CN109490793A - A kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM - Google Patents
A kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM Download PDFInfo
- Publication number
- CN109490793A CN109490793A CN201811185952.XA CN201811185952A CN109490793A CN 109490793 A CN109490793 A CN 109490793A CN 201811185952 A CN201811185952 A CN 201811185952A CN 109490793 A CN109490793 A CN 109490793A
- Authority
- CN
- China
- Prior art keywords
- lstm
- power supply
- wavelet decomposition
- voltage power
- high voltage
- 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
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/40—Testing power supplies
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
Abstract
The invention belongs to the test devices of electrical property；The detection device technical field of electric fault discloses a kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM；By the status change data { y of X-ray high voltage power supply1,y2,...,yt‑1,ytIt is used as initiation sequence, wavelet decomposition is carried out to the sequence at its preceding t-1 moment respectively, obtains the subsequence { D of wavelet decomposition under different scalej, j=1,2 ..., n }；LSTM model is established respectively using the data at the t-1 moment of n subsequence to be trained and predict, obtains the prediction result of each subsequence t moment；The prediction result of each subsequence t moment is subjected to linear superposition, obtains the predicted value of the t moment of the power supply status sequence；The relative error between prediction result and true value is calculated, prediction result is verified and is evaluated.The present invention improves precision of prediction and generalization ability with higher using the advantage of LSTM model by wavelet decomposition, has very great society value and realistic meaning.
Description
Technical field
The invention belongs to the test devices of electrical property；The detection device technical field more particularly to one kind of electric fault are based on
The X-ray high voltage power supply failure prediction method of wavelet decomposition and LSTM.
Background technique
Currently, the prior art commonly used in the trade is such that the development with the information age, industrial development are also more and more existing
Dai Hua, every field play the continuous improvement of various types of Large Complex Equipment performances of significant role and answering for system composition
Polygamy is continuously increased, so that the security reliability of large complicatedization equipment, maintainability, all the problems such as failure predication and diagnosis
Such as aviation, space flight, communication and the various fields of industrial application are become more and more important.Therefore, based on the data of magnanimity, to equipment
The prediction of failure causes serious safety and reliability consequence for reducing equipment failure, shortens downtime, reduces maintenance
Expense improves working efficiency and is of great significance.X-ray high voltage power supply is the main component of X-ray electronic system, and performance is determined
The service life of X-ray application system is determined.The high frequency miniaturization of X-ray high voltage power supply is its main development trend, but
It is high to easily lead to high-voltage power circuit integrated level, complexity, small fault may cause catastrophic accident.Traditional failure predication
Technology refers to diagnosis process using knowledge processing and knowledge reasoning as core to complete failure predication, and major defect includes: that search is empty
Between it is big, processing speed is slow, precision of prediction difference etc.；The method that traditional failure predication technology mainly uses mathematical prediction, such as mould
Paste theory and gray level model.Fuzzy theory is based on fuzzy set (fuzzy set), and superiority is mainly reflected in processing
Time-varying, time lag and the non-linear aspect of complication system.Gray scale prediction is established on the basis of gray theory, is an index letter
Number, since gray scale prediction model is to lead to too small amount of, imperfect information founding mathematical models and a kind of method made a prediction,
Cause gray level model very poor to the long-term forecast precision of failure.The experience and knowledge of expert can also be utilized by expert system
The system of foundation is predicted, using the knowledge reasoning in artificial intelligence, solves the problems, such as some professions, such as database, reasoning
Machine, knowledge base, but since expert system needs a large amount of practice and accumulation, cause the development cycle very long.A large amount of engineering
The task of habit is all the relevant input of processing timing, such as video analysis, music retrieval, relevant prediction of timing etc..
In conclusion problem of the existing technology is: long-term forecast precision of traditional failure predication technology to failure
It is very poor；Expert system needs a large amount of practice and accumulation, development cycle very long.
Solve the difficulty and meaning of above-mentioned technical problem:
Failure predication technology reduces the side such as life cycle cost and raising maintenance support efficiency in the safety for improving equipment
Face has played important function；The failure being likely to occur in the future is predicted, proposes to warn to user in time when predicting failure
It accuses, so that user can take measures that great serious accident is avoided to occur, has to existing system administration and maintenance and open
Invasive effect reaches timely failure predication and to the effective health control of electronic system.
