CN108804850A - A kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction - Google Patents
A kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction Download PDFInfo
- Publication number
- CN108804850A CN108804850A CN201810675418.0A CN201810675418A CN108804850A CN 108804850 A CN108804850 A CN 108804850A CN 201810675418 A CN201810675418 A CN 201810675418A CN 108804850 A CN108804850 A CN 108804850A
- Authority
- CN
- China
- Prior art keywords
- parameter
- aero
- engine
- data
- transition state
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Feedback Control In General (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to aero-engine performance parameter prediction technical fields, provide a kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction.The aero-engine transition state accelerator test data that the present invention is provided using certain research institute establishes training dataset and test data set；Data space reconstruct based on autocoder carries out a liter dimension to data set；It uses and model parameter is optimized by the population optimizing algorithm of representative of particle cluster algorithm；Finally transition state performance parameter is returned using showing outstanding random forest regression algorithm to high dimensional data, from application of engineering project, realizes effective prediction in real time.
Description
Technical field
The invention belongs to aero-engine performance parameter prediction technical fields, and in particular to a kind of boat based on Space Reconstruction
Empty engine transition state accelerator critical performance parameters prediction technique.
Background technology
Aero-engine for a long time high temperature, high pressure, high speed complex environment in run, the possibility to break down is at any time
Between passage and increase.And the performance of transition state accelerator be then directly related to take off and accelerate flight course into
Row.Since aero-engine mechanism is extremely complex, transition state procedure parameter is modeled very difficult.Therefore, using based on data
The aero-engine performance parameter advanced prediction method of driving can carry out model structure to avoid to the engine mechanism of process complexity
It builds, while the parameter state of engine transition state accelerator can be forecast in advance, ensure the security of the lives and property.
Many domestic and foreign scholars have done phase in terms of the aero-engine transition state performance parameter prediction based on data-driven
Close work.But traditional prediction algorithm is more demanding to model parameter and input feature vector, usually before modeling will use seek
Excellent algorithm is adjusted parameter；The engine parameters feature of different model to the influence degree of model prediction accuracy not yet
Together, it needs to re-start feature selecting；Transition state accelerator Parameter Variation is complicated, shadow of the model parameter to precision of prediction
The degree of sound is larger；Traditional regression forecasting algorithm is poor to the performance of higher-dimension numerical example, and transition state accelerator data sample
The not nearly enough description aero-engine performance of space dimensionality；Model generalization ability is poor, when engine changes, needs again to mould
Shape parameter and input feature vector are selected, and increase manpower and financial resources consumption to a certain extent.
Therefore it is difficult for model parameter selection, the present invention proposes a kind of parameter prediction being based on random forest (RF) algorithm,
RF algorithms are compared with conventional machines learn regression algorithm, with insensitive to multicollinearity, precision of prediction is high, convergence rate
Soon, adjustment parameter is few and meaning it can be readily appreciated that better to the effect of high dimensional data performance and be not in over-fitting etc.
Advantage.Random forest is more and more applied because of efficient and accurate feature in all trades and professions.It is tired for feature selecting
Difficulty, the present invention propose that a kind of sparse autocoder (SAE) based on neural network carries out lifting dimension to input feature vector.With tradition
Feature selecting algorithm (such as PCA) compare, SAE can be adjusted the lifting of intrinsic dimensionality according to algorithm, for RF this
Kind plays the very high algorithm that can also show accurate prediction effect of independent variable dimension the effect for promoting precision of prediction.To two
Kind algorithm optimizes model parameter using parameter optimization algorithm, from practical implementation angle, to the pass of engine
Key performance parameter, such as rotational speed of lower pressure turbine rotor, delivery temperature parameter are predicted.
Invention content
The present invention is directed to deficiencies of the prior art, proposes a kind of aero-engine transition based on Space Reconstruction
State accelerator performance parameter prediction technique.
