CN114792115A - Telemetry signal outlier removing method, device and medium based on deconvolution reconstruction network - Google Patents
Telemetry signal outlier removing method, device and medium based on deconvolution reconstruction network Download PDFInfo
- Publication number
- CN114792115A CN114792115A CN202210542432.XA CN202210542432A CN114792115A CN 114792115 A CN114792115 A CN 114792115A CN 202210542432 A CN202210542432 A CN 202210542432A CN 114792115 A CN114792115 A CN 114792115A
- Authority
- CN
- China
- Prior art keywords
- signal data
- deconvolution
- telemetry signal
- network
- layer
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Selective Calling Equipment (AREA)
Abstract
The embodiment of the invention discloses a method, a device and a medium for removing a telemetry signal outlier based on a deconvolution reconstruction network; the method comprises the following steps: constructing a deconvolution reconstruction network based on the convolution construction and the deconvolution reconstruction; training the deconvolution reconstruction network through training data consisting of the existing measured telemetering signal data and the corresponding measured telemetering signal data timestamp to obtain a trained deconvolution reconstruction network; inputting an original telemetry signal data timestamp into the trained deconvolution reconstruction network to obtain predicted telemetry signal data corresponding to the original telemetry signal data timestamp; calculating a distance between the raw telemetry signal data corresponding to the raw telemetry signal data timestamp and the predicted telemetry signal data; and when the distance is greater than a set threshold value, determining that the original telemetered signal data corresponding to the original telemetered signal data timestamp is a wild value, and removing the wild value.
Description
Technical Field
The embodiment of the invention relates to the technical field of spacecraft data processing, in particular to a method, a device and a medium for removing a telemetry signal outlier based on a deconvolution reconstruction network.
Background
The telemetering data of the spacecraft is an important strategic resource, and the operation state of the spacecraft can be monitored and abnormal forecast in real time by utilizing the telemetering data, so that the stable operation of the spacecraft is ensured. However, due to noise interference, data loss, sensor failure and the like in the spacecraft operation space, the telemetering data can generate instantaneous jump or error, so that an abnormal point, namely a telemetering signal outlier, is generated. Therefore, in order to obtain higher quality spacecraft telemetry data, these telemetry signal outliers need to be identified and rejected.
Telemetry signal outliers for spacecraft typically include two types: the first is isolated outlier, that is, there is a jump-type peak in the early time sequence, and the adjacent data are much smaller or larger than outlier; the second category is isolated continuous outliers, which are also typically a large spike in the time series, and the left and right neighbors of several outliers occurring in succession are much smaller or larger than the outliers.
Current conventional approaches to removing telemetry signal outliers include recursion-based methods, statistical-based methods, and fitting-based methods; the method based on the recursion relationship generally includes a forward difference algorithm, a 53H algorithm, wavelet transformation, an improved kalman filter algorithm, and the like. The statistical-based methods mainly include 3-sigma criteria, roman-lofski criteria, dicke criterion, grassblobs criteria, confidence based methods of ambiguity, and the like. The fitting-based methods mainly include polynomial fitting, least square fitting, fitting based on data driving, and the like. The three methods generally have certain limitations, and the limitations affect the effect of removing the telemetry signal outliers.
Disclosure of Invention
In view of this, embodiments of the present invention are intended to provide a method, an apparatus, and a medium for removing a telemetry signal outlier based on a deconvolution reconstruction network; the method can improve the removing effect of the field value of the telemetering signal under the condition that the deconvolution reconstruction network can learn quickly.
The technical scheme of the embodiment of the invention is realized as follows:
in a first aspect, an embodiment of the present invention provides a telemetry signal outlier removing method based on a deconvolution reconstruction network, including:
constructing a deconvolution reconstruction network based on the convolution construction and the deconvolution reconstruction;
training the deconvolution reconstructed network through training data consisting of the existing measured telemetering signal data and the corresponding measured telemetering signal data timestamp to obtain a trained deconvolution reconstructed network;
inputting an original telemetry signal data timestamp into the trained deconvolution reconstruction network to obtain predicted telemetry signal data corresponding to the original telemetry signal data timestamp;
calculating a distance between the raw telemetry signal data corresponding to the raw telemetry signal data timestamp and the predicted telemetry signal data;
and when the distance is greater than a set threshold value, determining that the original telemetered signal data corresponding to the original telemetered signal data timestamp is a wild value, and removing the wild value.
