CN115841149A - Confidence estimation-based satellite fault prediction model training and fault prediction method - Google Patents
Confidence estimation-based satellite fault prediction model training and fault prediction method Download PDFInfo
- Publication number
- CN115841149A CN115841149A CN202211240253.7A CN202211240253A CN115841149A CN 115841149 A CN115841149 A CN 115841149A CN 202211240253 A CN202211240253 A CN 202211240253A CN 115841149 A CN115841149 A CN 115841149A
- Authority
- CN
- China
- Prior art keywords
- satellite
- data
- fault prediction
- confidence
- unknown
- 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
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000012549 training Methods 0.000 title claims abstract description 44
- 239000013598 vector Substances 0.000 claims abstract description 80
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method for training a satellite fault prediction model and predicting a fault based on confidence estimation, which belongs to the field of satellite fault prediction and comprises the following steps: acquiring original satellite telemetry data and performing data preprocessing; inputting original satellite telemetry data into satellite fault prediction obtained by training of an unknown satellite fault prediction model training method based on confidence estimation, and calculating output characteristic vectors and deviation vectors; calculating a harmonic mean of each dimension of the deviation vector to obtain a confidence degree estimation value of the input data; and judging whether the confidence coefficient estimation value is greater than a confidence coefficient threshold value, if so, judging the satellite state to be an unknown fault, and if not, judging the satellite state to be no fault or a known fault. The method and the device can predict the known faults and unknown faults of the satellite platform and the load simultaneously according to the satellite telemetering data, reduce the influence of data noise on the satellite fault prediction model, and improve the reliability of the fault prediction model.
Description
Technical Field
The invention relates to a satellite fault prediction method based on confidence estimation in the field of satellite fault prediction, in particular to a prediction condition of unknown faults of a satellite.
Background
With the rapid development of satellite technology, the functions and structures of satellite loads and platforms become more complex, equipment becomes more precise, telemetering information becomes more comprehensive, higher requirements on the safety and reliability of the satellite loads and the platforms are provided, and some traditional fault prediction methods cannot meet the increasing satellite guarantee requirements.
Currently, in the field of satellite fault prediction, the existing satellite fault prediction methods are all used for predicting known faults. The satellite fault prediction method based on the expert system is limited by the coverage range of a knowledge base and cannot accurately predict unknown faults, and the satellite fault prediction method based on data driving is limited by the fact that model training data cannot give credible prediction to the unknown faults. The prediction output of the existing method is mostly point estimation, although the point estimation has strong discrimination capability in a boundary, the model always gives output in a value range for any data input, and when the satellite has unknown fault, the existing method cannot inform the model of the determination of the input data or not and cannot predict the unknown fault.
Disclosure of Invention
In order to solve the problems, the invention provides a satellite fault prediction model training and fault prediction method based on confidence degree estimation, and in the high-sensitivity application field of satellite fault prediction, the invention improves the reliability of the satellite fault prediction method and predicts unknown satellite faults.
The invention adopts the following technical scheme:
an unknown satellite fault prediction model training method based on confidence estimation comprises the following steps:
step S1: acquiring original satellite telemetry data and initializing confidence coefficient neural network parameters;
step S2: preprocessing original satellite telemetering data to obtain noiseless data, and adding Gaussian noise and Laplace noise to obtain noisy data;
and step S3: respectively inputting noisy data and non-noisy data into a confidence neural network to obtain corresponding characteristic vectors and deviation vectors;
and step S4: calculating a harmonic mean of each dimensionality of a deviation vector corresponding to the noisy data to obtain a confidence coefficient estimation S of the noisy data;
step S5: calculating a first loss function of the training process according to the feature vectors corresponding to the noisy data;
step S6: calculating a second loss function in the training process according to the confidence coefficient estimation value and the characteristic vector and the deviation vector corresponding to the noisy data and the non-noisy data;
step S7: summing the first loss function and the second loss function to obtain a total loss function in the model training process, and optimizing a confidence coefficient neural network parameter through a back propagation algorithm;
step S8: and (5) repeating the step (S2) to the step (S7) until the total loss function is converged to obtain the trained unknown satellite fault prediction model.
Wherein, the step S3 specifically comprises:
inputting noisy data and non-noisy data into a confidence degree neural network, and outputting a characteristic vector of the noisy data by the confidence degree neural network
And the deviation vector pick>
Obtaining feature vectors for noiseless data
And the deviation vector pick>
Wherein->
Is the feature vector->
Is greater than or equal to the ith dimension of (b)>
Is the deviation vector pick>
Of the ith dimension, f i Is the i-th dimension, σ, of the feature vector f i Is the ith dimension of the deviant vector σ, i =1, \ 8230and D, D is the length of the feature vector and deviant vector.
