CN109779791A - Abnormal data intelligent diagnosing method in a kind of solid propellant rocket - Google Patents
Abnormal data intelligent diagnosing method in a kind of solid propellant rocket Download PDFInfo
- Publication number
- CN109779791A CN109779791A CN201910224959.6A CN201910224959A CN109779791A CN 109779791 A CN109779791 A CN 109779791A CN 201910224959 A CN201910224959 A CN 201910224959A CN 109779791 A CN109779791 A CN 109779791A
- Authority
- CN
- China
- Prior art keywords
- abnormal
- neural network
- data
- parameters
- value
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 63
- 239000004449 solid propellant Substances 0.000 title abstract 2
- 238000013528 artificial neural network Methods 0.000 claims abstract description 44
- 230000008569 process Effects 0.000 claims abstract description 19
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 16
- 238000007476 Maximum Likelihood Methods 0.000 claims abstract description 6
- 238000012360 testing method Methods 0.000 claims description 59
- 239000007787 solid Substances 0.000 claims description 26
- 238000009826 distribution Methods 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 14
- 238000003745 diagnosis Methods 0.000 claims description 11
- 238000012353 t test Methods 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 3
- 230000007812 deficiency Effects 0.000 abstract 1
- 238000002360 preparation method Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 11
- 230000008439 repair process Effects 0.000 description 7
- 238000005259 measurement Methods 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 5
- 238000011084 recovery Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 238000003672 processing method Methods 0.000 description 3
- 238000000692 Student's t-test Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000001934 delay Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012886 linear function Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000002679 ablation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 210000004205 output neuron Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
Landscapes
- Testing Of Engines (AREA)
Abstract
The present invention proposes abnormal data intelligent diagnosing method in a kind of solid propellant rocket, abnormal point value is checked using t first, secondly according to BP-ML algorithm, the preparation method of weight in traditional artificial neural network is improved, is changed to that each parameter value { x of engine will be inputtediIt is assumed that a Gaussian process, meets AR (P) model, acquires weighted value by Maximum Likelihood Estimation Method, verify anomaly parameter, and carry out parameter value reconstruct.The present invention solves the problems, such as method deficiency under the conditions of the prior art, reduces the error of artificial micro-judgment, establishes abnormal data method for diagnosing faults.
Description
Technical Field
The invention belongs to the technical field of test measurement and control of solid rocket engines, and mainly relates to intelligent automatic diagnosis of abnormal data in a solid rocket engine test, which reduces errors of artificial experience judgment and establishes an intelligent diagnosis method of the abnormal data.
Background
The ground test of the solid rocket engine is a test project with high risk, high investment, high energy consumption and non-reversibility. With the development of the test technology, the reliability of the test measurement and control system flow is continuously improved, and the main parameters can be completely obtained without errors. However, the working conditions are complex in the test process, and the test data are abnormal due to the fact that the test data are inevitably affected by factors such as engine design, tool and sensor damage, cable high-temperature ablation and measurement block falling. The abnormal situations are divided into abnormal point values and abnormal parameters. The method for judging the abnormal point values comprises a t test, a Lauda criterion, a Shouyiller criterion and the like, wherein the t test is a statistical classical judgment method and belongs to a method in hypothesis tests of normal population mean and variance.
In the test process, the solid rocket engine is easily influenced by factors such as impact, temperature, structural deformation and the like, and the parameter signals measured in the test have nonlinear and non-stable characteristics. At present, when abnormal parameters are judged in the field of domestic solid rocket engines, a traditional data processing method is still adopted, the method is simple, when test data are abnormal, the reason that the fault belongs to the fault of a measurement and control system or the fault of the engine can not be judged, experts are required to perform experience interpretation, and certain errors exist, so that the application of the current more advanced data mining method in the field has certain necessity.
The solid rocket engine in the test process is easily influenced by factors such as impact, temperature, structural deformation and the like, the parameter signals measured in the test have nonlinear and non-stable characteristics, for the engine in the shaping stage, the performance parameters are in a relatively stable state, and the test environment and the statistical characteristics do not change along with the time change. Therefore, in practical engineering application, the solid engine ground test process is regarded as a stable random process, all the acquired parameter signals conform to the characteristics of stable random signals, and various data processing methods can be applied to the signals, such as machine learning, artificial intelligence and statistical methods which are widely applied at present, such as an artificial neural network model, a Logistic regression model, a time sequence and the like.