Summary of the invention
In view of the problems of the existing technology, the X-ray high pressure based on wavelet decomposition and LSTM that the present invention provides a kind of
Power source failure prediction method.
The invention is realized in this way a kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM,
The X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM includes:
Step 1, by the status change data { y of X-ray high voltage power supply1,y2,...,yt-1,ytIt is used as initiation sequence, point
The other sequence to its preceding t moment carries out wavelet decomposition, obtains the subsequence { D of wavelet decomposition under different scalej, j=1,
2,...,n}；
Step 2 establishes LSTM model using the data at the t-1 moment of n subsequence respectively and is trained and predicts,
Obtain the prediction result of each subsequence t moment；
The prediction result of each subsequence t moment is carried out linear superposition, obtains the power supply status sequence by step 3
The predicted value of t moment；
Step 4 calculates the relative error between prediction result and true value, is verified and evaluated to prediction result.
Further, the step 1 selects DB4 small echo, carries out 3 layers of wavelet decomposition to original series, obtains 3 high frequency
Sequence D1、D2、D3And low frequency sequence A3。
Further, the step 2 is further comprising the steps of:
3a) data are extracted: the length that sliding window is arranged is 20, sliding step 20, successively in subsequenceUpper extraction institute
Need training data；
3b) network training: LSTM carries out network using BPTT (Back Propagation Trough Time) algorithm
Training, network inputs include: the input value of current time network, the output valve of previous moment network, previous moment memory cell
State；Network output includes: output valve, the state of current time memory cell of current time network.
Further, the step 2 further includes the data that following calculation formula predicts each subsequence t moment:
Wherein, wh1And b1Respectively the first weight and the first biasing, w12And b2Respectively the second weight and the second biasing, w23
And b3Respectively third weight and third biasing, s indicates vector Ht(ht1,ht2,...,hts) include element number, and s is big
In equal to 1.
Further, the Relu function is defined as:
Reluf (x)=max (0, x)；
Vector Ht(ht1,ht2,...,hts) in each element ht1,ht2,...,htsPass through iterative calculation:
it=sigmoid (Whiht-1+WxiXt)；
ft=sigmoid (Whfht-1+WxfXt)；
ct=ft·ct-1+it·tanh(Whcht-1+WxcXt)；
ot=sigmoid (Whoht-1+WhxXt+Wcoct)；
ht=ot·tanh(ct)；
Wherein, i, f, o, c respectively indicate input gate, forget door, out gate, cell state；XtData before t moment
The sequence of the continuous data composition of middle selection, Whi、Wxi、Whf、Wxf、Who、Whx、Wco、Whc、WxcRespectively different weights, h1For
Initial state value is 0.
Further, the step 4 uses error criterion calculation formula:
Wherein, ytFor true value,For predicted value.
Another object of the present invention is to provide the X-ray high voltage power supplies described in a kind of application based on wavelet decomposition and LSTM
The X-ray high voltage power supply of failure prediction method.
Another object of the present invention is to provide the X-ray high voltage power supplies described in a kind of application based on wavelet decomposition and LSTM
The high-voltage power supply control system of failure prediction method.
In conclusion advantages of the present invention and good effect are as follows: the present invention is carried out using signal of the wavelet decomposition to acquisition
Processing, reduces the dimension of fault-signal, while can filter out extra signal component, prominent trouble unit, but not to letter
Number information for being included damages, and substantially increases the accuracy rate and diagnosis speed of fault diagnosis in practical applications；
The present invention predicts power supply status using depth network model LSTM, can reinforce subsequent back end pair
The back end perception of front, may be implemented to make full use of measurement data, greatly improve forecasting efficiency and accurate
Degree, while generalization ability with higher have great social value and realistic meaning.The present invention, which mainly explores, solves high pressure
The effective ways of power source failure prediction contain failure to construct On-line Fault control mathematical model, ensure equipment safety fortune
Row.
The method that the present invention uses deep learning will be very big using LSTM network using the powerful ability of deep learning
The precision of prediction is improved in degree.The relevant method of deep learning will provide new better solution to the problem of power source failure prediction
Method.Wherein LSTM (Long Short-Term Memory) model is a kind of time recurrent neural network, and a kind of most effective
Fault prediction model, allow network accumulating information in a longer period of time, and can solve gradient explosion and gradient disappears
The problem of mistake.
Detailed description of the invention
Fig. 1 is the X-ray high voltage power supply failure prediction method provided in an embodiment of the present invention based on wavelet decomposition and LSTM
Flow chart.
