CN111860447A - Method for accurately identifying telemetry time sequence data mode of random phase shift satellite - Google Patents
Method for accurately identifying telemetry time sequence data mode of random phase shift satellite Download PDFInfo
- Publication number
- CN111860447A CN111860447A CN202010764023.5A CN202010764023A CN111860447A CN 111860447 A CN111860447 A CN 111860447A CN 202010764023 A CN202010764023 A CN 202010764023A CN 111860447 A CN111860447 A CN 111860447A
- Authority
- CN
- China
- Prior art keywords
- data
- satellite
- time series
- different
- channel
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
A method of accurately identifying random phase shifted satellite telemetry timing data patterns, comprising: preprocessing data; establishing a time sequence data pattern recognition model consisting of a plurality of layers of sensors and a plurality of multi-channel characteristic network layers, wherein the model input is respectively sent into a plurality of channels of the multi-channel characteristic network layers, each channel comprises a convolution kernel and a pooling structure, the extracted characteristics of the plurality of channels in one multi-channel characteristic network layer are integrated to be used as the input of the next multi-channel characteristic network layer, and the like; integrating and connecting the feature vectors extracted by the last multi-channel feature network layer into a one-dimensional vector as the input of a multi-layer sensor, wherein the multi-layer sensor is used as the last identification classification layer; the method can accurately identify the mode with random phase offset in the satellite telemetering time sequence data, improves the precision of judging the corresponding satellite parameters or the running states of the satellite components and the platform, and is more suitable for the application conditions of the actual satellite in real-time monitoring and real-time fault diagnosis.
Description
Technical Field
The invention belongs to the technical field of computers, and particularly relates to a method for accurately identifying a telemetry time sequence data mode of a random phase shift satellite.
Background
Satellite telemetry timing data is the fundamental basis for monitoring the operational state of a satellite. The pattern recognition of the satellite telemetry time sequence data is to automatically recognize and judge the satellite parameters or the running states of satellite components and platforms by a computer. In actual work, firstly, the standard mode of the satellite telemetering time sequence data and the label thereof need to be labeled manually. This work is very heavy. Moreover, there is a problem that a phase difference often exists between the time series data corresponding to the artificially labeled standard pattern and the randomly acquired time series data in the real environment. The random phase difference brings great errors to the identification result of a general machine learning model, and the identification accuracy of satellite telemetry time sequence data is influenced.
In a real environment, the time window for viewing the time series data may start from an arbitrary point in time. And the sequence samples in the windows are different due to different starting points of the observation windows. Due to the randomness of the start of the observation window, the sequence samples from the same pattern are shifted left and right, which is called the random phase shift of the time series data. For satellite telemetry time sequence data with random phase offset, the accuracy of satellite time sequence data pattern recognition can be reduced by using the existing recognition model, and an ideal recognition effect cannot be achieved. The mode of accurately identifying the telemetry time sequence data of the random phase offset satellite is very important for accurately judging satellite parameters or the running states of satellite components and platforms, and is a necessary condition for ensuring the satellite fault diagnosis and the abnormality detection accuracy.
The existing time series data pattern recognition method generally extracts features of data to obtain feature vectors, for example, feature vectors are obtained by using statistical features, spectral features, wavelet features or principal component analysis methods. And then classifying and identifying the time sequence data patterns by utilizing various classification models. For example, classifiers such as k-nearest neighbors, random forests, AdaBoost, support vector machines, BP networks (i.e., multilayer perceptrons) and the like are used for classifying and identifying the time-series feature vectors, and determining which mode the input vectors belong to. Or calculating the similarity of two Time series modes by using Pearson Correlation Coefficient (PCC), Dynamic Time Warping (DTW) and other methods, and then judging whether the two Time series modes belong to a certain mode according to the similarity. In recent years, a large number of Recurrent Neural Networks (RNNs) such as Long Short-Term Memory networks (LSTM) are used for processing time series data, and can also perform time series data pattern recognition, which has better accuracy than a conventional machine learning model. Convolutional Neural Network (CNN) is one of the most popular deep learning models, and has the advantage of automatically learning data features, and is also applied to the field of time series. There are two main ways to solve the time series classification problem with CNN: (1) using a one-dimensional time sequence signal as an input, and adjusting a traditional CNN structure to realize a classification task; (2) based on the huge achievement of the CNN in the image field, the time series one-dimensional signal is converted into a two-dimensional image signal, and then the CNN is utilized for further processing.