Wherein the first loss function L in step S5 1 Comprises the following steps:
in the formula,
is a function of Softmax loss>
As a function of the central loss, α 1 And alpha 2 Are all non-negative multipliers;
softmax loss function
Comprises the following steps:
center loss function
Comprises the following steps:
in the formula,
for a corresponding class center in a satellite telemetry data sample batch, the class center is associated with a feature vector->
Is the same, 2' is the two-norm of the matrix, and N is the number of sample batches.
Wherein the second loss function L in step S6 2 Comprises the following steps:
L 2 ＝S*L 1 +βL LMS
in the formula, L 1 Is said first loss function, L LMS β is a non-negative multiplier for the MLS distance loss function;
the MLS distance loss function is:
an unknown satellite fault prediction method based on confidence estimation comprises the following steps:
step S1: acquiring satellite telemetry data to be subjected to fault prediction, wherein the satellite telemetry data comprises information acquisition frames of a satellite platform and a load;
step S2: carrying out data preprocessing on the acquired satellite telemetering data to obtain input data;
and step S3: carrying out forward propagation on input data through an unknown satellite fault prediction model obtained by training an unknown satellite fault prediction model training method based on confidence estimation, wherein the unknown satellite fault prediction model carries out the following steps on the input data:
step S31: carrying out forward propagation on input data through an unknown satellite fault prediction model to obtain a feature vector
And the deviation vector pick>
Wherein f is i Is the ith dimension, σ, of the feature vector f i Is the ith dimension of the deviant vector σ, i =1, \ 8230, D, D is the length of the eigenvector and deviant vector;
step S32: calculating harmonic mean of deviation vector sigma per dimension to obtain confidence degree estimation value
And step S4: judging whether the confidence coefficient estimated value S of the input satellite telemetering data is larger than a confidence coefficient threshold value Th s (ii) a If the confidence coefficient estimated value S is smaller than the confidence coefficient threshold Th s If the current state of the satellite is unknown, judging that an unknown fault exists, and not executing the step S5; if the confidence coefficient estimated value S is greater than or equal to the confidence coefficient threshold Th s If yes, continuing to execute the step S5;
step S5: and obtaining classification probability vectors of all classes by using the feature vectors output by the classification branches through a Softmax function, taking the position of the maximum value in the classification probability vectors as a classification result, and judging the satellite state to be the class without fault or with known fault according to the classification result.
Compared with the prior art, the invention has the following beneficial effects:
the unknown satellite fault prediction model training method based on confidence coefficient estimation and the unknown satellite fault prediction method based on confidence coefficient estimation are respectively used for a training process and a testing process of a satellite fault prediction model. The satellite fault prediction model has more accurate fault prediction capability, meanwhile, the influence of data noise on the satellite fault prediction model is reduced, and the reliability and robustness of the fault prediction model are improved. In addition, whether the characteristic distribution of the input sample deviates or not is judged by setting a threshold value to judge the size of the confidence degree estimation value, so that the satellite platform and the load known fault can be predicted, and the unknown fault can be predicted. Furthermore, the model training collapse phenomenon is avoided by the optimization of the first and second loss functions.
Drawings
FIG. 1 is a schematic flow chart diagram illustrating a method for confidence estimation based unknown satellite fault prediction model training in one embodiment of the present invention;
FIG. 2 is a schematic block diagram of an unknown satellite fault prediction model training method based on confidence estimation according to the present invention;
fig. 3 is a schematic flow chart of an unknown satellite fault prediction method based on confidence level estimation according to an embodiment of the present invention.
Detailed Description
The technical solution of the present invention will be described in detail with reference to the accompanying drawings and preferred embodiments.
In one embodiment, as shown in fig. 1 and fig. 2, the present invention provides a method for training an unknown satellite fault prediction model based on confidence level estimation, which includes the following steps 1 to 8. The specific training process of the unknown satellite fault prediction model training method based on confidence estimation is as follows:
step S1: acquiring original satellite telemetry data and initializing confidence coefficient neural network parameters;
and acquiring original satellite telemetry data, carrying out one-hot coding, and carrying out confidence coefficient neural network model parameter initialization.