In a second aspect, an embodiment of the present invention provides a telemetry signal outlier removing device based on a deconvolution reconstruction network, where the device includes: the method comprises the following steps of constructing a part, deconvoluting a reconstruction network, training a part, calculating a part, judging a part and removing a part; wherein the content of the first and second substances,
the construction portion configured to construct a deconvolution reconstructed network based on the convolution construction and the deconvolution reconstruction;
the training part is configured to train the deconvolution reconstructed network through training data consisting of existing measured telemetered signal data and corresponding measured telemetered signal data timestamps to obtain a trained deconvolution reconstructed network;
the deconvolution reconstruction network is used for inputting an original telemetry signal data time stamp to obtain predicted telemetry signal data corresponding to the original telemetry signal data time stamp;
the computing portion configured to compute a distance between the raw telemetry signal data corresponding to the raw telemetry signal data timestamp and the predicted telemetry signal data;
the judging part is configured to compare the distance with a set threshold value, and when the distance is greater than the set threshold value, the original telemetry signal data corresponding to the original telemetry signal data timestamp is determined to be a wild value, and the removing part is triggered;
the removal part is configured to remove the outliers.
In a third aspect, an embodiment of the present invention provides a computing device, where the computing device includes: a communication interface, a memory and a processor, the various components being coupled together by a bus system; wherein the content of the first and second substances,
the communication interface is used for receiving and sending signals in the process of receiving and sending information with other external network elements;
the memory for storing a computer program operable on the processor;
the processor is configured to, when running the computer program, execute the steps of the method for removing the outlier of the telemetry signal based on the deconvolution reconstruction network in the first aspect.
In a fourth aspect, an embodiment of the present invention provides a computer storage medium, where a telemetry signal outlier removing program based on a deconvolution reconstruction network is stored, and when executed by at least one processor, the telemetry signal outlier removing program based on a deconvolution reconstruction network implements the steps of the telemetry signal outlier removing method based on a deconvolution reconstruction network in the first aspect.
The embodiment of the invention provides a method, a device and a medium for removing a telemetry signal outlier based on a deconvolution reconstruction network; and predicting the telemetering data corresponding to the time stamp through a deconvolution reconstruction network after the telemetering signal data training is actually measured, and judging whether the telemetering data is a wild value or not according to the distance between the predicted value and the real value. Compared with the conventional scheme, the method avoids artificial interpolation and excessive parameter setting, eliminates the limitations caused by statistical distribution and time span, and improves the removal effect and reliability of the outlier under the condition that the deconvolution reconstruction network can learn quickly.
Drawings
Fig. 1 is a schematic flow chart of a telemetry signal outlier removing method based on a deconvolution reconstruction network according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a deconvolution reconstruction network according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a loss iteration condition of training and verification provided in the embodiment of the present invention;
FIG. 4 is a schematic diagram of a distance histogram and a kernel density estimation curve according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a process of performing outlier removal by using a DRN after training according to an embodiment of the present invention;
FIG. 6 is a graph illustrating normalized distance comparison of the major and minor axes of a track according to an embodiment of the present invention;
FIG. 7 is a graph illustrating training and validation loss comparison of track semi-major axis telemetry data provided in accordance with an embodiment of the present invention;
fig. 8 is a schematic composition diagram of a telemetry signal outlier removing device based on a deconvolution reconstruction network according to an embodiment of the present invention;
fig. 9 is a schematic composition diagram of another telemetry signal outlier removing apparatus based on a deconvolution reconstruction network according to an embodiment of the present disclosure;
fig. 10 is a schematic diagram of a specific hardware structure of a computing device according to an embodiment of the present invention.
Detailed Description
The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
For the conventional outlier removal method at present, in detail, firstly, the processed data is usually required to be continuous by the recursion-based method, and the real satellite telemetry signal is usually discontinuous due to the ground station position limitation, data packet loss and the like, so these methods need to interpolate the lost data first. The interpolation effect is greatly influenced by the data with the outliers; in addition, the method based on the recurrence relation has many hyper-parameters which need to be set manually, such as threshold values, the number of points participating in the recurrence and the like, so that the method has certain limitations. Secondly, the statistical-based method has certain limitation on the telemetry signal outlier removing effect with large variation amplitude. As the statistical laws presented by such data are quite diverse. It is possible that telemetry data over a period of time satisfies one distribution and the next period of time satisfies another distribution. In addition, fitting-based methods have poor telemetry signal outlier removal for long time stamp spans, so local fitting is usually used.
Based on the limitations exhibited by the above conventional scheme, referring to fig. 1, a method for removing a telemetry signal outlier based on a Deconvolution Reconstruction Network (DRN) provided by an embodiment of the present invention is shown, where the method may include:
s101: constructing a deconvolution reconstruction network based on the convolution construction and the deconvolution reconstruction;
s102: training the deconvolution reconstructed network through training data consisting of the existing measured telemetering signal data and the corresponding measured telemetering signal data timestamp to obtain a trained deconvolution reconstructed network;
s103: inputting an original telemetry signal data timestamp into the trained deconvolution reconstruction network to obtain predicted telemetry signal data corresponding to the original telemetry signal data timestamp;
s104: calculating a distance between the original telemetry signal data corresponding to the original telemetry signal data timestamp and the predicted telemetry signal data;
s105: and when the distance is greater than a set threshold value, determining that the original telemetered signal data corresponding to the original telemetered signal data timestamp is a wild value, and removing the wild value.