However, these conventional methods do not consider the random phase shift of the sequence pattern in the test data during the model test. The sequence mode with random offset cannot achieve ideal recognition effect by using the existing model. In addition, the existing method mostly adopts a network layer with a single channel, and the effect of extracting the characteristics of complex data is general.
Disclosure of Invention
In order to overcome the defects of the prior art and solve the problem of random phase offset of satellite telemetry time sequence data, the invention provides a method for accurately identifying a random phase shift satellite telemetry time sequence data mode, which is used for detecting the satellite telemetry time sequence data and improving the identification accuracy of the satellite telemetry time sequence data mode, thereby ensuring the accuracy of automatically identifying and judging satellite parameters or the running states of satellite components and platforms and the accuracy of satellite fault diagnosis and abnormality detection.
In order to achieve the purpose, the invention adopts the technical scheme that:
a method for accurately identifying a random phase shift satellite telemetry time series data pattern comprises the following steps:
step 1, preprocessing data, wherein the original time sequence data is random phase shift satellite telemetering time sequence data and is converted into a form which accords with neural network input specifications, namely time sequence data of a fixed time length window, and the fixed time length window is used as an observation window;
step 2, establishing a time sequence data pattern recognition model composed of a multilayer perceptron and a plurality of multi-channel characteristic network layers, inputting the obtained time sequence data of a fixed time length window as a model, respectively sending the model input into a plurality of channels of the multi-channel characteristic network layers, wherein each channel comprises a convolution kernel and a pooling structure, integrating the characteristics extracted from the plurality of channels in one multi-channel characteristic network layer, and then inputting the integrated characteristics as the input of the next multi-channel characteristic network layer, and so on;
step 3, integrating the feature vectors extracted by the last multi-channel feature network layer, connecting the integrated features into a one-dimensional vector as the input of a multi-layer sensor, wherein the multi-layer sensor is used as the last identification classification layer and consists of 2 full connection layers and a softmax output layer;
and 4, the mode type corresponding to the output maximum value of the softmax output layer is the recognition result.
The time sequence data of the fixed time length window corresponds to different modes to be identified, different identification modes represent different satellite parameters or operation states of the satellite component and the platform (namely, one time sequence data mode corresponds to one satellite parameter or operation state of the satellite component and the platform), different time sequence data exist in the observation window in the same mode, particularly different time sequence data representations exist in the observation window in the same mode with different phase offsets, and the method can accurately identify different time sequence data representations of the same mode under different phase offsets, so that the accuracy of judging the satellite parameters or the satellite component and the platform operation state is improved.
In the invention, the random phase shift satellite telemetry time sequence data is divided into stationary periodic time sequence data, non-stationary periodic time sequence data with obvious trend and other complex time sequence data except the stationary periodic time sequence data and the non-stationary periodic time sequence data according to periodicity.
For stable periodic time sequence data, dividing the data by adopting a sliding fixed time length window, taking the number of data points contained in one period of a time sequence according to the size of an observation window, carrying out normalization processing on the data, taking the maximum value MAX and the minimum value MIN of the time sequence data, mapping the maximum value to be 1, mapping the minimum value to be 0, and mapping the whole sequence to a [0,1] interval;
for non-stationary periodic time series data with an obvious trend, the whole time series data has a continuous rising or continuous falling trend, the slope of each data point of the time series data is calculated and called as the increasing rate or the decreasing rate, the change period rule of the slope, namely a slope sequence, is obtained, the obtained slope sequence is subjected to normalization processing, the maximum value and the minimum value of the slope sequence are taken, the maximum value is mapped into 1, the minimum value is mapped into 0, and the whole sequence is mapped into a [0,1] interval;
for other complex time sequence data except the two types of complex time sequence data, time window division data with preset length is adopted, normalization processing is carried out on the data, the maximum value MAX and the minimum value MIN of the time sequence data are taken, the maximum value MAX and the minimum value MIN are mapped into 1, the minimum value MAX and the minimum value MIN are mapped into 0, and the whole sequence is mapped into a [0,1] interval.