Technical scheme of the present invention：
A kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction, step is such as
Under：
The first step, Aero Engine Testing data prediction
(1) aero-engine transition state accelerator test data includes compressor inlet relative rotation speed PNNC2g, start
Machine inlet temperature T2, engine intake air pressure P2, blower outlet stagnation pressure P3, fuel flow WFB, fan physics rotating speed Nf, calm the anger
Machine physics rotating speed Nc, turbine-exit temperature T5, simulated flight height H and simulated flight Mach number Ma, totally 10 class parameter, as one
A sample；
(2) data storage and reading：Aero-engine transition state accelerator test data includes multiple boat hair test runs
Journey collection in worksite data combine the data of the multiple boat hair commissioning process collection in worksite of aero-engine accelerator, and
Unified storage, and establish aero-engine performance parameter experiment data warehouse；
(3) linear resampling：Aero-engine accelerator test data is analyzed, due to sampling time interval
It differs, therefore linear resampling method is used to carry out resampling processing to aero-engine accelerator test data, make data-signal
Sample frequency is identical；
(4) data screening and cleaning：To the aero-engine transition state accelerator test data after linear resampling into
Row visualization processing, the acceleration curve to obviously not meeting objective condition are cleared up；
Second step, the selection of random forest Parameters in Regression Model
There are two key parameters for random forest regression model, respectively：Ntree is regression tree in random forest regression model
Quantity, ntree is too small, and classification accuracy is relatively low, and ntree is excessive, calculate overlong time be unfavorable for predicting in real time；mtry
For the feature quantity of regression tree in random forest regression model, i.e., branch's number of each regression tree；Since random forest returns
It is discrete integer value to return the parameter of model, therefore two-dimensional grid search is selected to carry out traversal calculating to parameter ntree and mtry, is fitted
Response function (i.e. the return value of parameter optimization) selects MSE；The range of two-dimensional grid search optimizing is determined by following principle：
(1) ntree Search Ranges are determined by data error outside bag (OOB error rate), and the wherein outer data error of bag is
The error of recurrence generation is carried out as the data of single decision tree training sample to not being selected every time；To being predicted test data
Characteristic parameter rotational speed of lower pressure turbine rotor NcIt is such as attached with the change curve of the engine exhaust temperature EGT OOB and parameter ntree calculated
Shown in Fig. 2；Therefore ntree Search Ranges are determined as 50~500；
(2) mtry Search Ranges are determined as natural number 1 to test data feature sum；
Finally determine that random forest Parameters in Regression Model of the present invention is selected as ntree=300, mtry by grid optimizing algorithm
=D/3, wherein D are mode input variable number；
Third walks, and tranining database is built using sparse autocoder
After the parameter for determining random forest regression model, the relevant parameter of SAE is determined using SAE-RF mixed models；Because
10 parameter attributes of Aero Engine Testing data are difficult to meet aero-engine transition state parametric prediction model accuracy
It is required that therefore selection learns the rarefaction representation of Aero Engine Testing data, excavate more information from 10 dimension input quantities.Using
The SAE of 10-dim-10 structures builds the input vector of model.The particle cluster algorithm of discrete-continuous mixing is utilized in the present invention
(PSO) optimizing is carried out to SAE parameters, the important parameter of SAE includes learning rate α and reconstruct dimension dim；
It is as follows using the principle of particle cluster algorithm optimizing in the present invention：In 2 dimension parameter search spaces, there are one by n
Population X=(the X of parameter combination composition1, X2... Xn), wherein position of k-th of parameter combination in parameter search space
It is expressed as a 2 dimensional vector Xk=(xk1, xk2).If movement speed of k-th of parameter combination in search space is Vk=(Vk1,
Vk2)T, local optimum parameter is Pk=(Pk1, Pk2)T, global optimum's parameter of the parameter combination is Pg=(Pg1, Pg2)T.Every
In secondary iteration, the movement speed of parameter combination and the iterative formula of position are expressed as：
Wherein：W is inertia weight, and t is the number of current iteration, r1, r2For equally distributed random number in [0,1], c1,
c2For Studying factors constant.