By the technical scheme shown in fig. 1, the telemetry data corresponding to the time stamp is predicted by the deconvolution reconstruction network after the telemetry signal data training is actually measured, and whether the telemetry data is a wild value or not is judged according to the distance between the predicted value and the real value. Compared with the conventional scheme, the method avoids artificial interpolation and excessive parameter setting, eliminates the limitations caused by statistical distribution and time span, and improves the removal effect and reliability of the outlier.
For the technical solution shown in fig. 1, it should be noted that, in the embodiment of the present invention, a deconvolution reconstruction network including a convolution constructing portion and a deconvolution reconstruction portion is selected, where the convolution constructing portion can perform dimension increasing processing on data input to the network, and then perform dimension reducing operation on a high-dimensional tensor through the deconvolution reconstruction portion to adapt to telemetry signal data, and the network has no latent vector. In some examples, as shown in fig. 2, the deconvolution reconstruction network includes: an Input (Input) layer, a Batch normalized Full Connection (FC (BN), a Full Connection (Batch Normalization)) layer, a first reconstructed quench layer, a first convolutional layer, a second convolutional layer (Conv (BN + ReLU)) based on the Batch normalized BN and a modified Linear Unit activation function (ReLU), a first deconvolution DeConV layer, a second deconvolution layer (DeConV (BN + ReLU)) based on the Batch normalized BN and a modified Linear Unit activation function ReLU, a third deconvolution DeConV layer, a second reconstructed quench layer, and an Output (Output) layer of the Full Connection FC. As can be seen from fig. 2, the layers from the input layer to the ConV (BN + ReLU) layer belong to the convolution Construction (Convolutional Reconstruction) part, and the layers from the first deconvolution deconrv layer to the output (fc) layer belong to the deconvolution Reconstruction (Convolutional Reconstruction) part.
For the DRN shown in fig. 2, in some examples, the output signal of the input layer is identified as X 0 The output shape is1, the number of parameters is 0; the output signal of FC (BN) layer is marked as X 1 The output shape is 16, and the number of parameters is 96; the output signal of the first reconstruction layer is identified as X 2 The output shape is (4,4,1), and the number of parameters is 0; the output signal of the first convolution layer is marked as X 3 The output shape is (4,4,16), and the number of parameters is 32; the output signal of the second convolution layer based on the linear unit activation function of batch normalization and correction is marked as X 4 The output shape is (4,4,32), and the number of parameters is 672; the output signal of the first deconvolution layer is marked as X 5 The output shape is (4,4,32), and the number of parameters is 1056; the output signal of the second deconvolution layer based on the linear unit activation function of batch normalization and correction is labeled X 6 The output shape is (4,4,16), and the number of parameters is 592; the output signal of the third deconvolution layer is marked as X 7 The output shape is (4,4,1), and the number of parameters is 17; the output signal of the second reconstruction layer is identified as X 8 The output shape is 16, and the number of parameters is 0; the output signal of the fully connected output layer is marked as X 9 The output shape is 1, and the number of parameters is 17. Through the above example, it can be seen that, since the data is directed to the satellite telemetry signal, the total number of parameters of the whole DRN is not large, so the structure of the DRN is simple and the rapid training can be realized.
For the technical solution shown in fig. 1, in some examples, the training the deconvolution reconstruction network with training data composed of existing measured telemetry signal data and corresponding measured telemetry signal data timestamps to obtain a trained deconvolution reconstruction network includes:
aiming at the initialized deconvolution reconstruction network, inputting the actually measured telemetry signal data time stamp serving as input data into the deconvolution reconstruction network;
and training the deconvolution reconstruction network according to the output of the deconvolution reconstruction network and the measured telemetered signal data corresponding to the measured telemetered signal data timestamp, and determining the parameters corresponding to each layer contained in the deconvolution reconstruction network so as to obtain the trained deconvolution reconstruction network.
For the above example, it should be noted that, in the existing measured telemetry signal data, the outlier only occupies a very small portion of the entire data, and therefore, the measured telemetry signal data is suitable for the DRN network to train, and it is not necessary to train the DRN by using the telemetry data from which the outlier is removed.
Based on the above example, in the training process of the deconvolution reconstruction network, the parameters for DRN network training are set as: the learning rate is 0.01, the optimization method is Adam algorithm, and the learning rate attenuation rule is that if the training set is not updated for 5 generations, the learning rate is reduced to half of the original rate.
For the solution shown in fig. 1, in some examples, the method further comprises: and determining the threshold value for judging the removal outlier according to a distribution histogram of the distance between the outlier and the normal telemetering signal data and a kernel density estimation curve.
