In addition, different from higher-dimension remotely sensed image, acoustic matrix
The tensor model construction mode of column signal, creativeness of the invention be also embodied in mechanized equipment vibration signal frequency domain information, when
The information such as domain information and channel set up tensor model.
Invention content
The technical problems to be solved by the invention are to provide a kind of mechanical big data processing method based on tensor analysis,
By simple vibration measurement, using the processing method based on tensor resolution, efficiently and reliably carry out at multidimensional signal
Reason.To solve the above problems, the present invention is a kind of mechanical big data processing method based on tensor resolution, the present invention be by with
What lower scheme was realized:
This method comprises the following steps:
(1) gear-box time domain vibration signal is acquired using acceleration vibrating sensor;
(2) frequency domain information is built, and using Fast Fourier Transform (FFT), obtains the frequency domain information of acquisition signal;
(3) tensor model construction is modeled as a N by multiple dimension physical signals such as time, frequency and data channel
Rank tensor form;
X=G ×1P(1)×2P(2)×3P(3)…×nP(n)
In formula,Factor matrix, represent each mould it is important into
It is grouped as,It is core tensor, IiRepresent each dimension of tensor;
(4) Truncation Parameters are chosen, and using Higher-order Singular value decomposition side is blocked, solve tensor model:
In formula, λ is Truncation Parameters, σiIt is singular value, μiIt is left singular vector, viIt is right singular vector, b is measurement error
Constant, fiFor filtering factor;
(5) target tensor reconstructs, and target tensor is rebuild using rebuilding target tensor using Truncation Parameters are chosen:
In formula,The new factor matrix being made of Truncation Parameters,It is new core tensor,It is new target tensor, according to the teaming method inverse transformation of the former tensor signal that obtains that treated
Vibration signal can be expressed asλiIt is the Truncation Parameters of each dimension of the tensor asked for.
Compared with prior art, the present invention has following features:
1st, signal modeling can be solved the problems, such as into big data by tensor tool into tensor form in higher dimensional space;
2nd, using L-curve mode, to tensor, sparse ingredient is trimmed automatically, to reduce data dimension;
3rd, signal processing method of the exploratory development based on tensor analysis, it will provided for fields such as multidimensional data processing wide
Apply stage.
Description of the drawings
Fig. 1 is a kind of mechanical multidimensional processiug flow chart based on tensor analysis.
Fig. 2 is a kind of vibration multidimensional signal schematic diagram based on tensor analysis.
Fig. 3 is the tensor analysis Truncation Parameters schematic diagram that L-curve solves.
Fig. 4 is time domain waveform comparison diagram before and after embodiment middle gear case gear crack vibration signal noise reduction.
Fig. 5 is embodiment middle gear case gear crack vibration signal tensor resolution spectral contrast figure before and after the processing.
Specific embodiment
Vibration signal to be acquired in gear-box below, for carrying out signal de-noising by building multidimensional tensor model,
The present invention is described in detail.The entire flow figure of the invention is as shown in Figure 1, detailed process is:
1st, signal acquisition.Gear-box time domain vibration signal is acquired using acceleration vibrating sensor.In the present embodiment, if
The standby motor by 3HP drives, and experiment gear is mounted on the input shaft connect with associated electric motor, using VQ data collecting systems (packet
Include computer, data collecting instrument and NI capture cards) by the piezoelectric acceleration transducer that is mounted on 2 grades of parallel-shaft gearboxes
The vibration data in gear-box can be acquired.It is tooth root Gear with Crack that gear is tested used in experimentation, acquires vibration signal
When input shaft rotational frequency be 49.78Hz, sample frequency is 2.56kHz.It is observed from vibration signal time domain waveform, it can be with
Find out that signal is more mixed and disorderly, cannot get useful diagnostic message.
2nd, frequency domain information is built.Using Fast Fourier Transform (FFT), the frequency domain information of acquisition signal is obtained, is follow-up tensor structure
It builds and another dimensional information is provided.