K folding cross-validation methods are generally used for the estimation of the Generalization Ability in parameter selection, the particle based on discrete-continuous mixing
Group algorithm the parameter optimization of SAE-RF mixed models specific steps are as follows：
(1) initial position of one group { α, dim } as particle is randomly generated, determines inertia weight and Studying factors；
(2) training sample is equably divided into the k subset S not included mutually1, S2..., Sk；
(3) k for training SAE-RF mixed models, and is calculated as parameter according to the value of initial position where population
The average value of secondary accuracy rate, as K roll over the accuracy rate of cross validation；
(4) accuracy rate that K is rolled over cross validation calculates the local optimum of the generation population as the fitness of particle cluster algorithm
Position and global optimum position, iteration update position and speed；
(5) step (2) is repeated, until meeting fitness requirement or reaching maximum iteration；
(6) parameter optimization is completed, using result as the parameter of final SAE-RF mixed models；
4th step builds SAE-RF regression models, predicts Aero Engine Testing data, evaluation and foreca effect
Because magnitude difference is larger between the feature after the rarefaction representation of Aero Engine Testing data, therefore uses and be most worth method pair
Data sample is normalized, and avoids the model error caused by magnitude gap.Aviation is sent out according to the following formula in the present invention
In feature normalization to section [1,2] after the rarefaction representation of motivation transition state accelerator test data：
Aero-engine transition state accelerator test data after being reconstructed using dimension is to rotational speed of lower pressure turbine rotor NcWith hair
Motivation delivery temperature EGT carries out regression forecasting, calculates response evaluation index；
Main evaluation index includes：
(1) relative error (RE)
Relative error formula is：
Wherein,For i-th the moment sample predicted value, yiFor i-th the moment sample observation, N be the sample length
Degree；
(2) mean square error (MSE)
Mean square error is a kind of measurement of difference between response prediction sequence and observation sequence, is calculated according to following formula：
WhereinFor i-th the moment sample predicted value, yiFor i-th the moment sample observation, N be the sample length
Degree.
Beneficial effects of the present invention：The present invention is tested using the aero-engine transition state accelerator that certain research institute provides
Data establish training dataset and test data set；Data space reconstruct based on autocoder carries out a liter dimension to data set；
It uses and model parameter is optimized with the population optimizing algorithm that particle cluster algorithm (PSO) is representative；It finally utilizes to high dimension
Transition state performance parameter is returned according to the random forest regression algorithm for showing outstanding, from application of engineering project, is realized
Effective prediction in real time.
Description of the drawings
Fig. 1 is aero-engine transition state accelerator critical performance parameters prediction model Establishing process figure.
Fig. 2 is the outer data error of bag and model decision tree quantitative relation curve graph.
Fig. 3 is grid search random forest regression model optimized parameter result figure.
Fig. 4 is particle cluster algorithm optimizing parametric procedure curve graph, and upper figure is the particle cluster algorithm for predicting rotational speed of lower pressure turbine rotor
Optimum results figure, figure below are the particle cluster algorithm optimum results figure for predicting delivery temperature.
Fig. 5 is the prediction curve and observation curve of 10 groups of samples, and upper figure is rotational speed of lower pressure turbine rotor prediction result figure, and figure below is
Delivery temperature prediction result figure.
Fig. 6 is the predicted value and observation departure degree schematic diagram of 10 groups of samples, and upper figure is rotational speed of lower pressure turbine rotor predicted value
Departure degree schematic diagram, figure below are delivery temperature predicted value departure degree schematic diagram.
Fig. 7 is the relative error distribution map of 10 groups of samples, and upper figure is rotational speed of lower pressure turbine rotor Relative Error distribution map, under
Figure is delivery temperature Relative Error distribution map.
Fig. 8 is the forecasting sequence and observation sequence mean square error distribution map of 10 groups of samples, and upper figure is that rotational speed of lower pressure turbine rotor is pre-
Mean square error figure is surveyed, figure below is that delivery temperature predicts mean square error figure.
Specific implementation mode
Below in conjunction with attached drawing and technical solution, the specific implementation mode that further illustrates the present invention.
The data that the present invention uses are certain type aero-engine transition state accelerator racks that certain domestic research institute provides
100 groups of test data.