Preferably, the number of channels of each multi-channel feature network layer is the same, in each multi-channel feature network layer, the outputs of the pooling structures of all the channels are connected to the splice layer, and the output of the splice layer of the current multi-channel feature network layer is used as the input of the next multi-channel feature network layer.
Preferentially, in each multi-channel feature network layer, the convolution kernels and pooling structures of each channel are different, each channel accumulates different convolution kernels to extract features with different dimensionalities, the different convolution kernels represent different receptive fields, and the features with different scales can be fused in a feature integration stage, so that the adaptability of the network to input data is improved.
Preferentially, the number of channels in each multi-channel feature network layer is 3, the sizes of convolution kernels corresponding to 3 channels are 1 × 1, 1 × 3 and 1 × 5 respectively, the number of convolution kernels is 32, 64, 128 and 256, the number of convolution kernels is gradually increased along with the continuous deepening of the network, and the maximum pooling method is selected for the pooling structure.
Because the time phase shift of the observation window is random, one observation window can contain the time sequence data of the same mode or two different modes; if the time sequence data in the observation window corresponds to the same mode, the identification result is the mode; if the time sequence data in the observation window correspond to A, B two different modes, and the coverage rate of the A mode in the window is more than 70%, the recognition result is the A mode, otherwise, if the coverage rate of the B mode in the window is more than 70%, the recognition result is the B mode; if the coverage rate of the A, B pattern in the window is not more than 70%, the result is identified as an uncertain or unknown pattern.
Compared with the prior art, the method realizes the detection of the random phase shift satellite telemetering time sequence data based on the convolutional neural network, and identifies the random phase shift satellite telemetering time sequence data mode by using a model formed by a plurality of multi-channel characteristic network layers and a plurality of layers of sensors by utilizing the thought of combination of multiple channels and multiple scales aiming at the characteristics of high dimensionality and complex characteristics of the satellite telemetering time sequence data. Compared with the traditional machine learning model, the method has higher identification precision.
Drawings
FIG. 1 is a flow chart of the present invention for identification.
Fig. 2 is a diagram of a network model architecture according to the present invention.
FIG. 3 is a schematic diagram of a training data pattern labeling rule used in the present invention.
Fig. 4 is a graph of satellite shunt regulator shunt current.
Fig. 5 is an observation graph of different phases of shunt currents of the satellite shunt regulator in a terrestrial shadow working state.
Detailed Description
The following further describes embodiments of the present invention with reference to the accompanying drawings. FIG. 1 is a flow chart of the present invention, which comprises the following steps according to the flow chart shown in FIG. 1:
(1) obtaining random phase shifted time series data samples
The method comprises the steps of firstly sliding original time sequence data through a fixed time length window (observation window) and intercepting to obtain a plurality of time sequence data samples with the same size as the observation window. Since the observation window randomly selects the starting point of the divided data, the time sequence data samples from the same mode have left and right offsets, namely the time sequence data samples with random phase offsets.
(2) Labeling training data pattern labels
When training and testing the timing data pattern recognition problem, the correct label for any input timing data must be given. The input data randomly selects time sequence data with fixed length through an observation window. But the randomly chosen time series data of the observation window may not match the standard pattern exactly. Therefore, the present invention sets forth a method to label the pattern labels of any random phase offset timing data samples.
In the invention, an observation window is randomly placed on the standard mode sequence which is manually marked with the label, and then the mode of the new sequence is determined according to the coverage rate of the new sequence in the window on the original sequence. As shown in FIG. 3, there are 2 canonical patterns A and B, and a sequence of manually labeled canonical patterns A, B, B, A. Since the observation window randomly selects the starting point, there may occur a case where 2 standard patterns, i.e., (a, B), (B, B), and (B, a), are overlapped. The invention makes a new label rule according to the coverage rate, as shown in the following table:
where a% and b% each represent coverage in the original sequence pattern, and a% + b% is 1. C represents an uncertain (or unknown) pattern.