Fig. 2 is the X-ray high voltage power supply failure prediction method provided in an embodiment of the present invention based on wavelet decomposition and LSTM
Implementation flow chart.
Fig. 3 is LSTM cell schematics provided in an embodiment of the present invention.
Fig. 4 is LSTM prediction model schematic diagram provided in an embodiment of the present invention.
Fig. 5 is the predicted value and actual value comparison schematic diagram provided in an embodiment of the present invention obtained applied to current sequence.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
The present invention is based on the X-ray high voltage power supply failure prediction methods of wavelet decomposition and LSTM, to make up traditional failure
Predicting Technique is very poor to the long-term forecast precision of failure；Expert system needs a large amount of practice and accumulation, development cycle very long
Problem；The present invention has stronger generalization ability, realizes higher precision of prediction；Electricity based on a large amount of X-ray high voltage power supply
Delta data and temperature variation data are flowed, the X-ray high voltage power supply fault prediction model based on wavelet decomposition and LSTM is proposed.
Application principle of the invention is explained in detail with reference to the accompanying drawing.
As shown in Figure 1, the X-ray high voltage power supply failure predication provided in an embodiment of the present invention based on wavelet decomposition and LSTM
Method the following steps are included:
S101: by the status change data { y of X-ray high voltage power supply1,y2,...,yt-1,ytIt is used as initiation sequence, respectively
Wavelet decomposition is carried out to the sequence at its preceding t moment, obtains the subsequence { D of wavelet decomposition under different scalej, j=1,2 ...,
n}；
S102: LSTM model is established respectively using the data at the t-1 moment of n subsequence and is trained and predicts, is obtained
To the prediction result of each subsequence t moment；
S103: the prediction result of each subsequence t moment is subjected to linear superposition, obtains the t of the power supply status sequence
The predicted value at moment
S104: the relative error between prediction result and true value is calculated, prediction result is verified and is evaluated.
Application principle of the invention is further described with reference to the accompanying drawing.
As shown in Fig. 2, the X-ray high voltage power supply failure predication provided in an embodiment of the present invention based on wavelet decomposition and LSTM
Method the following steps are included:
Step 1, by the sequence { y at t-1 moment1,y2,...,yt-1,ytWavelet decomposition is carried out, obtain n subsequence
{Dj, j=1,2 ..., n }, wherein each subsequence is made of the data at corresponding t-1 moment:
It carries out being related to following two aspects when wavelet decomposition:
First is that the selection of wavelet function, the big more options in the time series predicting model based on wavelet decomposition
The small echo of Daubechies series, the present invention select DB4 small echo；
Second is that the determination of Decomposition order, the number of plies determines that there is no a specific methods, in the time based on wavelet decomposition
The common number of plies determines that method is by testing different Decomposition orders on sequence prediction, and then selection is most suitable for the ginseng of the model
Number.Decomposition order is more, and noise can be fewer, and signal can be smoother, but some useful information can also lose, to cause
The decline of precision., whereas if Decomposition order is inadequate, our signal just preserves more noises, also has an impact to precision.
In view of deep learning has stronger learning ability to abstract characteristics, Decomposition order of the present invention is 3 layers.
In conclusion the present invention selects DB4 small echo, 3 layers of wavelet decomposition are carried out to original series, obtain 3 sub- sequences of high frequency
Arrange D1、D2、D3And low frequency sequence A3。
Step 2 establishes LSTM model using the data at the t-1 moment of n subsequence, by each subsequence respectively
It is divided into training set and forecast set, training set is trained using LSTM network, and is predicted with forecast set, is obtained each
The prediction result of subsequence t moment
2a) data are extracted: the length that sliding window is arranged is 20, sliding step 20, successively in subsequenceUpper extraction institute
Need training data；
2b) network training: LSTM carries out network using BPTT (Back Propagation Trough Time) algorithm
Training, network inputs include: the input value of current time network, the output valve of previous moment network, previous moment memory cell
State；Network output includes: output valve, the state of current time memory cell of current time network.