For the above example, it should be noted that, for the trained DRN, for normal data in the telemetry signal data, the distance between the input and the output can be reflected in the prediction error of the DRN network; for the outlier, the distance of the outlier is far larger than the prediction error, which does not accord with the training rule of the normal data, so that the threshold value for distinguishing the normal data from the outlier can be determined by analyzing the distribution level of all differences. Regarding the level of variance distribution, the kernel density estimation is used to estimate the unknown density function, and is one of the non-parametric test methods, the kernel density estimation includes a plurality of histogram indicators, and in the embodiment of the present invention, the distances of the normal data are all within the threshold of 0.03.
For the solution shown in fig. 1, in some examples, for satellite telemetry signal data, the distance between the two data may be considered as the difference between the two data, and therefore, the calculating the distance between the original telemetry signal data corresponding to the original telemetry signal data timestamp and the predicted telemetry signal data includes: calculating a difference between the raw telemetry signal data corresponding to the raw telemetry signal data time stamp and the predicted telemetry signal data.
Based on the foregoing technical solution, in the embodiment of the present invention, a simulation experiment is performed on 6 different types of telemetry signal data of a certain model of hyperspectral satellite, the telemetry signal data in the simulation experiment is expressed by using a J2000 coordinate system, and the 6 types include: an X-axis position (X-position) in a J2000 coordinate system, a Y-axis position (Y-position) in a J2000 coordinate system, a Z-axis position (Z-position) in a J2000 coordinate system, an X-axis Velocity (X-Velocity) in a J2000 coordinate system, a Y-axis Velocity (Y-Velocity) in a J2000 coordinate system, and a Z-axis Velocity (Z-Velocity) in a J2000 coordinate system. Since this model satellite does not have the capability to delay downloading data, the telemetry signal data is limited to the location of the ground observation station and will appear as a composition of discrete data points over a period of time. For the training data set, 32 training data are obtained in each batch, the ratio of a training set to a verification validation set in the training data is 3:1, the verification set is verified after each training round epoch, and 20 rounds of epochs are respectively trained for each group of data. The details of the training data used in the embodiments of the present invention are shown in table 1:
| name (R) | Number of training sets | Number of verification sets | Training data aggregation | Number of outliers |
| X axis position (X position) | 46677 | 15558 | 62235 | 123 |
| Y axis position (Y position) | 46677 | 15558 | 62235 | 243 |
| Z axis position (Z position) | 46677 | 15558 | 62235 | 204 |
| X-axis speed (X Velocity) | 46677 | 15558 | 62235 | 121 |
| Y-axis speed (Y Velocity) | 46677 | 15558 | 62235 | 146 |
| Z-axis Velocity (Z Velocity) | 46677 | 15558 | 62235 | 133 |
TABLE 1
The iterative condition of loss (loss) of training and verification is shown in fig. 3, and it can be seen from fig. 3 that the training and verification loss rarely fluctuates during the training of DRN and can be converged rapidly in 20 rounds. Further, as for the distance histogram and kernel density estimation curve for determining the threshold, as shown in fig. 4, as can be seen from the histograms and kernel density estimation curves of 6 types of data, the threshold is preferably 0.03.
In the simulation experiment, the process of performing outlier removal by using the DRN after training is shown in fig. 5, and in fig. 5, (a1) to (a6) respectively represent the X, Y, and Z axis positions in the J2000 coordinate system and the original discrete signal data of the X, Y, and Z axis velocities in the J2000 coordinate system; (b1) and (b6) respectively represent the X, Y and Z axis positions in the J2000 coordinate system and the distances between the normalized predicted values corresponding to the X, Y and Z axis speeds in the J2000 coordinate system and the normalized raw discrete signal data, and the outliers can be removed by a threshold value of 0.03 as can be seen from (b1) to (b 6). The positions of the X axis, the Y axis and the Z axis in the J2000 coordinate system and the discrete signal data after removing the outlier corresponding to the speeds of the X axis, the Y axis and the Z axis in the J2000 coordinate system are respectively shown as (c1) to (c6), and as can be seen from fig. 5, the distance of the outlier is far greater than the distance of the normal data, and in an extreme case, if the DRN fully learns the real track dynamics and the control law, then the distance of the normal data in the original data can be perfect, and then the distance of the normal data is equal to 0.
In addition to the above data simulation, in order to further embody the superiority of the technical solution proposed by the embodiment of the present invention, the semi-spindle telemetry data of the track is adopted, and the telemetry signal outlier removing method (hereinafter, referred to as DRN) based on the deconvolution reconstruction network proposed by the above technical solution is compared with the MLP automatic encoder network (MLP-AE) and the convolution automatic encoder network (CAE) in a simulation mode, and the simulation conditions are consistent with the above data simulation experiment. The normalized distances of the semi-major axes of the tracks after the training of the three networks are shown in fig. 6, and it can be seen from fig. 6 that although the three networks can identify most outliers, the performance of the DRN is better. This is because DRN can find more outliers than MLP-AE and CAE. In addition, as shown in fig. 7, the loss of training and verification of the DRN on the orbit semi-major axis telemetry data is lower than that of MLP-AE and CAE, which means that the DRN has better training effect than the other two networks, and therefore more outliers can be found.