Step S2: preprocessing the original satellite telemetry data to obtain noiseless data, and adding Gaussian noise and Laplace noise to obtain noisy data;
all the original satellite telemetry data is subjected to data preprocessing operation of normalization and principal component analysis, and meanwhile, gaussian noise and Laplace noise are added to training data.
And step S3: inputting the noisy data and the non-noisy data into a confidence neural network to obtain corresponding characteristic vectors and deviation vectors;
the method comprises the following specific steps: inputting noisy data and non-noisy data into a confidence degree neural network, and outputting a characteristic vector of the noisy data by the confidence degree neural network
And the deviation vector pick>
Wherein +>
Is a feature vector
Is greater than or equal to the ith dimension of (b)>
Is the deviation vector pick>
I =1, \ 8230, D, D is the length of the eigenvectors and the deviation vectors; obtaining a feature vector for the noiseless data>
And the deviation vector pick>
Wherein f is i Is the ith dimension, σ, of the feature vector f i Is the ith of the deviation vector σDimension, i =1, \ 8230, D, D is the length of the eigenvectors and the disparity vectors.
And step S4: calculating the harmonic mean of each dimension of the deviation vector corresponding to the noisy data to obtain the data confidence estimation of the added noise;
computing noisy data deviation vectors
And the average number of chords in each dimension, a confidence estimate is obtained>
Step S5: calculating a first loss function of the training process according to the noisy data feature vector;
the first loss function is:
wherein,
is a function of Softmax loss>
As a function of the central loss, α 1 And alpha 2 Are all non-negative multipliers. The Softmax loss function @>
Comprises the following steps:
the center loss function
Comprises the following steps:
wherein,
for a corresponding class center in a satellite telemetry data sample batch, the class center is associated with a feature vector->
Is the same, 2' is the two-norm of the matrix, N is the number of sample batches
Step S6: calculating a second loss function in the training process according to the confidence coefficient estimation value and the characteristic vector and the deviation vector corresponding to the noisy data and the non-noisy data;
the second loss function L 2 Comprises the following steps:
L 2 ＝S*L 1 +βL LMS
wherein L is 1 Is said first loss function, L LMS β is a non-negative multiplier for the MLS distance loss function. The MLS distance loss function L LMS Comprises the following steps:
step S7: summing the first loss function and the second loss function to obtain a total loss function in the model training process, and optimizing model parameters through a back propagation algorithm;
calculating a total loss function L of the model training process and optimizing model parameters through a Back Propagation (BP) algorithm, wherein the total loss function L is calculated by the following formula:
L＝L 1 +L 2
model parameters are optimized by a back propagation algorithm.
Step S8: and repeating the step S2 to the step S7 until the total loss function is converged, obtaining model parameters after training, and further obtaining an unknown satellite fault prediction model after training.
The confidence estimation is used in the training process of the satellite fault prediction model, and the confidence estimation value can accurately reflect the cheap condition of data distribution of satellite telemetry, so that the trained satellite fault prediction model can predict the unknown fault condition of the satellite, and the reliability and robustness of the satellite fault prediction model are improved; meanwhile, the loss function is optimized according to the satellite fault prediction model, the training collapse phenomenon in the training process is avoided, and the accuracy of the known satellite fault prediction is improved.
In another embodiment, as shown in fig. 3, the present invention provides a method for predicting known satellite faults and unknown faults based on confidence level estimation, where the method uses an unknown satellite fault prediction model trained by the method for training an unknown satellite fault prediction model based on confidence level estimation described in the foregoing embodiment to predict known satellite faults and unknown faults, and specifically includes the following steps:
step S1: acquiring satellite telemetry data to be subjected to fault prediction, wherein the satellite telemetry data comprises information acquisition frames of a satellite platform and a load;
step S2: carrying out data preprocessing on the acquired satellite telemetry data to obtain input data;
and step S3: carrying out forward propagation on input data through an unknown satellite fault prediction model obtained by training an unknown satellite fault prediction model training method based on confidence coefficient estimation, wherein the unknown satellite fault prediction model carries out the following steps on the input data:
step S31: carrying out forward propagation on input data through an unknown satellite fault prediction model to obtain a feature vector
And the deviation vector pick>
Wherein f is i Is the i-th dimension, σ, of the feature vector f i Is the ith dimension of the deviant vector σ, i =1, \ 8230, D, D is the length of the eigenvector and deviant vector;
step S32: calculating harmonic mean of deviation vector sigma per dimension to obtain confidence degree estimation value
And step S4: judging whether the confidence coefficient estimated value S of the input satellite telemetering data is larger than a confidence coefficient threshold value Th s (ii) a If the confidence coefficient estimated value S is smaller than the confidence coefficient threshold Th s If the current state of the satellite is unknown, judging that an unknown fault exists, and not executing the step S5; if the confidence coefficient estimated value S is greater than or equal to the confidence coefficient threshold Th s If yes, continuing to execute the step S5;
step S5: and obtaining classification probability vectors of all classes by using the feature vectors output by the classification branches through a Softmax function, taking the position of the maximum value in the classification probability vectors as a classification result, and judging the satellite state to be the class without fault or with known fault according to the classification result.