The relation among various parameters of the solid rocket engine test is non-stable and non-linear, and is influenced by different field factors of the specific test, the obtained data are different, the specific function relation form among various parameters is difficult to obtain, the artificial neural network model has non-linear mapping and generalization capability, and a 3-layer BP neural network can realize approximation (according to Kolrnogorov theorem) on any non-linear function, as shown in figure 1. It can be used to represent the interrelationship between parameters in the experimental data analysis process. The artificial neural network is composed of an input layer, a hidden layer and an output layer. The learning process consists of two processes of signal forward propagation and error backward propagation, and the weight is adjusted for a plurality of times until the error output by the network is reduced to an acceptable degree or the learning times are set in advance. The invention designs a new weight estimation algorithm, and learns to obtain a nonlinear relation between a dependent variable and an independent variable. When certain abnormal parameters exist, the abnormal parameters can be detected according to the nonlinear relation. Meanwhile, the abnormal parameters can be reconstructed and recovered.
Disclosure of Invention
The method aims at solving the problems that the later analysis and processing process of the test data of the existing solid rocket engine does not relate to abnormal data intelligent diagnosis, the traditional data processing method is basically still adopted, the method is simple, when the test data is abnormal, the fault reason can not be judged to belong to the measurement and control system fault or the engine fault, and the expert needs to perform experience interpretation, so that certain errors exist. The invention provides an intelligent diagnosis method for abnormal data in a solid rocket engine, which judges an abnormal point value in an engine test by a classical t test method; a new method for detecting abnormal parameters of the engine based on the artificial neural network model is designed, a weight estimation algorithm is improved, the method is used for finding a hidden fault rule from engine test historical data, the abnormal parameters can be effectively detected, artificial experience judgment errors are reduced, and intelligent diagnosis of the abnormal data is realized.
The technical scheme of the invention is as follows:
the intelligent diagnosis method for abnormal data in the solid rocket engine is characterized by comprising the following steps: the method comprises the following steps:
step 1: collecting test data of the solid rocket engine, checking whether the selected test data meet normal distribution, if so, performing the step 2, otherwise, selecting other test data to perform the check again;
step 2: judging abnormal points in the selected test data by adopting a t test method and removing the abnormal points;
and step 3: and (3) carrying out abnormal parameter detection on the test data subjected to the t detection processing in the step (2) by using a BP-ML algorithm and a neural network, and if the abnormal parameters exist, carrying out parameter value reconstruction on the abnormal parameters.
Further preferably, the method for intelligently diagnosing the abnormal data in the solid rocket engine is characterized in that: in the step 1, whether the test data meet normal distribution or not is judged by calculating the skewness and kurtosis of the test data.
Further preferably, the method for intelligently diagnosing the abnormal data in the solid rocket engine is characterized in that: the BP-ML algorithm in the step 3 is obtained by adopting a maximum likelihood estimation method to realize a weight vector estimation method in the artificial neural network BP algorithm.
Further preferably, the method for intelligently diagnosing the abnormal data in the solid rocket engine is characterized in that: the specific process of carrying out abnormal parameter detection in the step 3 is as follows:
step 3.1: training a neural network by using data without abnormal parameters as sample data to obtain a trained neural network;
step 3.2: inputting actual test data into a neural network, repairing the neural network, and if the correlation between the repaired output value and a target value is poor, inferring that abnormal parameters exist in the input value of the neural network; if the neural network input values have abnormal parameters, judging the parameter values in the neural network input values respectively to obtain specific abnormal parameters;
step 3.3: and 3, taking the abnormal parameters obtained in the step 3.2 as output parameters of the neural network, removing the abnormal parameters from the input parameters of the neural network, and reconstructing the neural network to obtain the repaired abnormal parameters.