3rd, tensor model construction.Using Tucker decomposition models (1), pass through multiple dimensions such as time, frequency and data channel
Degree physical signal is modeled as a N rank tensor form;
X=G ×1P(1)×2P(2)×3P(3)…×nP(n) (1)
In formula,Factor matrix, represent each mould it is important into
It is grouped as,It is core tensor.Vibration signal can be expressed asIn formula, IiIt represents to open
Measure each dimension.Fig. 2 is a kind of vibration signal schematic diagram based on tensor analysis;It is decomposed by Tucker and establishes vibration signal
Into tensor formVibration signal can be expressed asIn formula, ItRepresent time dimension, IcIt represents
Data channel dimension, IsRepresent frequency dimension.
4th, Truncation Parameters are chosen.Tensor mould is solved using Higher-order Singular value decomposition method (Truncated HOSVD) is blocked
Type.For the sparse intrinsic propesties of high dimensional data, remove the garbage ingredient in data using data truncation mode, and then realize
The dimension-reduction treatment of big data.Wherein data are blocked, need to carry out data automatic identification.This patent proposes bent using L
Collimation method asks for Truncation Parameters.
In formula, λ is Truncation Parameters, σiIt is singular value, μiIt is left singular vector, viIt is right singular vector.Using L-curve method
Solve the Truncation Parameters λ of three dimensions of three rank tensorst, λc, λsCurve, correspondingly λtIt is the time Truncation Parameters asked for, λcIt is
The Truncation Parameters for the data channel asked for, λsIt is the frequency Truncation Parameters asked for.In the present embodiment, λc=8.1247, λs=1,
λt=6.8819, as shown in Figure 3.It is respective integer value to three parameter roundings, respectivelyJust
In subsequently to three tensors progress dimension reductions.
5th, target tensor reconstructs.Target tensor is rebuild using weight using formula (3) and (4) using Truncation Parameters are chosen
New structure target tensor.
In formula,The new factor matrix being made of Truncation Parameters,It is new core tensor,It is new target tensor.According to the teaming method inverse transformation of the former tensor signal that obtains that treated
Then vibration signal can be expressed asλiIt is the Truncation Parameters of each dimension of the tensor asked for.
It is difficult from there are random noise time domain waveform shown in original vibration signal such as Fig. 4 (a) of acquisition in the present embodiment
To identify useful information ingredient.It by using tensor resolution, is trimmed in higher dimensional space, obtains filtered signal such as Fig. 4
(b) shown in.This it appears that shock characteristic from Fig. 4 (b), it was demonstrated that the validity of proposed method.From being given from Fig. 5
The front and rear signal spectrum comparison diagram of reason, it can be seen that the on the one hand method therefor of this patent effectively puies forward initial data
On the other hand pure processing saves the detail section of data as much as possible again.
Claims (1)
1. a kind of mechanical multidimensional big data processing method based on tensor resolution, which is characterized in that this method comprises the following steps:
(1) gear-box time domain vibration signal is acquired using acceleration vibrating sensor;
(2) frequency domain information is built, and using Fast Fourier Transform (FFT), obtains the frequency domain information of acquisition signal;
(3) tensor model construction is modeled as a N rank by multiple dimension physical signals such as time, frequency and data channel
Amount form;
X=G ×1P(1)×2P(2)×3P(3)…×nP(n)
In formula,It is factor matrix, represents the important component group of each mould
Into,It is core tensor, IiRepresent each dimension of tensor;
(4) Truncation Parameters are chosen, and using Higher-order Singular value decomposition side is blocked, solve tensor model:
In formula, λ is Truncation Parameters, σiIt is singular value, μiIt is left singular vector, viIt is right singular vector, b is measurement error constant,
fiFor filtering factor.