The first step, Aero Engine Testing data prediction
(1) Aero Engine Testing data include compressor inlet relative rotation speed PNNC2g, engine intake temperature T2, hair
Motivation inlet pressure P2, blower outlet stagnation pressure P3, fuel flow WFB, fan physics rotating speed Nf, compressor physics rotating speed Nc, whirlpool
Take turns outlet temperature T5, simulated flight height H, simulated flight Mach number Ma totally 10 groups of parameters；
(2) data integration：The txt file of 100 groups of data is read out and is integrated and unifies storage and establishes aeroplane engine
Machine test data data warehouse.
(3) resampling：Due to the sampling interval etc., resampling processing is carried out to data first.Concrete operation step is as follows,
Using the method for interpolation, new sample frequency will be drafted and be inserted into the time series of initial data as interpolation, statistics is specified to adopt
The number of initial data between sampling point.If including only an initial data, which is corresponded to as the sampled point
Data；If including two initial data, are used as the time point corresponding after the two initial data are done average value processing
Data；If not including initial data, a upper time point at the time point in time rating sequence and future time point were corresponded to
Data of the average value of data as the time point.
(4) data screening and cleaning：Visualization processing is carried out to data, to acceleration curve progress simple clustering and clearly
Reason.
Second step, the selection of random forest Parameters in Regression Model
The present invention determines that the parameter optimization range based on grid search, the Search Range of final selected ntree are by attached drawing 2
50~500, optimizing step-length is 10；The Search Range of selected mtry is 1~D, and optimizing step-length is 1.Wherein D is mode input vector
Dimension, i.e. the dimension of test data sample.3-fold cross-validation methods are used when calculating fitness function value, by 90 groups of experiments
Data sample is divided into 3 parts, carries out 3 prediction tasks every time：
Using 3 subtasks mean square error average value as the corresponding fitness function value of this group of parameter：
MSE=(mse1+mse2+mse3)/3
Final optimizing result is as shown in Fig. 3.Considered by the influence factors such as time cost and calculation amount, final argument selection
As a result it is ntree=300, mtry=D/3, wherein D is the dimension of mode input vector.
Third walks, and tranining database is built using sparse autocoder
The important parameter of sparse autocoder in the present invention includes learning rate α and reconstruct dimension dim, and wherein α is to connect
Continuous value, dim is discrete integer value, therefore discrete-continuous Hybrid Particle Swarm is used to carry out optimizing in two dimensions of parameter, and
Equally use 3-fold cross-validation methods.It is 10 that initial population group number, which is arranged, maximum iteration 50, particle initial bit
It sets and is randomly set to { [0,1], { 1,2 ... 50 } }, particle initial velocity is randomly provided it should be noted that limitation dim not got
Position less than 1 or more than 50, inertia weight are set asStudying factors are set as c1=c2=0.5+ln2.
SAE parameter optimization results are as shown in Fig. 4.Therefore office when selecting to reach the iterations of fitness function minimum value
Portion's optimal solution is as parametric results：Parameter is set as dim=46, α=0.7060 when predicting rotational speed of lower pressure turbine rotor；To exhaust temperature
Parameter is set as dim=20, α=1.7428 when degree prediction.
4th step：SAE-RF regression models are built, Aero Engine Testing data are predicted, evaluation and foreca effect
In view of the Aero Engine Testing data volume grade gap after reconstruct is larger, it is normalized using most value method
Processing accelerates its convergence rate and the predictablity rate caused by magnitude gap is avoided to decline.By the rarefaction representation of test data
Feature afterwards is normalized to as follows in section [1,2]：
It, will wherein 90 groups conduct training numbers in 100 groups of test datas that certain research institute of the country of the present invention provides
According to, remaining 10 groups are used as prediction data, be respectively completed in transition state accelerator key parameter rotational speed of lower pressure turbine rotor and row
The prediction task of temperature degree calculates relative error distribution and the mean square error of 10 groups of test datas, evaluation model prediction effect.