Advantageous effects
Compared with the prior art, the invention has the characteristics that: (1) the test data characteristics of the solid rocket engine are met; (2) the application is wide; (3) constructing a complete abnormal data inspection method; (4) and completing the abnormal data reconstruction process.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1: a three-layer neural network schematic diagram;
FIG. 2: target value and repair value schematic;
FIG. 3: the target value and the repair value are related to each other;
FIG. 4: a schematic diagram of CCDD correlation coefficients;
FIG. 5: the target value and the repair value are related to each other;
FIG. 6: target value and repair value schematic;
FIG. 7: the target value and the repair value are related to each other;
FIG. 8: a schematic diagram of CCDD correlation coefficients;
FIG. 9: the weight w of the neural network;
FIG. 10: the value of the abnormal parameter wP9 after being repaired is compared with the target value;
FIG. 11: exception parameter wP 9.
Detailed Description
The abnormal data inspection method according to the present invention uses t-test on the premise that the measurement values follow the theoretical basis of normal distribution. Although it is understood theoretically that the measured performance parameter characteristics do not change along with the time change in the steady-state stage of the ground test of the solid rocket engine and conform to normal distribution, the measured performance parameter characteristics are interfered by different conditions in the test process, and the distribution of the measured performance parameter characteristics has different degrees of deviation, so that the test data needs to be subjected to normal distribution test before the t test is used.
According to the skewness and kurtosis formula, when the random variable { xnWhen the normal distribution is obeyed, its skewness k10, kurtosis k2＝3，
The calculation formula is as follows:
in the t test, when the variance is unknown, the test method of the single normal overall mean is as follows:
assume that the population obeys a normal distribution N (μ, σ)2),μ,σ2Are all unknown parameters, (X)1,...,XN) Is a sample of total capacity n, to be examined for hypothesis:statistics are typically constructed with sample-corrected variances instead of the overall variances.
Wherein,when it is assumed that H0When the T distribution is satisfied, T follows the T distribution with the degree of freedom n-1, and when the value of | T | is large, H should be negated on the assumption that it is unlikely to be satisfied0Therefore, for a given 0<α<1, obtaining the critical value t of the test from the t distribution tableα/2(n-1) reacting
p{|T|≥tα/2(n-1)}＝α (4)
Therefore it is examined
If | T | ≧ Tα/2(n-1), rejecting hypothesis H0I.e. as the global mean and mu0The difference is obvious; if T < Tα/2(n-1), then H is accepted0I.e. as the global mean and mu0There was no significant difference. This utilization of statistics obeying t-distributions as test statisticsThe method of measuring the amount is called t-test.
Selecting steady state data of a pressure value p which is one of the most main parameters measured in the solid rocket engine test as a test object. Selecting data point values in a certain hair styling test 10.5 s-17.5 s, wherein the sampling rate is 5000Sa/s, the time interval is 0.0002s, and calculating the skewness k of the measured data according to the formulas (1) and (2)10.2279, kurtosis k2When 2.7308 is satisfied, the normal distribution is good and can be tested by t test.
The neural network model has strong generalization capability and can be used for nonlinearly mapping the relationship among various parameters, so that thrust data which has large influence on engine design is selected as output neuron nodes, and the input layer neuron nodes select the parameters of pressure, displacement, strain and temperature related to an output layer. Let the system equation have an m-dimensional input vector and an n-dimensional output vector, { x }iIs the value of each input parameter point, { fiIs the output parameter point value, x ═ x1,x2,…,xm)T,f＝(f1,f2,…,fn)TThe network corresponds to m input nodes, n output nodes, ziFor the system identification function, also called activation function, a linear function, a ramp function, an s-shaped function, etc. can be chosen. The system equation is as follows:
let the neural network have M layers, the number of nodes of the l layer be nl,fi lIs the output of the l-th layer node i, then
Wherein,is the state of the l-th level node i,coefficient row vector, flLayer l-1 outputs the column vector.