(3) Data pre-processing
After the tagged random phase-shifted time series data is obtained, it is pre-processed according to the method described above in order to match the network input.
Specifically, the satellite telemetry time sequence data has different laws, and the periodicity of the satellite telemetry time sequence data can be divided into stationary periodic time sequence data, non-stationary periodic time sequence data with obvious trend and other complex time sequence data.
For stable periodic time sequence data, dividing the data by adopting a sliding fixed time length window (namely an observation window), wherein the size of the observation window is equal to the number of data points contained in one period of a time sequence, carrying out normalization processing on the data, taking the maximum value MAX and the minimum value MIN of the time sequence data, mapping the maximum value to 1, mapping the minimum value to 0, and mapping the whole sequence to a [0,1] interval;
for non-stationary periodic time series data with an obvious trend, the whole time series data has a continuous rising or continuous falling trend, the slope of each data point of the time series data is calculated and called as the increasing rate or the decreasing rate, the change period rule of the slope, namely a slope sequence, is obtained, the obtained slope sequence is subjected to normalization processing, the maximum value and the minimum value of the slope sequence are taken, the maximum value is mapped into 1, the minimum value is mapped into 0, and the whole sequence is mapped into a [0,1] interval;
for other complex time sequence data, dividing the data by adopting a time window with a preset length, normalizing the data, taking the maximum value MAX and the minimum value MIN of the time sequence data, mapping the maximum value MAX to 1, mapping the minimum value MAX to 0, and mapping the whole sequence to a [0,1] interval.
In this embodiment, different data preprocessing methods are selected for time series data with different characteristics, and a time series data set T ═ T is obtained0,t1,……,tn}. n represents the number of samples in the dataset.
(4) Training random phase shift timing sequence data identification network
The invention provides a novel time sequence data pattern recognition model by utilizing the thought of combination of multiple channels and multiple scales aiming at the characteristics of high time sequence dimension and complex features, and the model is composed of multiple multi-channel feature network layers and multiple layers of sensors as shown in figure 2. And time sequence data in the observation window is used as model input and is respectively sent into a plurality of channels of the first multi-channel characteristic network layer, each channel comprises a convolution kernel and a pooling structure, and the convolution kernels and the pooling structures of different channels are different. And integrating the extracted features of a plurality of channels in one multi-channel feature network layer, and then using the integrated features as the input of the next multi-channel feature network layer, and so on. Multiple channels accumulate multiple different convolution kernels to extract features of different dimensions. Different convolution kernels represent different receptive fields, and features of different scales can be fused in the feature integration stage, so that the adaptability of the network to input data is improved.
And after integrating the feature vectors extracted by the last multi-channel feature network layer, connecting the integrated features into a one-dimensional vector as the input of the multi-layer sensor. The multilayer perceptron is used as a final identification classification layer and consists of 2 full connection layers and a softmax output layer.
In fig. 2, the number of channels in each layer of the multi-channel feature network layer is 3, different channels use different convolution kernels, and each channel uses different convolution kernels to extract features of input data in different dimensions. The sizes of convolution kernels corresponding to 3 channels are 1 multiplied by 1, 1 multiplied by 3 and 1 multiplied by 5 respectively. The number of the convolution kernels is selected from 32, 64, 128 and 256, and the number of the convolution kernels is gradually increased along with the continuous deepening of the network. The pooling structure selects the maximum pooling method. In order to ensure that the sizes of the feature graphs output by each channel are consistent before feature integration, pooling structures with different sizes are adopted, and adaptability of the network structure to different sizes is improved. And finally, aggregating the output characteristics of the multiple channels as the input of the next multi-channel characteristic network layer.
In the present embodiment, an arbitrary sample t in the time series data seti(i is more than or equal to 0 and less than or equal to n), the input network is sent into three network channels. Input data tiA new feature data of a different size is obtained through each channel. The new feature data generated by each channel passes through different pooling structures to obtain feature maps with the same size.