For the model predication value of rotational speed of lower pressure turbine rotor and experimental observation curve it can be seen from attached drawing 5 and attached drawing 6,
The two is close to coincideing, and predicted value is with observation almost without deviation；It is bent for the model predication value and experimental observation of delivery temperature
Line, effect is then slightly poorer to the prediction effect to rotational speed of lower pressure turbine rotor, and when the sampling test starts, predicted value and observation is inclined
It is larger from degree.Reason is that measurable test data feature is associated with closely with rotational speed of lower pressure turbine rotor, but aero-engine gas
Road thermal part is complicated, and thermodynamic relation is inherently difficult to describe, in addition the rigorous service condition of sensor.Attached drawing 7 is opened up
The 10 groups of relative error for being predicted each observation time point of sample distributions are shown.Even if can be seen that delivery temperature by attached drawing 8
Prediction effect is slightly poorer to rotational speed of lower pressure turbine rotor, but the mean square error of 10 groups of prediction data samples can be controlled in aeroplane engine completely
Below machine key performance parameter prediction software requirement.
The mean square error of 1 forecast sample of table
In conclusion based on the sparse autocoder of particle cluster algorithm optimizing by aero-engine transition state accelerator
After test data carries out dimension reconstruct, using random forest regression algorithm to key parameter such as rotational speed of lower pressure turbine rotor, delivery temperature
It can achieve the desired results etc. the accuracy rate predicted, the status predication in aero-engine can be applied to be led with fault diagnosis
Domain.
Claims (1)
1. a kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction, feature exist
In steps are as follows：
The first step, Aero Engine Testing data prediction
(1) aero-engine transition state accelerator test data includes compressor inlet relative rotation speed PNNC2g, engine intake
Temperature T2, engine intake air pressure P2, blower outlet stagnation pressure P3, fuel flow WFB, fan physics rotating speed Nf, compressor physics
Rotating speed Nc, turbine-exit temperature T5, simulated flight height H and simulated flight Mach number Ma, totally 10 class parameter, as a sample；
(2) data storage and reading：Aero-engine transition state accelerator test data includes that multiple boat hair commissioning process are existing
Field gathered data combines the data of the multiple boat hair commissioning process collection in worksite of aero-engine accelerator, and uniformly
Storage, and establish aero-engine performance parameter experiment data warehouse；
(3) linear resampling：Aero-engine accelerator test data is analyzed, not due to sampling time interval
Deng, therefore linear resampling method is used to carry out resampling processing to aero-engine accelerator test data, so that data-signal is adopted
Sample frequency is identical；
(4) data screening and cleaning：It can to the aero-engine transition state accelerator test data progress after linear resampling
It is handled depending on change, the acceleration curve to obviously not meeting objective condition is cleared up；
Second step, the selection of random forest Parameters in Regression Model
There are two key parameters for random forest regression model, respectively：Ntree is the number of regression tree in random forest regression model
Amount；Mtry is the feature quantity of regression tree in random forest regression model, i.e., branch's number of each regression tree；Select two dimension
Grid search carries out traversal calculating to parameter ntree and mtry, and fitness function selects MSE；The model of two-dimensional grid search optimizing
It encloses and is determined by following principle：
(1) ntree Search Ranges are determined by data error outside bag, wherein the outer data error of bag is to not being selected conduct every time
The data of single decision tree training sample return the error of generation；Turn to being predicted test data characteristic parameter low pressure rotor
Fast NcWith the change curve of the engine exhaust temperature EGT OOB and parameter ntree calculated；Therefore ntree Search Ranges are determined as 50