At this time, according to the existing BP algorithm of the artificial neural network, the weight coefficient in the system equationUsing the instructor learning method, giving the objective function g (w)min：
I.e. after given input parameters X, F, the objective function is minimized
In order to output the value of the output,the weight vector can be obtained by the method in optimization, namely steepest gradient descent method
In order to improve the accuracy and efficiency of the output value of the system equation, the invention improves the BP algorithm of the artificial neural network, changes the estimation method of the weight vector into maximum likelihood estimation (ML), and the new algorithm is named as BP-ML algorithm,
according to the formula (1) and (2) in the text, the parameter values { x) of the engine are inputiThe normal distributions are met, and independent equal distributions are satisfied, because each parameter value { x }iAre process quantities that all vary with time, assuming { x }iIt is a Gaussian process, satisfying the AR (P) model, that is, formula (11) at this time replaces formula (9) in the BP algorithm,
Xk+p+1＝a1Xk+1+a2Xk+2+...+apXk+p+εk+p+1. (11)
in which white noise epsilon is assumedm～i.i.d.N(0,σ2)，(X1,X2,...,XM) A normal distribution with mean μ and covariance Σ is satisfied. Let (x)p+1,xp+2,…xp+m)TIs one input sample of the ar (p) sequence. Then:
(xp+1,xp+2,…xp+m)～N(μ,Σ)
where Σ is the covariance matrix:
our aim is based on the observed value xp+1,xp+2,…xp+mEstimating the unknown parameter σ2,a1,a2,...,apThe value of (a) is2,a1,a2,...,ap) Let x1,x2,...,xpAre unknown parameters. Equation (11) may be varied as:
so based on the observed value xp+1,xp+2,…xp+mWe can get
Then (epsilon)p+1,εp+2,…,εp+m) Has a joint probability density of
The likelihood function is
And the likelihood equation is
Wherein i is more than or equal to 1 and less than or equal to p, and l is more than or equal to 1 and less than or equal to p.
First, since we assume the model to be a p-order autoregressive model, a1,a2,...,apNot equal to 0, according toI is more than or equal to 1 and less than or equal to p can obtain epsilonp+1＝εp+2＝ε2p0. And then can solve sigma2The maximum likelihood estimate of (a) is:
according to the followingL is more than or equal to 1 and less than or equal to p
When l is 1,2,.. times.p
Wherein epsilonj＝xj-a1xj-p-a2xj-p+1-...-apxj-1Substituting the formula into (19) to obtain
From (20), the parameter a can be solved1,a2,...,apThe value of parameter μ, Σ can be obtained.
I.e. the objective function g (w)minIn the formula (9), the equation has been modeled by establishing an AR (P), and the unknown parameters are solved by a maximum likelihood estimation method, wherein aiThe value is wiSubstituting the value into equation (10) to obtain g (w)minAt the minimum, the sample value is outputtedThe purpose of signal reconstruction is achieved. The BP-ML method has more estimated parameters and is more complex, but the precision and the error of the BP algorithm are improved.
The following detailed description of embodiments of the invention is intended to be illustrative, and not to be construed as limiting the invention.
When the neural network is trained, 12 data in a certain batch of extraction tests are selected as training samples, and the training samples can typically represent the test conditions of the solid rocket engine. In specific application, the process of estimating the weight by the BP-ML algorithm is more complex and is suitable for the condition of higher precision requirement, so the BP method is also applied in the following examples.
Criteria for determination
The error determination of the recovery value is based on the root mean square of the signal as a criterion,
the root mean square (SMSE) equation is as follows:
the effect of obtaining the recovery value according to the neural network is evaluated by adopting correlation Coefficients (DDCC) of different time delays, the correlation degree of the recovery value and the target value can be judged, the DDCC value is closer to 1, the higher the correlation degree is, the higher the accuracy of obtaining the recovery value is, and the more accurate the established neural network is. The correlation coefficient threshold is given as 5%, i.e. |1-C | ≦ 0.05, then the correlation is considered high.
Wherein: sTIn order to be the target output,for recovery values, τ is the delay value.