Referring to fig. 8, it shows a telemetry signal outlier removing apparatus 80 based on a deconvolution reconstruction network according to an embodiment of the present invention, where the apparatus 80 includes: construction section 801, deconvolution reconstruction network 802, training section 803, calculation section 804, decision section 805, and removal section 806; wherein, the first and the second end of the pipe are connected with each other,
the construction part 801 is configured to construct a deconvolution reconstruction network 802 based on the convolution construction and the deconvolution reconstruction;
the training section 803 is configured to train the deconvolution reconstruction network 802 by using training data composed of existing measured telemetry signal data and corresponding measured telemetry signal data timestamps, to obtain a trained deconvolution reconstruction network 802;
the deconvolution reconstruction network 802 is configured to input an original telemetry signal data timestamp to obtain predicted telemetry signal data corresponding to the original telemetry signal data timestamp;
the calculating portion 804 configured to calculate a distance between the raw telemetry signal data corresponding to the raw telemetry signal data time stamp and the predicted telemetry signal data;
the decision section 805 configured to compare the distance with a set threshold, and when the distance is greater than the set threshold, determine that the original telemetry signal data corresponding to the original telemetry signal data timestamp is a outlier, and trigger the removal section 806;
the removing part 806 is configured to remove the outliers.
In the above solution, the deconvolution reconstruction network 802 includes: the device comprises an input layer, a batch standardized full-connection FC (BN) layer, a first reconstruction layer, a first convolution layer, a second convolution layer based on a batch standardized and corrected linear unit activation function, a first deconvolution layer, a second deconvolution layer based on a batch standardized and corrected linear unit activation function, a third deconvolution layer, a second reconstruction layer and a full-connection output layer.
In the scheme, the output shape of the input layer is 1, and the number of parameters is 0; the output shape of the FC (BN) layer is 16, and the number of parameters is 96; the output shape of the first reconstruction layer is (4,4,1), and the number of parameters is 0; the output shape of the first convolution layer is (4,4,16), and the number of parameters is 32; the output shape of the second convolution layer based on the batch normalization and correction linear unit activation function is (4,4,32), and the number of parameters is 672; the output shape of the first deconvolution layer is (4,4,32), and the number of parameters is 1056; the output shape of the second deconvolution layer based on the batch standardization and correction linear unit activation function is (4,4,16), and the number of parameters is 592; the output shape of the third deconvolution layer is (4,4,1), and the number of parameters is 17; the output shape of the second reconstruction layer is 16, and the number of parameters is 0; the output shape of the fully connected output layer is 1, and the number of parameters is 17.
In the above scheme, the training section 803 is configured to:
inputting the measured telemetry signal data time stamp as input data to the deconvolution reconstruction network 802 for the initialized deconvolution reconstruction network 802;
training the deconvolution reconstructed network 802 according to the output of the deconvolution reconstructed network 802 and the measured telemetry signal data corresponding to the measured telemetry signal data timestamp, and determining parameters corresponding to each layer included in the deconvolution reconstructed network 802 to obtain the trained deconvolution reconstructed network 802.
In the above scheme, in the training process of the deconvolution reconstructed network 802, the network training parameters are set as: the learning rate is 0.01, the optimization method is Adam algorithm, and the learning rate attenuation rule is that if the training set is not updated in 5 generations, the learning rate is reduced to half of the original learning rate.
In the above scheme, referring to fig. 9, the apparatus further includes: a determining section 807 configured to determine the threshold value for making the outlier removal decision based on the partition histogram of the distance between the outlier and the normal telemetry signal data and the kernel density estimation curve.
In the above scheme, the calculating part 804 is configured to calculate a difference between the raw telemetry signal data corresponding to the raw telemetry signal data time stamp and the predicted telemetry signal data.
It is understood that in this embodiment, "part" may be part of a circuit, part of a processor, part of a program or software, etc., and may also be a unit, and may also be a module or a non-modular.
In addition, each component in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit. The integrated unit can be realized in a hardware mode, and can also be realized in a software functional module mode.
Based on the understanding that the technical solution of the present embodiment essentially or partly contributes to the prior art, or all or part of the technical solution may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the method of the present embodiment. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Therefore, the present embodiment provides a computer storage medium, where a telemetry signal outlier removing program based on a deconvolution reconstruction network is stored, and when executed by at least one processor, the telemetry signal outlier removing program based on a deconvolution reconstruction network implements the steps of the telemetry signal outlier removing method based on a deconvolution reconstruction network in the foregoing technical solution.