And finally, splicing the output characteristics of the previous layer by the splicing layer according to an end-to-end mode. And splicing the characteristic graphs output by the three channels to be used as the input of the next network layer.
When the time sequence tiAfter passing through all the multi-channel characteristic network layers, the multi-channel characteristic network layers are converted into one-dimensional vectors which are used as the input of a full connection layer.
(6) Inputting test data for identification to obtain an identification result
And sending the test time sequence data into a trained random phase shift identification network, and identifying any one piece of time sequence data to obtain an identification result. In a satellite system, different time series data patterns represent different satellite parameters or the operating states of the satellite components and platforms. The same mode may have different timing data within the observation window, and particularly, the same mode with different phase offsets may have different timing data representations within the observation window.
For example, fig. 4 shows a graph of the shunt current of the satellite shunt regulator, and it is obvious that the graph mode of the satellite in the normal operation state is greatly different from the graph mode of the satellite in the earth shadow operation state. Therefore, the method can accurately identify and judge which working state the satellite is in according to the time sequence data, and is a basis for satellite fault diagnosis. If the current is small or even 0 in the ground shadow operation, it is not abnormal. However, if in a normal working state, the shunt current is smaller or falls to 0, which may be a potential abnormality or a fault, that is, after the time sequence data mode is judged by using the method and the current working state of the satellite is obtained, simple satellite fault diagnosis can be realized by combining the shunt current of the shunt regulator and the like. Several of the situations shown in fig. 5 may occur when randomly observing the satellite shunt regulator shunt current curve, while the situation in fig. 5 is all the case when the satellites are in earth shadow operation, only the phase deviation is different. Fig. 5(a) shows the state of entering the terrestrial shadow from the normal state, (b) to (c) show the state of being in the terrestrial shadow, and (e) and (f) show the state of returning from the terrestrial shadow to the normal state. The phases of the curves are different, which indicates that the satellite should be identified as a terrestrial shadow mode in different stages of the terrestrial shadow working state when time series data pattern identification is carried out.
Claims (9)
1. A method for accurately identifying a random phase shift satellite telemetry time series data pattern is characterized by comprising the following steps:
step 1, preprocessing data, wherein the original time sequence data is random phase shift satellite telemetering time sequence data and is converted into a form which accords with neural network input specifications, namely time sequence data of a fixed time length window, and the fixed time length window is used as an observation window;
step 2, establishing a time sequence data pattern recognition model composed of a multilayer perceptron and a plurality of multi-channel characteristic network layers, inputting the obtained time sequence data of a fixed time length window as a model, respectively sending the model input into a plurality of channels of the multi-channel characteristic network layers, wherein each channel comprises a convolution kernel and a pooling structure, integrating the characteristics extracted from the plurality of channels in one multi-channel characteristic network layer, and then inputting the integrated characteristics as the input of the next multi-channel characteristic network layer, and so on;
step 3, integrating the feature vectors extracted by the last multi-channel feature network layer, connecting the integrated features into a one-dimensional vector as the input of a multi-layer sensor, wherein the multi-layer sensor is used as the last identification classification layer and consists of 2 full connection layers and a softmax output layer;
and 4, the mode type corresponding to the output maximum value of the softmax output layer is the recognition result.
2. The method as claimed in claim 1, wherein the time series data of the fixed time length window corresponds to different patterns to be identified, different identification patterns represent different satellite parameters or operation states of the satellite component and the platform, that is, one time series data pattern corresponds to one satellite parameter or operation states of the satellite component and the platform, the same pattern has different time series data in the observation window, and the same pattern with different phase offsets has different time series data representations in the observation window, the method can accurately identify different time series data representations of the same pattern with different phase offsets, thereby improving the accuracy of determining the satellite parameters or the operation states of the satellite component and the platform.
3. The method for accurately identifying a random phase shifted satellite telemetry timing data pattern as claimed in claim 1, wherein the random phase shifted satellite telemetry timing data is periodically divided into stationary periodic timing data, non-stationary periodic timing data with a significant trend and other complex timing data than the stationary periodic timing data and the non-stationary periodic timing data.