The neural network anomaly parameter test is described in detail below:
the method comprises the following steps: establishing a neural network, and training to obtain a network weight value; selecting twelve main parameters of thrust F and pressure P of a certain batch of model test data as research objects by using a three-layer BP artificial neural network, wherein the working time is 0-18 s, and the time interval is 0.2s, namelyThe number of each path point value is 91, the neuron nodes of the input layer select relevant parameters influencing the nodes of the output layer, the thrust F values and the pressure P values of 8-sent data are respectively selected, and the pressure P values of key parameters reflecting the working conditions of the engine are selected by the output layer1As target output, P1Pressure value for first issue data, i.e. input parameter P1，P2，P4，P5，P6，P7，P9，P10，F1，F2，F4，F5，F6，F7，F9，F108-path pressure, 8-path thrust and 16-path parameters in total, and the output target value is P1The hidden layer neuron node number selection needs to comprehensively consider training efficiency and training effect, the larger the node number is, the better the node number is, if the node number is too large, the operation amount is increased, the training is too slow, even the overtraining is caused, and the output value is diverged; if the number of the nodes is too small, the training result is poor, and the output is not ideal, so that after data analysis, the number of the nodes is selected to be 50, the training effect is optimal, and the error value is minimum; the activation function selects a bipolar sigmoid function because its range is [ -1,1]And the maximum pressure value reaches 105kPa, thus reducing all input and target values by 10-5And (4) finishing.
And assigning random numbers to the initial weight matrix, starting training the network, storing the weight at the moment after the error reaches a preset requirement, inputting actual engine test data as a test sample, calculating the weight instead of a human weight, and detecting parameters.
FIG. 3 shows the correlation between the target value and the repair value, and it can be seen that the correlation between the source signal and the separated signal is high, and the correlation is consistent regardless of trend or value range; and under different time delays, the correlation coefficient is improved to different degrees, and when the tau is 7, the value of CCDD is maximum, which shows that the correlation degree is highest. SMSE 2.8687. The C is 1.0008, i.e. the error accuracy between the target value and the repair copy is 0.08%.
To verify the reliability of the recovered value, the pressure P of another data is selected2As an output, when τ is 21, the correlation coefficient C is 1.0005, that is, the error accuracy between the target value and the copy correction is 0.05%, and the degree of correlation is the highest.
Step two: judging abnormal parameters; in order to avoid misjudging the normal data as abnormal data, the error threshold value should be larger than the training error threshold value during detection. And storing the weight value at the moment, namely inputting actual engine test data as a test sample, calculating the weight value instead of a person, and detecting the parameters. Because the pressure and thrust data of a certain batch of collected model extraction test data are taken as main parameters, the average acquisition rate is 100%, and no abnormal condition occurs, P is used9The 40 point values in the process are artificially enlarged by ten times to construct an abnormal parameter wP9Inputting a value P1，P2，P4，P5，P6，P7，wP9，P10，F1，F2，F4，F5，F6，F7，F9，F108-path pressure, 8-path thrust and 16-path parameters in total, and the output target value is P1Performing neural network restoration to obtain restored P1When the delay value tau is any value, |1-C | ≧ 0.05, that is, the correlation between the restored P1 and the target value P1 is poor, so that abnormal parameters in the input value can be reversely deduced. Similarly, on the premise of a neural network model constructed by the user, if the thrust F and the pressure P of a certain shot are judged to be abnormal, the thrust pressure value in the input value is replaced by a parameter needing to be judged, the output target value is unchanged, and the output target value is arbitrarily P1And obtaining the repaired values, comparing the repaired values, observing whether the correlation coefficient is within a specified threshold value, namely |1-C | < 0.05, if so, having no abnormal parameter, if not, having an abnormal parameter, further sequentially replacing the detection values, and judging the specific abnormal parameter.
Step three: repairing abnormal parameters; according to the neural network model trained by us, determining wP9After the abnormal parameter is set, if abnormal data restoration is required, the output layer neuron node in fig. 1 is set as the target parameter wP to be restored9Removing mesh from input layer neuron nodesMarking parameters, respectively establishing data of each key parameter, reconstructing a neural network, and training in the same way as before.
The invention comprises the following application steps:
the first step is as follows: judging an abnormal point value; selecting steady state data of a pressure value P which is the most main parameter in a certain solid rocket engine test. Firstly, checking the normality of data, and respectively calculating the skewness and the kurtosis of the measured data according to the data (1) and the data (2), wherein the closer the skewness is to 0, the closer the kurtosis is to 3, the better the normal distribution characteristic is shown, and the t test can be applied; and secondly, according to the formulas (3), (4) and (5), completing the t inspection process, judging abnormal points and removing the abnormal points.
The second step is that: and training a neural network, checking out abnormal parameters, and reconstructing parameter values.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made in the above embodiments by those of ordinary skill in the art without departing from the principle and spirit of the present invention.