The unknown satellite fault prediction method based on confidence estimation provided by the embodiment of the invention uses a satellite fault prediction model to perform satellite fault prediction, and the satellite fault prediction model is a model obtained by training with an unknown satellite fault prediction model training method based on confidence estimation. The satellite fault prediction model outputs a confidence coefficient estimation value to the input satellite telemetering data, and the confidence coefficient estimation value can accurately reflect the cheap condition of data distribution of satellite telemetering, so that the trained satellite fault prediction model can predict the unknown fault condition of the satellite, and the reliability and robustness of the satellite fault prediction model are improved;
the technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent should be subject to the appended claims.
Claims (5)
1. An unknown satellite fault prediction model training method based on confidence estimation is characterized by comprising the following steps:
step S1: acquiring original satellite telemetry data and initializing confidence coefficient neural network parameters;
step S2: preprocessing original satellite telemetering data to obtain noiseless data, and adding Gaussian noise and Laplace noise to obtain noisy data;
and step S3: respectively inputting noisy data and non-noisy data into a confidence neural network to obtain corresponding characteristic vectors and deviation vectors;
and step S4: calculating a harmonic mean of each dimensionality of a deviation vector corresponding to the noisy data to obtain a confidence coefficient estimation S of the noisy data;
step S5: calculating a first loss function of the training process according to the feature vector corresponding to the noisy data;
step S6: calculating a second loss function in the training process according to the confidence coefficient estimation value and the characteristic vector and the deviation vector corresponding to the noisy data and the non-noisy data;
step S7: summing the first loss function and the second loss function to obtain a total loss function in the model training process, and optimizing a confidence coefficient neural network parameter through a back propagation algorithm;
step S8: and (5) repeating the step (S2) to the step (S7) until the total loss function is converged to obtain the trained unknown satellite fault prediction model.
2. The unknown satellite fault prediction model training method based on confidence estimation according to claim 1, wherein the step S3 specifically comprises:
inputting noisy data and non-noisy data into a confidence degree neural network, and outputting a characteristic vector of the noisy data by the confidence degree neural network
Sum deviation vector
Obtaining feature vectors for noiseless data
Sum deviation vector
Wherein
Is a feature vector
The (c) th dimension of (a),
is a deviation vector
Of the ith dimension, f i Is the i-th dimension, σ, of the feature vector f i Is the ith dimension of the deviant vector σ, i =1, \ 8230and D, D is the length of the feature vector and deviant vector.
3. The method of claim 2, wherein the first loss function L in step S5 is a function of a first loss, L 1 Comprises the following steps:
in the formula,
in order to be a function of the Softmax loss,
as a function of the central loss, α 1 And alpha 2 Are all non-negative multipliers;
softmax loss function
Comprises the following steps:
center loss function
Comprises the following steps:
in the formula,
for corresponding class centers, class centers and feature vectors in satellite telemetry data sample batches
Is the same, 2' is the two-norm of the matrix, and N is the number of sample batches.