During LSTM is trained by BPTT algorithm, two parts are updated comprising forward calculation and error-duration model, and
The square error of the output result of moment output layers all in LSTM and desired output result is defined as objective function.LSTM exists
Before progress forward calculation, it is necessary first to be initialized with every network weight of the minimum close to 0 to LSTM, then
It inputs training sample data and carries out forward calculation, obtain the output of output layer.After the output for obtaining LSTM by forward calculation, meter
It calculates objective function and network parameter is updated using back-propagation algorithm；
Predict that the calculation formula of the data of each subsequence t moment is as follows:
Wherein, wh1And b1Respectively the first weight and the first biasing, w12And b2Respectively the second weight and the second biasing, w23
And b3Respectively third weight and third biasing, s indicates vector Ht(ht1,ht2,...,hts) in include element number, and s
More than or equal to 1.
Also, Relu function is defined as:
Reluf (x)=max (0, x)；
Also, vector Ht(ht1,ht2,...,hts) in each element ht1,ht2,...,htsPass through following formula iteration meter
It calculates:
it=sigmoid (Whiht-1+WxiXt)；
ft=sigmoid (Whfht-1+WxfXt)；
ct=ft·ct-1+it·tanh(Whcht-1+WxcXt)；
ot=sigmoid (Whoht-1+WhxXt+Wcoct)；
ht=ot·tanh(ct)；
Wherein, i, f, o, c respectively indicate input gate, forget door, out gate, cell state；XtData before t moment
The sequence of the continuous data composition of middle selection, Whi、Wxi、Whf、Wxf、Who、Whx、Wco、Whc、WxcRespectively different weights, h1For
Initial state value is 0.
Fig. 3, which is shown, predicts LSTM cell schematics used in the present invention.Fig. 4 shows LSTM in the present invention and predicts mould
Type schematic diagram.As shown in figure 3, three kinds of doors more than being exactly in place of the difference of LSTM unit and tradition RNN (Input, Forget,
) and a memory unit (Memory Cell) Output.Wherein, A indicates that output gate, B indicate forget gate, C table
Show that input gate, D indicate memory cell.
As shown in figure 4, LSTM network model includes 4 layers, as described below.Wherein, B indicates LSTM layers, and A indicates the LSTM layers
In LSTM unit, C indicate hidden layer 1, E indicate hidden layer 2, F indicate output layer, D indicate neural unit.
(1) LSTM layers: having 100 LSTM units, this certain quantity also can be adjusted according to the actual situation；
(2): there are 60 neurons in hidden layer 1, maps for dimensionality reduction and to abstract feature, this certain quantity
Also it can be adjusted according to the actual situation；
(3) hidden layer 2: effect has 30 neurons with hidden layer 1, this certain quantity can also be according to the actual situation
It is adjusted；
(4) output layer: 1 neuron, as predicted value.
Above-mentioned LSTM model structure include 1 layer of output layer, 1 layer LSTM layers, 2 layers of hidden layer.Wherein 2 layers of hidden layer are used for
Feature learning and dimensionality reduction, 1 layer of output layer are used to export the prediction result corresponding to list entries.The LSTM structure ratio of this model
Traditional LSTM model structure is simplified, therefore training difficulty is lower than traditional LSTM, and has good generalization ability.
It should be noted that the parameter of above layers is all by compared to other parameters, having obtained from many experiments
Have the characteristics that trained difficulty is lower, generalization ability is stronger.
The prediction result of each subsequence t moment is carried out linear superposition, obtains the power supply status sequence by step 3
The predicted value of t momentWherein j-th of subsequence in prediction resultWeight coefficient αjThe method meter of energy accounting can be used
It obtains:
Jth subsequenceENERGY EjIt calculates as follows:
Weight coefficient αjIt can be obtained by following formula:
The predicted value of the t moment of power supply status sequenceIt calculates as follows:
Step 4 calculates the error criterion between predicted value and true value, is verified and evaluated to prediction result；
The present invention is using MAPE (mean absolute percentage error), MAE (mean absolute error) the two indexs to this hair
Bright prediction technique and other prediction techniques carry out application condition, and error criterion calculation formula is as follows:
Wherein, ytFor true value,For predicted value.
Application effect of the invention is explained in detail below with reference to emulation.
The present invention carries out experimental verification using two different current signals, and every kind of signal has 10000 data points, selects
Wherein 8000 data progress network trainings are taken, wherein 2000 data is chosen and carries out prediction of result.Experimental result such as Fig. 5 (a),
Shown in Fig. 5 (b), wherein red curve indicates true waveform, and green curve indicates the waveform of prediction.