~500；
(2) mtry Search Ranges are determined as natural number 1 to test data feature sum；
Finally determine that random forest Parameters in Regression Model is selected as ntree=300, mtry=D/3 by grid optimizing algorithm, wherein
D is mode input variable number；
Third walks, and tranining database is built using sparse autocoder
After the parameter for determining random forest regression model, the relevant parameter of SAE is determined using SAE-RF mixed models；Using 10-
The SAE of dim-10 structures builds the input vector of model；SAE parameters are carried out using the particle cluster algorithm of discrete-continuous mixing
The important parameter of optimizing, SAE includes learning rate α and reconstruct dimension dim；
It is specific as follows using particle cluster algorithm optimizing：In 2 dimension parameter search spaces, there are one what is be made of n parameter combination
Population X=(X1, X2..., Xn), wherein position of k-th of parameter combination in parameter search space is expressed as one 2
Dimensional vector Xk=(xk1, xk2)；If movement speed of k-th of parameter combination in parameter search space is Vk=(Vk1, Vk2)T,
Local optimum parameter is Pk=(Pk1, Pk2)T, global optimum's parameter of the parameter combination is Pg=(Pg1, Pg2)T；In each iteration
In, the movement speed of parameter combination and the iterative formula of position are expressed as：
Wherein：W is inertia weight, and t is the number of current iteration, r1, r2For equally distributed random number in [0,1], c1, c2To learn
Practise factor constant；
K rolls over cross-validation method and estimates for the Generalization Ability in parameter selection, the particle cluster algorithm based on discrete-continuous mixing
SAE-RF mixed model parameter optimizations, are as follows：
(1) initial position of one group { α, dim } as particle is randomly generated, determines inertia weight and Studying factors；
(2) training sample is equably divided into the k subset S not included mutually1, S2..., Sk；
(3) according to the value of initial position where population as parameter, for training SAE-RF mixed models, and k standard is calculated
The average value of true rate, as K roll over the accuracy rate of cross validation；
(4) accuracy rate that K is rolled over cross validation calculates the local optimum position of the generation population as the fitness of particle cluster algorithm
With global optimum position, iteration updates position and speed；
(5) step (2) is repeated, until meeting fitness requirement or reaching maximum iteration；
(6) parameter optimization is completed, using result as the parameter of final SAE-RF mixed models；
4th step builds SAE-RF regression models, predicts Aero Engine Testing data, evaluation and foreca effect
Data sample is normalized using most value method, avoids the model error caused by magnitude gap；According to the following formula
It will be in the feature normalization after the rarefaction representation of aero-engine transition state accelerator test data to section [1,2]：
Aero-engine transition state accelerator test data after being reconstructed using dimension is to rotational speed of lower pressure turbine rotor NcIt is arranged with engine
Temperature degree EGT carries out regression forecasting, calculates response evaluation index；
Main evaluation index includes：
(1) relative error RE
Relative error formula is：
Wherein,For i-th the moment sample predicted value, yiFor i-th the moment sample observation, N be the sample length；
(2) mean square error MSE
Mean square error is a kind of measurement of difference between response prediction sequence and observation sequence, is calculated according to following formula：
WhereinFor i-th the moment sample predicted value, yiFor i-th the moment sample observation, N be the sample length.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810675418.0A CN108804850B (en) | 2018-06-27 | 2018-06-27 | Method for predicting parameters of aircraft engine in transient acceleration process based on spatial reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810675418.0A CN108804850B (en) | 2018-06-27 | 2018-06-27 | Method for predicting parameters of aircraft engine in transient acceleration process based on spatial reconstruction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108804850A true CN108804850A (en) | 2018-11-13 |