Referring to fig. 10, which illustrates a specific hardware structure of a computing device 100 capable of implementing the deconvolution reconstruction network-based telemetry signal outlier removing device 80 according to an embodiment of the present invention, the computing device 100 may be a wireless device, a mobile or cellular phone (including a so-called smart phone), a Personal Digital Assistant (PDA), a video game console (including a video display, a mobile video game device, and a mobile video conference unit), a laptop computer, a desktop computer, a television set-top box, a tablet computing device, an e-book reader, a fixed or mobile media player, and so on. The computing device 100 includes: a communication interface 1001, a memory 1002, and a processor 1003; the various components are coupled together by a bus system 1004. It is understood that the bus system 1004 is used to enable connected communication between these components. The bus system 1004 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, the various buses are designated in figure 10 as the bus system 1004. Wherein, the first and the second end of the pipe are connected with each other,
the communication interface 1001 is configured to receive and transmit signals during information transmission and reception with other external network elements;
the memory 1002 is used for storing a computer program capable of running on the processor 1003;
the processor 1003 is configured to, when running the computer program, execute the steps of the telemetry signal outlier removing method based on the deconvolution reconstruction network 802 in the above technical solution.
It is to be understood that the memory 1002 in embodiments of the present invention may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The non-volatile Memory may be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash Memory. The volatile Memory may be a Random Access Memory (RAM) which serves as an external cache. By way of example, and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 1002 of the subject systems and methods is intended to comprise, without being limited to, these and any other suitable types of memory.
And the processor 1003 may be an integrated circuit chip having signal processing capability. In implementation, the steps of the method may be implemented by integrated logic circuits of hardware in the processor 1003 or instructions in the form of software. The Processor 1003 may be a general-purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, or discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in the memory 1002, and the processor 1003 reads the information in the memory 1002 and completes the steps of the method in combination with the hardware.
It is to be understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or any combination thereof. For a hardware implementation, the Processing units may be implemented within one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), general purpose processors, controllers, micro-controllers, microprocessors, other electronic units designed to perform the functions described herein, or a combination thereof.
For a software implementation, the techniques described herein may be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The software codes may be stored in a memory and executed by a processor. The memory may be implemented within the processor or external to the processor.
It can be understood that the exemplary technical solutions of the above-mentioned telemetry signal outlier removing device 80 and the computing device 100 based on the deconvolution reconstruction network belong to the same concept as the technical solution of the above-mentioned telemetry signal outlier removing method based on the deconvolution reconstruction network, and therefore, the details that are not described in detail in the technical solutions of the above-mentioned telemetry signal outlier removing device 80 and the computing device 100 based on the deconvolution reconstruction network can be referred to the description of the technical solution of the above-mentioned telemetry signal outlier removing method based on the deconvolution reconstruction network. The embodiments of the present invention will not be described in detail herein.
It should be noted that: the technical schemes described in the embodiments of the present invention can be combined arbitrarily without conflict.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of the changes or substitutions within the technical scope of the present invention, and shall cover the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (10)
1. A telemetry signal outlier removing method based on a deconvolution reconstruction network is characterized by comprising the following steps:
constructing a deconvolution reconstruction network based on the convolution construction and the deconvolution reconstruction;
training the deconvolution reconstructed network through training data consisting of the existing measured telemetering signal data and the corresponding measured telemetering signal data timestamp to obtain a trained deconvolution reconstructed network;
inputting an original telemetry signal data timestamp into the trained deconvolution reconstruction network to obtain predicted telemetry signal data corresponding to the original telemetry signal data timestamp;
calculating a distance between the original telemetry signal data corresponding to the original telemetry signal data timestamp and the predicted telemetry signal data;
and when the distance is greater than a set threshold value, determining that the original telemetered signal data corresponding to the original telemetered signal data timestamp is a wild value, and removing the wild value.
2. The method of claim 1, wherein said deconvolution reconstruction network comprises: the system comprises an input layer, a batch standardized full connection FC (BN) layer, a first reconstruction layer, a first convolution layer, a second convolution layer based on a batch standardized and corrected linear unit activation function, a first deconvolution layer, a second deconvolution layer based on a batch standardized and corrected linear unit activation function, a third deconvolution layer, a second reconstruction layer and a full connection output layer.
3. The method of claim 2, wherein the output shape of the input layer is 1, the number of parameters is 0; the output shape of the FC (BN) layer is 16, and the number of parameters is 96; the output shape of the first reconstruction layer is (4,4,1), and the number of parameters is 0; the output shape of the first convolution layer is (4,4,16), and the number of parameters is 32; the output shape of the second convolution layer based on the batch normalization and correction linear unit activation function is (4,4,32), and the number of parameters is 672; the output shape of the first deconvolution layer is (4,4,32), and the number of parameters is 1056; the output shape of the second deconvolution layer based on the batch standardization and correction linear unit activation function is (4,4,16), and the number of parameters is 592; the output shape of the third deconvolution layer is (4,4,1), and the number of parameters is 17; the output shape of the second reconstruction layer is 16, and the number of parameters is 0; the output shape of the fully connected output layer is 1, and the number of parameters is 17.