Claims (4)
1. An intelligent diagnosis method for abnormal data in a solid rocket engine is characterized by comprising the following steps: the method comprises the following steps:
step 1: collecting test data of the solid rocket engine, checking whether the selected test data meet normal distribution, if so, performing the step 2, otherwise, selecting other test data to perform the check again;
step 2: judging abnormal points in the selected test data by adopting a t test method and removing the abnormal points;
and step 3: and (3) carrying out abnormal parameter detection on the test data subjected to the t detection processing in the step (2) by using a BP-ML algorithm and a neural network, and if the abnormal parameters exist, carrying out parameter value reconstruction on the abnormal parameters.
2. The intelligent diagnosis method for abnormal data in a solid rocket engine according to claim 1, wherein: in the step 1, whether the test data meet normal distribution or not is judged by calculating the skewness and kurtosis of the test data.
3. The intelligent diagnosis method for abnormal data in a solid rocket engine according to claim 1, wherein: the BP-ML algorithm in the step 3 is obtained by adopting a maximum likelihood estimation method to realize a weight vector estimation method in the artificial neural network BP algorithm.
4. The intelligent diagnosis method for abnormal data in a solid rocket engine according to claim 3, characterized in that: the specific process of carrying out abnormal parameter detection in the step 3 is as follows:
step 3.1: training a neural network by using data without abnormal parameters as sample data to obtain a trained neural network;
step 3.2: inputting actual test data into a neural network, repairing the neural network, and if the correlation between the repaired output value and a target value is poor, inferring that abnormal parameters exist in the input value of the neural network; if the neural network input values have abnormal parameters, judging the parameter values in the neural network input values respectively to obtain specific abnormal parameters;
step 3.3: and 3, taking the abnormal parameters obtained in the step 3.2 as output parameters of the neural network, removing the abnormal parameters from the input parameters of the neural network, and reconstructing the neural network to obtain the repaired abnormal parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910224959.6A CN109779791B (en) | 2019-03-24 | 2019-03-24 | Intelligent diagnosis method for abnormal data in solid rocket engine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910224959.6A CN109779791B (en) | 2019-03-24 | 2019-03-24 | Intelligent diagnosis method for abnormal data in solid rocket engine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109779791A true CN109779791A (en) | 2019-05-21 |
| CN109779791B CN109779791B (en) | 2021-01-01 |
Family
ID=66490920
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910224959.6A Active CN109779791B (en) | 2019-03-24 | 2019-03-24 | Intelligent diagnosis method for abnormal data in solid rocket engine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109779791B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110239745A (en) * | 2019-06-13 | 2019-09-17 | 北京深蓝航天科技有限公司 | The multiple-motor parallel connection rocket control device and control method for having power redundant ability |
| CN110414063A (en) * | 2019-06-29 | 2019-11-05 | 万翼科技有限公司 | Model restorative procedure and Related product |
| CN110531622A (en) * | 2019-09-05 | 2019-12-03 | 沈阳航空航天大学 | A kind of Thrust of Solid Rocket Motor control method based on radial base neural net |
| CN111259927A (en) * | 2020-01-08 | 2020-06-09 | 西北工业大学 | Rocket engine fault diagnosis method based on neural network and evidence theory |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101907088A (en) * | 2010-05-27 | 2010-12-08 | 中国人民解放军国防科学技术大学 | Fault diagnosis method based on one-class support vector machines |
| CN104239698A (en) * | 2014-09-01 | 2014-12-24 | 中国航天科技集团公司第四研究院四〇一所 | Time sequence correcting method of solid rocket engine vibration distortion signal |
| CN106460727A (en) * | 2014-04-03 | 2017-02-22 | 赛峰飞机发动机公司 | Method and device for monitoring a parameter of a rocket engine |
| CN108520301A (en) * | 2018-04-17 | 2018-09-11 | 重庆大学 | A kind of circuit intermittent fault diagnostic method based on depth confidence network |
| CN108803554A (en) * | 2017-05-04 | 2018-11-13 | 上海电气集团股份有限公司 | A kind of intelligent diagnosing method of generator signal |