| CN108804850B CN108804850B (en) | 2020-09-11 |
Family
ID=64072129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810675418.0A Active CN108804850B (en) | 2018-06-27 | 2018-06-27 | Method for predicting parameters of aircraft engine in transient acceleration process based on spatial reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108804850B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472105A (en) * | 2018-11-22 | 2019-03-15 | 上海华力微电子有限公司 | Semiconductor product yield Upper bound analysis method |
| CN109657945A (en) * | 2018-12-06 | 2019-04-19 | 华中科技大学 | A kind of industrial process fault diagnosis method based on data-driven |
| CN113065206A (en) * | 2021-03-24 | 2021-07-02 | 北京航空航天大学 | Transition state control method and device, electronic equipment and storage medium |
| CN113468803A (en) * | 2021-06-09 | 2021-10-01 | 淮阴工学院 | Improved WOA-GRU-based flood flow prediction method and system |
| CN113899557A (en) * | 2020-06-22 | 2022-01-07 | 中国航发商用航空发动机有限责任公司 | Method and device for determining characteristics of air system of aircraft engine |
| CN114675535A (en) * | 2022-03-07 | 2022-06-28 | 大连理工大学 | Aero-engine transition state optimization control method based on reinforcement learning |
| CN117740727A (en) * | 2024-02-19 | 2024-03-22 | 南京信息工程大学 | Textile component quantitative inversion method based on infrared hyperspectrum |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105608004A (en) * | 2015-12-17 | 2016-05-25 | 云南大学 | CS-ANN-based software failure prediction method |
| US20170061305A1 (en) * | 2015-08-28 | 2017-03-02 | Jiangnan University | Fuzzy curve analysis based soft sensor modeling method using time difference Gaussian process regression |
| CN106503788A (en) * | 2016-10-28 | 2017-03-15 | 中国矿业大学 | Least square method supporting vector machine Forecasting Methodology based on self adaptation population |
| CN106503459A (en) * | 2016-10-26 | 2017-03-15 | 南京信息工程大学 | A kind of improvement random forest temperature record method of quality control |
| WO2017098862A1 (en) * | 2015-12-08 | 2017-06-15 | 国立研究開発法人物質・材料研究機構 | Fuel oil discrimination sensor equipped with receptor layer composed of hydrocarbon-group-modified microparticles, and fuel oil discrimination method |
| CN106980877A (en) * | 2017-03-15 | 2017-07-25 | 陕西中爆安全网科技有限公司 | A kind of Prediction of Blasting Vibration method based on particle cluster algorithm Support Vector Machines Optimized |
| CN107273633A (en) * | 2017-06-29 | 2017-10-20 | 中南大学 | Varying delay method of estimation and flow time lag method of estimation is hydrocracked between multiple operation |
-
2018
- 2018-06-27 CN CN201810675418.0A patent/CN108804850B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170061305A1 (en) * | 2015-08-28 | 2017-03-02 | Jiangnan University | Fuzzy curve analysis based soft sensor modeling method using time difference Gaussian process regression |
| WO2017098862A1 (en) * | 2015-12-08 | 2017-06-15 | 国立研究開発法人物質・材料研究機構 | Fuel oil discrimination sensor equipped with receptor layer composed of hydrocarbon-group-modified microparticles, and fuel oil discrimination method |
| CN105608004A (en) * | 2015-12-17 | 2016-05-25 | 云南大学 | CS-ANN-based software failure prediction method |
| CN106503459A (en) * | 2016-10-26 | 2017-03-15 | 南京信息工程大学 | A kind of improvement random forest temperature record method of quality control |
| CN106503788A (en) * | 2016-10-28 | 2017-03-15 | 中国矿业大学 | Least square method supporting vector machine Forecasting Methodology based on self adaptation population |
| CN106980877A (en) * | 2017-03-15 | 2017-07-25 | 陕西中爆安全网科技有限公司 | A kind of Prediction of Blasting Vibration method based on particle cluster algorithm Support Vector Machines Optimized |
| CN107273633A (en) * | 2017-06-29 | 2017-10-20 | 中南大学 | Varying delay method of estimation and flow time lag method of estimation is hydrocracked between multiple operation |
Non-Patent Citations (2)
| Title |
|---|
| 刘继辉等: "基于随机森林回归的制丝过程参数影响权重分析", 《烟草科技》 * |
| 胡欢: "基于SQP方法的航空发动机过渡态最优控制研究", 《中国优秀硕士学位论文全文数据库.工程科技II辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472105A (en) * | 2018-11-22 | 2019-03-15 | 上海华力微电子有限公司 | Semiconductor product yield Upper bound analysis method |
| CN109472105B (en) * | 2018-11-22 | 2023-09-19 | 上海华力微电子有限公司 | Semiconductor product yield upper limit analysis method |
| CN109657945A (en) * | 2018-12-06 | 2019-04-19 | 华中科技大学 | A kind of industrial process fault diagnosis method based on data-driven |