4. The method for accurately identifying the telemetry time series data mode of the random phase shift satellite as claimed in claim 3, wherein in the step 1, for stable periodic time series data, the data is divided by sliding a window with a fixed time length, the size of an observation window is the number of data points included in one period of the time series, the data is normalized, the maximum value MAX and the minimum value MIN of the time series data are taken, the maximum value is mapped to 1, the minimum value is mapped to 0, and the whole series is mapped to a [0,1] interval;
for non-stationary periodic time series data with an obvious trend, the whole time series data has a continuous rising or continuous falling trend, the slope of each data point of the time series data is calculated and called as the increasing rate or the decreasing rate, the change period rule of the slope, namely a slope sequence, is obtained, the obtained slope sequence is subjected to normalization processing, the maximum value and the minimum value of the slope sequence are taken, the maximum value is mapped into 1, the minimum value is mapped into 0, and the whole sequence is mapped into a [0,1] interval;
for other complex time sequence data except the two types of complex time sequence data, time window division data with preset length is adopted, normalization processing is carried out on the data, the maximum value MAX and the minimum value MIN of the time sequence data are taken, the maximum value MAX and the minimum value MIN are mapped into 1, the minimum value MAX and the minimum value MIN are mapped into 0, and the whole sequence is mapped into a [0,1] interval.
5. The method of claim 1, wherein the number of channels in each of the multi-channel signature network layers is the same, the output of the pooling structure of all channels in each of the multi-channel signature network layers is connected to a splice layer, and the output of the splice layer of a current multi-channel signature network layer is used as the input of a next multi-channel signature network layer.
6. The method for accurately identifying the random phase shift satellite telemetry time series data mode as claimed in claim 1 or 5, wherein in each multi-channel feature network layer, the convolution kernel and pooling structure of each channel are different, each channel accumulates different convolution kernels to extract features with different dimensions, different convolution kernels represent different receptive fields, and features with different dimensions can be fused in a feature integration stage, so that the adaptability of the network to input data is increased.
7. The method for accurately identifying the random phase shift satellite telemetry time series data mode as claimed in claim 1 or 5, wherein the number of channels in each multichannel feature network layer is 3, the sizes of convolution kernels corresponding to 3 channels are 1 x 1, 1 x 3 and 1 x 5 respectively, the number of convolution kernels is selected from 32, 64, 128 and 256, the number of convolution kernels is gradually increased along with the continuous deepening of the network, and the maximum pooling method is selected for the pooling structure.
8. The method of claim 1, wherein if the time series data in the observation window corresponds to the same pattern, the pattern is identified; if the time sequence data in the observation window correspond to A, B two different modes, and the coverage rate of the A mode in the window is more than 70%, the recognition result is the A mode, otherwise, if the coverage rate of the B mode in the window is more than 70%, the recognition result is the B mode; if the coverage rate of the A, B pattern in the window is not more than 70%, the result is identified as an uncertain or unknown pattern.