Table 1 is the application condition of the method for the present invention and KF (Kalman filtering) method and BP method:
The simulation experiment result analysis:
The present invention passes through relative to other conventional methods such as KF (Kalman filtering), BP mind it can be seen from Fig. 5 and table 1
There is more acurrate prediction effect through network technique.The present invention can not only reach higher precision of prediction, additionally it is possible to make Modeling Calculation mistake
Journey is simply easily operated.Prediction technique of the invention enhances the study energy to curent change sequence details by wavelet decomposition
Power, and reinforce subsequent back end to the back end perception of front using LSTM model, it may be implemented to measurement number
According to make full use of, while generalization ability with higher greatly improves forecasting efficiency and accuracy.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (8)
1. a kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM, which is characterized in that described based on small
The X-ray high voltage power supply failure prediction method of Wave Decomposition and LSTM includes:
Step 1, by the status change data { y of X-ray high voltage power supply1,y2,...,yt-1,ytIt is used as initiation sequence, it is right respectively
The sequence at its preceding t moment carries out wavelet decomposition, obtains the subsequence { D of wavelet decomposition under different scalej, j=1,2 ...,
n}；
Step 2 establishes LSTM model using the data at the t-1 moment of n subsequence respectively and is trained and predicts, obtains
The prediction result of each subsequence t moment；
The prediction result of each subsequence t moment is carried out linear superposition, when obtaining the t of the power supply status sequence by step 3
The predicted value at quarter；
Step 4 calculates the relative error between prediction result and true value, is verified and evaluated to prediction result.
2. the X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM, feature exist as described in claim 1
In the step 1 selects DB4 small echo, carries out 3 layers of wavelet decomposition to original series, obtains 3 high frequency subsequence D1、D2、D3And
Low frequency sequence A3。
3. the X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM, feature exist as described in claim 1
In the step 2 is further comprising the steps of:
3a) data are extracted: the length that sliding window is arranged is 20, sliding step 20, successively in subsequenceInstruction needed for upper extraction
Practice data；
3b) network training: LSTM is trained network using BPTT algorithm, and network inputs include: current time network it is defeated
Enter value, the output valve of previous moment network, the state of previous moment memory cell；Network output includes: current time network
The state of output valve, current time memory cell.
4. the X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM, feature exist as described in claim 1
In the step 2 further includes the data that following calculation formula predicts each subsequence t moment:
Wherein, wh1And b1Respectively the first weight and the first biasing, w12And b2Respectively the second weight and the second biasing, w23And b3
Respectively third weight and third biasing, s indicates vector Ht(ht1,ht2,...,hts) include element number, and s be greater than etc.
In 1.
5. the X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM, feature exist as claimed in claim 3
In the Relu function is defined as:
Relu f (x)=max (0, x)；
Vector Ht(ht1,ht2,...,hts) in each element ht1,ht2,...,htsPass through iterative calculation:
it=sigmoid (Whiht-1+WxiXt)；
ft=sigmoid (Whfht-1+WxfXt)；
ct=ft·ct-1+it·tanh(Whcht-1+WxcXt)；
ot=sigmoid (Whoht-1+WhxXt+Wcoct)；
ht=ot·tanh(ct)；
Wherein, i, f, o, c respectively indicate input gate, forget door, out gate, cell state；XtIt is selected from the data before t moment
The sequence of the continuous data composition taken, Whi、Wxi、Whf、Wxf、Who、Whx、Wco、Whc、WxcRespectively different weights, h1It is initial
State value is 0.
6. the X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM, feature exist as described in claim 1
In the step 4 uses error criterion calculation formula:
Wherein, ytFor true value,For predicted value.
7. a kind of pre- using the X-ray high voltage power supply failure based on wavelet decomposition and LSTM described in Claims 1 to 5 any one
The X-ray high voltage power supply of survey method.