4. The method of claim 1, wherein the training the deconvolution reconstructed network with training data comprising existing measured telemetry signal data and corresponding measured telemetry signal data timestamps to obtain a trained deconvolution reconstructed network comprises:
aiming at the initialized deconvolution reconstruction network, inputting the actually measured telemetry signal data time stamp serving as input data into the deconvolution reconstruction network;
and training the deconvolution reconstruction network according to the output of the deconvolution reconstruction network and the measured telemetered signal data corresponding to the measured telemetered signal data timestamp, and determining the parameters corresponding to each layer contained in the deconvolution reconstruction network so as to obtain the trained deconvolution reconstruction network.
5. The method of claim 4, wherein in training the deconvolution reconstructed network, the network training parameters are set to: the learning rate is 0.01, the optimization method is Adam algorithm, and the learning rate attenuation rule is that if the training set is not updated in 5 generations, the learning rate is reduced to half of the original learning rate.
6. The method of claim 1, further comprising: and determining the threshold value for judging the removal outlier according to a distribution histogram of the distance between the outlier and the normal telemetering signal data and a nuclear density estimation curve.
7. The method of claim 1, wherein the calculating the distance between the raw telemetry signal data corresponding to the raw telemetry signal data time stamp and the predicted telemetry signal data comprises:
calculating a difference between the raw telemetry signal data corresponding to the raw telemetry signal data time stamp and the predicted telemetry signal data.
8. A telemetry signal outlier removing device based on a deconvolution reconstruction network, the device comprising: the method comprises the following steps of constructing a part, deconvolution reconstruction network, training part, calculating part, judging part and removing part; wherein the content of the first and second substances,
the construction portion configured to construct a deconvolution reconstructed network based on the convolution construction and the deconvolution reconstruction;
the training part is configured to train the deconvolution reconstructed network through training data consisting of existing measured telemetered signal data and corresponding measured telemetered signal data timestamps to obtain a trained deconvolution reconstructed network;
the deconvolution reconstruction network is used for inputting an original telemetry signal data time stamp to obtain predicted telemetry signal data corresponding to the original telemetry signal data time stamp;
the computing portion configured to compute a distance between the raw telemetry signal data corresponding to the raw telemetry signal data timestamp and the predicted telemetry signal data;
the judging part is configured to compare the distance with a set threshold value, and when the distance is greater than the set threshold value, determine the original telemetry signal data corresponding to the original telemetry signal data timestamp as a wild value and trigger the removing part;
the removal part is configured to remove the outliers.
9. A computing device, wherein the computing device comprises: a communication interface, a memory and a processor, the various components being coupled together by a bus system; wherein the content of the first and second substances,
the communication interface is used for receiving and sending signals in the process of receiving and sending information with other external network elements;
the memory for storing a computer program operable on the processor;
the processor, when running the computer program, is configured to perform the steps of the deconvolution reconstruction network-based telemetry signal outlier removal method of any of claims 1-7.
10. A computer storage medium, characterized in that the computer storage medium stores a deconvolution reconstruction network-based telemetry signal outlier removal program, which when executed by at least one processor implements the steps of the deconvolution reconstruction network-based telemetry signal outlier removal method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210542432.XA CN114792115B (en) | 2022-05-17 | 2022-05-17 | Telemetry signal outlier removing method, device and medium based on deconvolution reconstruction network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210542432.XA CN114792115B (en) | 2022-05-17 | 2022-05-17 | Telemetry signal outlier removing method, device and medium based on deconvolution reconstruction network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114792115A true CN114792115A (en) | 2022-07-26 |
| CN114792115B CN114792115B (en) | 2023-04-07 |
Family
ID=82463172
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210542432.XA Active CN114792115B (en) | 2022-05-17 | 2022-05-17 | Telemetry signal outlier removing method, device and medium based on deconvolution reconstruction network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114792115B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204449A (en) * | 2016-07-06 | 2016-12-07 | 安徽工业大学 | A kind of single image super resolution ratio reconstruction method based on symmetrical degree of depth network |