(5) target tensor reconstructs, and target tensor is rebuild using rebuilding target tensor using Truncation Parameters are chosen:
In formula,The new factor matrix being made of Truncation Parameters,It is
New core tensor,It is new target tensor, according to the teaming method inverse transformation of the former tensor signal that obtains that treatedIt shakes
Dynamic signal can be expressed asλiIt is the Truncation Parameters of each dimension of the tensor asked for.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810061289.6A CN108267311A (en) | 2018-01-22 | 2018-01-22 | A kind of mechanical multidimensional big data processing method based on tensor resolution |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810061289.6A CN108267311A (en) | 2018-01-22 | 2018-01-22 | A kind of mechanical multidimensional big data processing method based on tensor resolution |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108267311A true CN108267311A (en) | 2018-07-10 |
Family
ID=62776436
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810061289.6A Pending CN108267311A (en) | 2018-01-22 | 2018-01-22 | A kind of mechanical multidimensional big data processing method based on tensor resolution |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108267311A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111462106A (en) * | 2020-04-09 | 2020-07-28 | 中山易美杰智能科技有限公司 | Method for generating tensor for recognizing input of deep learning image and application of tensor |
CN114235413A (en) * | 2021-12-28 | 2022-03-25 | 频率探索智能科技江苏有限公司 | Method for constructing three-order tensor model of multi-path signal |
CN114235412A (en) * | 2021-12-28 | 2022-03-25 | 频率探索智能科技江苏有限公司 | Third order tensor rank decomposition method |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104331404A (en) * | 2013-07-22 | 2015-02-04 | 中国科学院深圳先进技术研究院 | A user behavior predicting method and device based on net surfing data of a user's cell phone |
CN104751169A (en) * | 2015-01-10 | 2015-07-01 | 哈尔滨工业大学(威海) | Method for classifying rail failures of high-speed rail |
CN105160699A (en) * | 2015-09-06 | 2015-12-16 | 电子科技大学 | Tensor-approximation-based multi-solution body drawing method of mass data |
CN105262441A (en) * | 2015-09-08 | 2016-01-20 | 西安交通大学 | Infrared image-based photovoltaic array fault grading method |
CN105389585A (en) * | 2015-10-20 | 2016-03-09 | 深圳大学 | Random forest optimization method and system based on tensor decomposition |
CN106646595A (en) * | 2016-10-09 | 2017-05-10 | 电子科技大学 | Earthquake data compression method based on tensor adaptive rank truncation |
CN107329933A (en) * | 2017-07-14 | 2017-11-07 | 北京知觉科技有限公司 | Fault detection method and device based on Fibre Optical Sensor vibration signal |
CN107507253A (en) * | 2017-08-15 | 2017-12-22 | 电子科技大学 | Based on the approximate more attribute volume data compression methods of high order tensor |
CN107515843A (en) * | 2017-09-04 | 2017-12-26 | 四川易诚智讯科技有限公司 | Based on the approximate anisotropy data compression method of tensor |
-
2018
- 2018-01-22 CN CN201810061289.6A patent/CN108267311A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104331404A (en) * | 2013-07-22 | 2015-02-04 | 中国科学院深圳先进技术研究院 | A user behavior predicting method and device based on net surfing data of a user's cell phone |
CN104751169A (en) * | 2015-01-10 | 2015-07-01 | 哈尔滨工业大学(威海) | Method for classifying rail failures of high-speed rail |
CN105160699A (en) * | 2015-09-06 | 2015-12-16 | 电子科技大学 | Tensor-approximation-based multi-solution body drawing method of mass data |
CN105262441A (en) * | 2015-09-08 | 2016-01-20 | 西安交通大学 | Infrared image-based photovoltaic array fault grading method |
CN105389585A (en) * | 2015-10-20 | 2016-03-09 | 深圳大学 | Random forest optimization method and system based on tensor decomposition |
CN106646595A (en) * | 2016-10-09 | 2017-05-10 | 电子科技大学 | Earthquake data compression method based on tensor adaptive rank truncation |
CN107329933A (en) * | 2017-07-14 | 2017-11-07 | 北京知觉科技有限公司 | Fault detection method and device based on Fibre Optical Sensor vibration signal |