4. The method for training the unknown satellite fault prediction model based on the confidence coefficient estimation as claimed in claim 2, wherein the second loss function L in the step S6 2 Comprises the following steps:
L 2 ＝S*L 1 +βL LMS
in the formula, L 1 Is said first loss function, L LMS β is a non-negative multiplier for the MLS distance loss function;
the MLS distance loss function is:
5. an unknown satellite fault prediction method based on confidence estimation is characterized by comprising the following steps:
step S1: acquiring satellite telemetry data to be subjected to fault prediction, wherein the satellite telemetry data comprises information acquisition frames of a satellite platform and a load;
step S2: carrying out data preprocessing on the acquired satellite telemetering data to obtain input data;
and step S3: carrying out forward propagation on input data through an unknown satellite fault prediction model obtained by training an unknown satellite fault prediction model training method based on confidence estimation, wherein the unknown satellite fault prediction model carries out the following steps on the input data:
step S31: carrying out forward propagation on input data through an unknown satellite fault prediction model to obtain a feature vector
Sum deviation vector
Wherein f is i Is the i-th dimension, σ, of the feature vector f i Is the ith dimension of the offset vector σ, i =1, \8230, D is the length of the feature vector and the offset vector;
step S32: calculating harmonic mean of deviation vector sigma per dimension to obtain confidence degree estimation value
And step S4: judging whether the confidence coefficient estimated value S of the input satellite telemetering data is larger than a confidence coefficient threshold value Th s (ii) a If the confidence coefficient estimated value S is smaller than the confidence coefficient threshold Th s If the current state of the satellite is unknown, judging that an unknown fault exists, and not executing the step S5; if the confidence coefficient estimated value S is greater than or equal to the confidence coefficient threshold Th s If yes, continuing to execute the step S5;
step S5: and obtaining classification probability vectors of all classes by using the feature vectors output by the classification branches through a Softmax function, taking the position of the maximum value in the classification probability vectors as a classification result, and judging the satellite state to be the class without fault or with known fault according to the classification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211240253.7A CN115841149A (en) | 2022-10-11 | 2022-10-11 | Confidence estimation-based satellite fault prediction model training and fault prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211240253.7A CN115841149A (en) | 2022-10-11 | 2022-10-11 | Confidence estimation-based satellite fault prediction model training and fault prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115841149A true CN115841149A (en) | 2023-03-24 |
Family
ID=85575521
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211240253.7A Pending CN115841149A (en) | 2022-10-11 | 2022-10-11 | Confidence estimation-based satellite fault prediction model training and fault prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115841149A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116304884A (en) * | 2023-05-11 | 2023-06-23 | 西安衍舆航天科技有限公司 | Spacecraft telemetry data health prediction method, system, equipment and storage medium |
-
2022
- 2022-10-11 CN CN202211240253.7A patent/CN115841149A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116304884A (en) * | 2023-05-11 | 2023-06-23 | 西安衍舆航天科技有限公司 | Spacecraft telemetry data health prediction method, system, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106600059B (en) | Intelligent power grid short-term load prediction method based on improved RBF neural network | |
| CN108181591B (en) | Battery SOC value prediction method based on improved BP neural network | |
| CN114475350A (en) | Intelligent charging system and working method thereof | |
| CN113852432B (en) | Spectrum Prediction Sensing Method Based on RCS-GRU Model | |
| CN107944648A (en) | A kind of accurate Forecasting Methodology of large ship speed of a ship or plane rate of fuel consumption | |
| CN111967535A (en) | Fault diagnosis method and device for temperature sensor in grain storage management scene | |
| CN115841149A (en) | Confidence estimation-based satellite fault prediction model training and fault prediction method | |
| CN111355633A (en) | Mobile phone internet traffic prediction method in competition venue based on PSO-DELM algorithm | |
| CN115099296A (en) | Sea wave height prediction method based on deep learning algorithm | |
| CN113791351B (en) | Lithium battery life prediction method based on transfer learning and difference probability distribution | |
| CN115905855A (en) | Improved meta-learning algorithm MG-copy | |
| Sun et al. | Regression assisted matrix completion for reconstructing a propagation field with application to source localization | |
| CN114003900A (en) | Network intrusion detection method, device and system for secondary system of transformer substation | |
| CN117312835A (en) | Fuel cell performance prediction method based on mixed probability data driving model | |
| CN110288002B (en) | Image classification method based on sparse orthogonal neural network | |
| CN113837226B (en) | Aircraft telemetry parameter anomaly detection method based on uncertainty estimation | |
| CN113537821A (en) | Rapid assessment method and system for state of power system | |
| CN113408371A (en) | Early fault diagnosis method and device | |
| CN112968740A (en) | Satellite spectrum sensing method based on machine learning | |
| CN113033690A (en) | Carrier roller fault detection method for double-projection neighborhood embedding based on sound | |
| CN117148170B (en) | Battery energy storage system and energy storage test method thereof | |
| Qi et al. | Analysis and Prediction of Energy Consumption in Neural Networks Based on Machine Learning | |
| CN112231933B (en) | Feature selection method for radar electromagnetic interference effect analysis | |
| KR102679889B1 (en) | Method and Apparatus for Learning of Active learning algorithm with long-range observation | |
| CN116776926B (en) | Optimized deployment method, device, equipment and medium for dialogue model |
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 |