9. The method for accurately identifying the telemetry time series data mode of the random phase-shift satellite according to claim 1, wherein the operation states of the satellite component and the platform are correspondingly obtained according to the obtained telemetry time series data mode, and further the preliminary diagnosis of the satellite fault is realized by combining shunt current of a shunt regulator.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010764023.5A CN111860447A (en) | 2020-08-01 | 2020-08-01 | Method for accurately identifying telemetry time sequence data mode of random phase shift satellite |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010764023.5A CN111860447A (en) | 2020-08-01 | 2020-08-01 | Method for accurately identifying telemetry time sequence data mode of random phase shift satellite |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111860447A true CN111860447A (en) | 2020-10-30 |
Family
ID=72954223
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010764023.5A Pending CN111860447A (en) | 2020-08-01 | 2020-08-01 | Method for accurately identifying telemetry time sequence data mode of random phase shift satellite |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111860447A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112949753A (en) * | 2021-03-26 | 2021-06-11 | 西安交通大学 | Satellite telemetry time sequence data anomaly detection method based on binary relation |
| CN113268551A (en) * | 2021-05-18 | 2021-08-17 | 朱厚强 | Data monitoring method and device for unified measurement and control system |
| CN114881110A (en) * | 2022-04-02 | 2022-08-09 | 西安交通大学 | Real-time detection method for total pressure change mode in on-orbit spacecraft cabin |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934130A (en) * | 2019-02-28 | 2019-06-25 | 中国空间技术研究院 | The in-orbit real-time fault diagnosis method of satellite failure and system based on deep learning |
-
2020
- 2020-08-01 CN CN202010764023.5A patent/CN111860447A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934130A (en) * | 2019-02-28 | 2019-06-25 | 中国空间技术研究院 | The in-orbit real-time fault diagnosis method of satellite failure and system based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| JUNPENG BAO等: ""MTCNN: A Deep Neural Network for Recognizing Stochastic Phase Shifted Time Series Data"", 《ELECTRICAL ENGINEERING AND COMPUTER SCIENCE (EECS)》 * |
| 鲍军鹏等: ""卫星时序数据挖掘节点级并行与优化方法"", 《北京航空航天大学学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112949753A (en) * | 2021-03-26 | 2021-06-11 | 西安交通大学 | Satellite telemetry time sequence data anomaly detection method based on binary relation |
| CN112949753B (en) * | 2021-03-26 | 2023-10-24 | 西安交通大学 | Binary relation-based satellite telemetry time sequence data anomaly detection method |
| CN113268551A (en) * | 2021-05-18 | 2021-08-17 | 朱厚强 | Data monitoring method and device for unified measurement and control system |
| CN114881110A (en) * | 2022-04-02 | 2022-08-09 | 西安交通大学 | Real-time detection method for total pressure change mode in on-orbit spacecraft cabin |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108846835B (en) | Image change detection method based on depth separable convolutional network | |
| CN111860447A (en) | Method for accurately identifying telemetry time sequence data mode of random phase shift satellite | |
| CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
| CN110532932B (en) | Method for identifying multi-component radar signal intra-pulse modulation mode | |
| US20230152187A1 (en) | Fault signal locating and identifying method of industrial equipment based on microphone array | |
| CN113673346B (en) | Motor vibration data processing and state identification method based on multiscale SE-Resnet | |
| CN111507370A (en) | Method and device for obtaining sample image of inspection label in automatic labeling image | |
| CN111275198B (en) | Bearing abnormality detection method and system | |
| CN111860446A (en) | Detection system and method for unknown mode of satellite remote measurement time sequence data | |
| US11443137B2 (en) | Method and apparatus for detecting signal features | |
| CN112785441B (en) | Data processing method, device, terminal equipment and storage medium | |
| CN116453438B (en) | Display screen parameter detection method, device, equipment and storage medium | |
| CN112508105A (en) | Method for detecting and retrieving faults of oil extraction machine | |
| CN110766075A (en) | Tire area image comparison method and device, computer equipment and storage medium | |
| Schäfer et al. | Classifying Land Cover from Satellite Images Using Time Series Analytics. | |
| CN112613443A (en) | Robustness communication radiation source intelligent identification method based on deep learning | |
| CN114022715A (en) | Method and device for detecting lead sealing defect of high-voltage cable terminal | |
| US20040096107A1 (en) | Method and computer program product for determining an efficient feature set and an optimal threshold confidence value for a pattern recogniton classifier | |
| CN108507607B (en) | Weak signal detection method based on kernel function | |
| CN115267035B (en) | Chromatograph fault diagnosis analysis method and system | |
| CN116720095A (en) | Electrical characteristic signal clustering method for optimizing fuzzy C-means based on genetic algorithm | |
| CN116561641A (en) | Industrial equipment fault diagnosis method and system based on multi-view generation algorithm | |
| CN111062448A (en) | Equipment type recognition model training method, equipment type recognition method and device | |
| CN116662800A (en) | Rolling bearing fault diagnosis method based on self-adaptive attention mechanism | |
| CN115082713A (en) | Method, system and equipment for extracting target detection frame by introducing space contrast information |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20201030 |