CN107507253A (en) * | 2017-08-15 | 2017-12-22 | 电子科技大学 | Based on the approximate more attribute volume data compression methods of high order tensor |
CN107515843A (en) * | 2017-09-04 | 2017-12-26 | 四川易诚智讯科技有限公司 | Based on the approximate anisotropy data compression method of tensor |
Non-Patent Citations (2)
Title |
---|
吕小光: "结构矩阵计算及在数字图像复原中的应用", 《中国博士学位论文全文数据库 信息科技辑》 * |
蒋涉权: "基于张量分析的麦克风阵列语音信号降噪方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111462106A (en) * | 2020-04-09 | 2020-07-28 | 中山易美杰智能科技有限公司 | Method for generating tensor for recognizing input of deep learning image and application of tensor |
CN114235413A (en) * | 2021-12-28 | 2022-03-25 | 频率探索智能科技江苏有限公司 | Method for constructing three-order tensor model of multi-path signal |
CN114235412A (en) * | 2021-12-28 | 2022-03-25 | 频率探索智能科技江苏有限公司 | Third order tensor rank decomposition method |
CN114235413B (en) * | 2021-12-28 | 2023-06-30 | 频率探索智能科技江苏有限公司 | Method for constructing multi-channel signal third-order tensor model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107192553B (en) | Gear-box combined failure diagnostic method based on blind source separating | |
CN105115594B (en) | Gear-box vibration signal fault signature extracting method based on Wavelet Entropy and information fusion | |
CN104215456B (en) | Plane clustering and frequency-domain compressed sensing reconstruction based mechanical fault diagnosis method | |
CN104807534B (en) | Equipment eigentone self study recognition methods based on on-line vibration data | |
CN103217213B (en) | Modal parameter identification method based on response signal time-frequency joint distribution characteristics | |
CN108152033A (en) | A kind of compound Weak fault diagnostic method of the gear-box of sparse disjunctive model | |
CN108267311A (en) | A kind of mechanical multidimensional big data processing method based on tensor resolution | |
CN105928702B (en) | Variable working condition box bearing method for diagnosing faults based on form PCA | |
DE102012109393A1 (en) | Apparatus and method for evaluating vibrations | |
CN113405795B (en) | Method for identifying weak faults of joint RV reducer | |
CN108801630A (en) | The gear failure diagnosing method of single channel blind source separating | |
CN105043767A (en) | Rolling ball bearing outer ring spalling fault double-impact feature extraction method and rolling ball bearing outer ring spalling fault double-impact feature extraction system | |
CN109520611B (en) | Monitoring method for working condition of earthquake simulation vibration table | |
CN103278867A (en) | Automatic meteorological station data quality control method based on empirical mode decomposition | |
CN106383028A (en) | Gear case fault diagnosis method | |
CN103926097B (en) | Method for collecting and extracting fault feature information of low-speed and heavy-load device | |
CN112733612A (en) | Cross-domain rotating machinery fault diagnosis model establishing method and application thereof | |
CN110782041B (en) | Structural modal parameter identification method based on machine learning | |
CN107966287A (en) | A kind of adaptive dynamoelectric equipment Weak fault feature extracting method | |
CN113793263B (en) | Parallel residual error network high-resolution image reconstruction method for multi-scale cavity convolution | |
CN109946076B (en) | Planetary wheel bearing fault identification method of weighted multi-scale dictionary learning framework | |
CN108181098A (en) | A kind of high pedestal jib crane low-speed heave-load unit failure feature extracting method | |
CN111189646A (en) | Vehicle NVH self-diagnosis method and device, vehicle, controller and medium | |
CN114964769A (en) | Wind power gear box vibration signal fault diagnosis method | |
CN103152298B (en) | Blind signal reconstruction method based on distribution-type compressed sensing system |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180710 |
|
RJ01 | Rejection of invention patent application after publication |