-
2019
- 2019-03-24 CN CN201910224959.6A patent/CN109779791B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101907088A (en) * | 2010-05-27 | 2010-12-08 | 中国人民解放军国防科学技术大学 | Fault diagnosis method based on one-class support vector machines |
| CN106460727A (en) * | 2014-04-03 | 2017-02-22 | 赛峰飞机发动机公司 | Method and device for monitoring a parameter of a rocket engine |
| CN104239698A (en) * | 2014-09-01 | 2014-12-24 | 中国航天科技集团公司第四研究院四〇一所 | Time sequence correcting method of solid rocket engine vibration distortion signal |
| CN108803554A (en) * | 2017-05-04 | 2018-11-13 | 上海电气集团股份有限公司 | A kind of intelligent diagnosing method of generator signal |
| CN108520301A (en) * | 2018-04-17 | 2018-09-11 | 重庆大学 | A kind of circuit intermittent fault diagnostic method based on depth confidence network |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110239745A (en) * | 2019-06-13 | 2019-09-17 | 北京深蓝航天科技有限公司 | The multiple-motor parallel connection rocket control device and control method for having power redundant ability |
| CN110414063A (en) * | 2019-06-29 | 2019-11-05 | 万翼科技有限公司 | Model restorative procedure and Related product |
| CN110531622A (en) * | 2019-09-05 | 2019-12-03 | 沈阳航空航天大学 | A kind of Thrust of Solid Rocket Motor control method based on radial base neural net |
| CN110531622B (en) * | 2019-09-05 | 2022-04-05 | 沈阳航空航天大学 | Thrust control method of solid rocket engine based on radial basis function neural network |
| CN111259927A (en) * | 2020-01-08 | 2020-06-09 | 西北工业大学 | Rocket engine fault diagnosis method based on neural network and evidence theory |
| CN111259927B (en) * | 2020-01-08 | 2022-08-05 | 西北工业大学 | Rocket engine fault diagnosis method based on neural network and evidence theory |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109779791B (en) | 2021-01-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109779791B (en) | Intelligent diagnosis method for abnormal data in solid rocket engine | |
| CN108038300B (en) | Optical fiber state evaluation method based on improved membership function combined with neural network | |
| CN108897286B (en) | Fault detection method based on distributed nonlinear dynamic relation model | |
| CN108256173B (en) | Gas circuit fault diagnosis method and system for dynamic process of aircraft engine | |
| CN107704962B (en) | Steam flow interval prediction method based on incomplete training data set | |
| EP1443375A1 (en) | Fault diagnosis | |
| CN109325293B (en) | Multi-fault diagnosis method based on correlation model | |
| Riad et al. | Evaluation of neural networks in the subject of prognostics as compared to linear regression model | |
| CN115081484B (en) | Aeroengine sensor fault diagnosis method based on CRJ-OSELM algorithm | |
| CN110222844A (en) | A kind of compressor performance prediction technique based on artificial neural network | |
| CN114266278A (en) | Dual-attention-network-based method for predicting residual service life of equipment | |
| CN115618732A (en) | Nuclear reactor digital twin key parameter autonomous optimization data inversion method | |
| CN113283004A (en) | Aero-engine degradation state fault diagnosis method based on transfer learning | |
| Cao et al. | Bearing fault diagnosis method based on GMM and coupled hidden Markov model | |
| CN113988210A (en) | Method and device for restoring distorted data of structure monitoring sensor network and storage medium | |
| CN114166509A (en) | Motor bearing fault prediction method | |
| CN117521512A (en) | Bearing residual service life prediction method based on multi-scale Bayesian convolution transducer model | |
| CN114861349A (en) | Rolling bearing RUL prediction method based on model migration and wiener process | |
| CN116992295A (en) | Reconstruction method and device for machine pump equipment monitoring missing data for machine learning | |
| CN112232570A (en) | Forward active total electric quantity prediction method and device and readable storage medium | |
| CN111863153A (en) | Method for predicting total amount of suspended solids in wastewater based on data mining | |
| CN117013533A (en) | Power distribution network state estimation method, system, storage medium and equipment | |
| CN116579233A (en) | Method for predicting residual life of mechanical equipment | |
| CN109657285A (en) | The detection method of turbine rotor transient stress | |
| CN114048837A (en) | Deep neural network model reinforcement method based on distributed brain-like map |
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 |