| CN113899557A (en) * | 2020-06-22 | 2022-01-07 | 中国航发商用航空发动机有限责任公司 | Method and device for determining characteristics of air system of aircraft engine |
| CN113065206A (en) * | 2021-03-24 | 2021-07-02 | 北京航空航天大学 | Transition state control method and device, electronic equipment and storage medium |
| CN113468803A (en) * | 2021-06-09 | 2021-10-01 | 淮阴工学院 | Improved WOA-GRU-based flood flow prediction method and system |
| CN113468803B (en) * | 2021-06-09 | 2023-09-26 | 淮阴工学院 | WOA-GRU flood flow prediction method and system based on improvement |
| CN114675535A (en) * | 2022-03-07 | 2022-06-28 | 大连理工大学 | Aero-engine transition state optimization control method based on reinforcement learning |
| CN114675535B (en) * | 2022-03-07 | 2024-04-02 | 大连理工大学 | Aeroengine transition state optimizing control method based on reinforcement learning |
| CN117740727A (en) * | 2024-02-19 | 2024-03-22 | 南京信息工程大学 | Textile component quantitative inversion method based on infrared hyperspectrum |
| CN117740727B (en) * | 2024-02-19 | 2024-05-14 | 南京信息工程大学 | Textile component quantitative inversion method based on infrared hyperspectrum |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108804850B (en) | 2020-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11436395B2 (en) | Method for prediction of key performance parameter of an aero-engine transition state acceleration process based on space reconstruction | |
| CN108804850A (en) | A kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction | |
| CN111738512B (en) | Short-term power load prediction method based on CNN-IPSO-GRU hybrid model | |
| CN106446540B (en) | A kind of Wind turbines health status real time evaluating method | |
| CN107992648A (en) | A kind of Adaptive radial basis function neural network carries out aeroengine thrust algorithm for estimating | |
| CN112417769B (en) | Data-driven-based high-supercharged engine air passage structure optimization design method | |
| CN114462319B (en) | Active regulation and control method for combustion performance of aero-engine and intelligent prediction model | |
| CN101802728A (en) | Gaseous emission is carried out system and method based on the virtual sensing of experience set | |
| CN109460629A (en) | A kind of cooling fan performance optimization method based on approximate model method | |
| CN108427845A (en) | A kind of Pb-Zn deposits mining process carbon emission short term prediction method | |
| CN109255726A (en) | A kind of ultra-short term wind power prediction method of Hybrid Intelligent Technology | |
| CN114912195B (en) | Aerodynamic sequence optimization method for commercial vehicle | |
| CN107644297A (en) | A kind of energy-saving of motor system amount calculates and verification method | |
| CN116204779B (en) | Method, system and readable storage medium for judging operation state of energy storage salt cavern | |
| CN106909719A (en) | Integrated arithmetic of linearity regression | |
| CN111832839B (en) | Energy consumption prediction method based on sufficient incremental learning | |
| Shuang et al. | An adaptive compressor characteristic map method based on the Bezier curve | |
| Hosseinimaab et al. | Optimizing the performance of a single-shaft micro gas turbine engine by modifying its centrifugal compressor design | |
| CN111368370A (en) | Aircraft engine fault diagnosis method based on class-specific redirection extreme learning machine learning algorithm | |
| Sánchez Moreno et al. | Robustness of optimisation algorithms for transonic aerodynamic design | |
| Botticelli et al. | Application of machine learning to gasoline direct injection systems: Towards a data-driven development | |
| CN109190800A (en) | A kind of sea surface temperature prediction technique based on spark frame | |
| Zaglauer et al. | Evolutionary algorithms for the automatic calibration of simulation models for the virtual engine application | |
| CN115221782A (en) | Hybrid prediction method and system for energy consumption of large public building | |
| Qin | Software reliability prediction model based on PSO and SVM |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zhang Shuo Inventor after: Sun Ximing Inventor after: Sun Xiaoyu Inventor after: Li Jibang Inventor before: Sun Xiaoyu Inventor before: Zhang Shuo Inventor before: Li Jibang Inventor before: Sun Ximing |
|
| CB03 | Change of inventor or designer information |