8. a kind of pre- using the X-ray high voltage power supply failure based on wavelet decomposition and LSTM described in Claims 1 to 5 any one
The high-voltage power supply control system of survey method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811185952.XA CN109490793A (en) | 2018-10-11 | 2018-10-11 | A kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811185952.XA CN109490793A (en) | 2018-10-11 | 2018-10-11 | A kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109490793A true CN109490793A (en) | 2019-03-19 |
Family
ID=65689688
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811185952.XA Pending CN109490793A (en) | 2018-10-11 | 2018-10-11 | A kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109490793A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112311488A (en) * | 2020-11-05 | 2021-02-02 | 上海应用技术大学 | Spectrum sensing method based on wavelet decomposition |
| CN112465264A (en) * | 2020-12-07 | 2021-03-09 | 湖北省食品质量安全监督检验研究院 | Food safety risk grade prediction method and device and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105559777A (en) * | 2016-03-17 | 2016-05-11 | 北京工业大学 | Electroencephalographic identification method based on wavelet packet and LSTM-type RNN neural network |
| CN107102969A (en) * | 2017-04-28 | 2017-08-29 | 湘潭大学 | The Forecasting Methodology and system of a kind of time series data |
| CN108520320A (en) * | 2018-03-30 | 2018-09-11 | 华中科技大学 | A kind of equipment life prediction technique based on multiple shot and long term memory network and Empirical Bayes |
-
2018
- 2018-10-11 CN CN201811185952.XA patent/CN109490793A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105559777A (en) * | 2016-03-17 | 2016-05-11 | 北京工业大学 | Electroencephalographic identification method based on wavelet packet and LSTM-type RNN neural network |
| CN107102969A (en) * | 2017-04-28 | 2017-08-29 | 湘潭大学 | The Forecasting Methodology and system of a kind of time series data |
| CN108520320A (en) * | 2018-03-30 | 2018-09-11 | 华中科技大学 | A kind of equipment life prediction technique based on multiple shot and long term memory network and Empirical Bayes |
Non-Patent Citations (1)
| Title |
|---|
| 杨洪富 等: "基于LSTM的航空发动机排气温度预测", 《航空计算技术》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112311488A (en) * | 2020-11-05 | 2021-02-02 | 上海应用技术大学 | Spectrum sensing method based on wavelet decomposition |
| CN112465264A (en) * | 2020-12-07 | 2021-03-09 | 湖北省食品质量安全监督检验研究院 | Food safety risk grade prediction method and device and electronic equipment |
| WO2022121312A1 (en) * | 2020-12-07 | 2022-06-16 | 湖北省食品质量安全监督检验研究院 | Food safety risk level prediction method and apparatus, and electronic device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | A directed acyclic graph network combined with CNN and LSTM for remaining useful life prediction | |
| Qu et al. | A neural-network-based method for RUL prediction and SOH monitoring of lithium-ion battery | |
| CN110321603B (en) | Depth calculation model for gas path fault diagnosis of aircraft engine | |
| CN113343633B (en) | Thermal runaway fault classification and risk prediction method and system for power lithium battery | |
| CN109993270A (en) | Lithium ion battery residual life prediction technique based on grey wolf pack optimization LSTM network | |
| CN108197648A (en) | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models | |
| CN109991685A (en) | A kind of precipitation prediction technique and device based on more LSTM Model Fusions | |
| Gong et al. | State-of-health estimation of lithium-ion batteries based on improved long short-term memory algorithm | |
| CN110070229A (en) | The short term prediction method of home electrical load | |
| CN109034054B (en) | Harmonic multi-label classification method based on LSTM | |
| CN108520320A (en) | A kind of equipment life prediction technique based on multiple shot and long term memory network and Empirical Bayes | |
| CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
| CN109242212A (en) | A kind of wind-powered electricity generation prediction technique based on change Mode Decomposition and length memory network | |
| CN107993012A (en) | A kind of adaptive electric system on-line transient stability appraisal procedure of time | |
| CN110222844A (en) | A kind of compressor performance prediction technique based on artificial neural network | |
| CN113591728A (en) | Electric energy quality disturbance classification method based on integrated deep learning | |
| CN113469470B (en) | Energy consumption data and carbon emission correlation analysis method based on electric brain center | |
| CN114660497A (en) | Lithium ion battery service life prediction method aiming at capacity regeneration phenomenon | |
| CN102034111A (en) | Method for identifying and detecting aircraft structural damage conditions in diversified way | |
| CN109490793A (en) | A kind of X-ray high voltage power supply failure prediction method based on wavelet decomposition and LSTM | |
| CN113935513A (en) | CEEMDAN-based short-term power load prediction method | |
| Hu et al. | A hybrid model for predicting the degradation trend of hydropower units based on deep learning | |
| Wei et al. | Research on group behavior model based on neural network computing | |
| CN112465136A (en) | Hydroelectric generating set degradation prediction method and system | |
| CN113884936B (en) | ISSA coupling DELM-based lithium ion battery health state prediction method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190319 |