| CN107123089A (en) * | 2017-04-24 | 2017-09-01 | 中国科学院遥感与数字地球研究所 | Remote sensing images super-resolution reconstruction method and system based on depth convolutional network |
| US20180149753A1 (en) * | 2016-11-30 | 2018-05-31 | Yujin Robot Co., Ltd. | Ridar apparatus based on time of flight and moving object |
| US20180189564A1 (en) * | 2016-12-30 | 2018-07-05 | International Business Machines Corporation | Method and system for crop type identification using satellite observation and weather data |
| CN110427893A (en) * | 2019-08-06 | 2019-11-08 | 西安电子科技大学 | A kind of specific emitter identification method, apparatus and computer storage medium based on convolutional neural networks |
| CN111258325A (en) * | 2020-01-08 | 2020-06-09 | 哈尔滨工业大学 | High-precision and high-performance attitude fault-tolerant control method and device for ground remote sensing satellite and computer storage medium |
| CN112132006A (en) * | 2020-09-21 | 2020-12-25 | 西南交通大学 | Intelligent forest land and building extraction method for cultivated land protection |
| CN113985408A (en) * | 2021-09-13 | 2022-01-28 | 南京航空航天大学 | Inverse synthetic aperture radar imaging method combining gate unit and transfer learning |
| CN114169232A (en) * | 2021-11-29 | 2022-03-11 | 中国科学技术大学 | Full-time-period three-dimensional atmospheric pollutant reconstruction method and device, computer equipment and storage medium |
-
2022
- 2022-05-17 CN CN202210542432.XA patent/CN114792115B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204449A (en) * | 2016-07-06 | 2016-12-07 | 安徽工业大学 | A kind of single image super resolution ratio reconstruction method based on symmetrical degree of depth network |
| US20180149753A1 (en) * | 2016-11-30 | 2018-05-31 | Yujin Robot Co., Ltd. | Ridar apparatus based on time of flight and moving object |
| US20180189564A1 (en) * | 2016-12-30 | 2018-07-05 | International Business Machines Corporation | Method and system for crop type identification using satellite observation and weather data |
| CN107123089A (en) * | 2017-04-24 | 2017-09-01 | 中国科学院遥感与数字地球研究所 | Remote sensing images super-resolution reconstruction method and system based on depth convolutional network |
| CN110427893A (en) * | 2019-08-06 | 2019-11-08 | 西安电子科技大学 | A kind of specific emitter identification method, apparatus and computer storage medium based on convolutional neural networks |
| CN111258325A (en) * | 2020-01-08 | 2020-06-09 | 哈尔滨工业大学 | High-precision and high-performance attitude fault-tolerant control method and device for ground remote sensing satellite and computer storage medium |
| CN112132006A (en) * | 2020-09-21 | 2020-12-25 | 西南交通大学 | Intelligent forest land and building extraction method for cultivated land protection |
| CN113985408A (en) * | 2021-09-13 | 2022-01-28 | 南京航空航天大学 | Inverse synthetic aperture radar imaging method combining gate unit and transfer learning |
| CN114169232A (en) * | 2021-11-29 | 2022-03-11 | 中国科学技术大学 | Full-time-period three-dimensional atmospheric pollutant reconstruction method and device, computer equipment and storage medium |
Non-Patent Citations (6)
| Title |
|---|
| HE Z等: "A fast real-time algorithm for outlier elimination based on polynomial modeling", 《2019 IEEE 8TH DATA DRIVEN CONTROL AND LEARNING SYSTEMS CONFERENCE (DDCLS)》 * |
| ZUO J等: "Research on image super-resolution algorithm based on mixed deep convolutional networks", 《COMPUTERS AND ELECTRICAL ENGINEERING》 * |
| 元伟等: "基于盲反卷积的超分辨率图像盲复原算法", 《天津大学学报(自然科学与工程技术版)》 * |
| 孙国栋等: "不均匀光照和重影的仪表图像二值化方法", 《应用光学》 * |
| 张砚等: "基于外存和凸集投影法的遥感图像超分辨率方法", 《清华大学学报(自然科学版)》 * |
| 王云艳等: "基于反卷积高层特征的遥感地物图像分类", 《计算机工程与应用》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114792115B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jwo et al. | Neural network-based GPS GDOP approximation and classification | |
| CN109165691B (en) | Training method and device for model for identifying cheating users and electronic equipment | |
| Chen et al. | Distributed data association for multi-target tracking in sensor networks | |
| CN115358487A (en) | Federal learning aggregation optimization system and method for power data sharing | |
| CN112257848B (en) | Method for determining logic core layout, model training method, electronic device and medium | |
| CN115700494A (en) | Rail transit monitoring data cleaning method and system based on Bayesian inference | |
| CN114244835A (en) | Decentralized self-adaptive collaborative training method and device based on block chain | |
| CN115409091A (en) | Method, device, equipment and medium for unsupervised satellite anomaly detection based on TDRAE | |
| CN111752691A (en) | AI (artificial intelligence) calculation graph sorting method, device, equipment and storage medium | |
| CN114792115B (en) | Telemetry signal outlier removing method, device and medium based on deconvolution reconstruction network | |
| WO2021196800A1 (en) | Method and apparatus for obtaining artificial intelligence model, device, and storage medium | |
| CN116894189A (en) | Model training method, device, equipment and readable storage medium | |
| CN112131274A (en) | Method, device and equipment for detecting time series abnormal points and readable storage medium | |
| CN111783883A (en) | Abnormal data detection method and device | |
| CN113342474B (en) | Method, equipment and storage medium for predicting customer flow and training model | |
| CN113127274A (en) | Disk failure prediction method, device, equipment and computer storage medium | |
| CN113207080B (en) | Data cooperative transmission method and device in VANET and electronic equipment | |
| CN113051854B (en) | Self-adaptive learning type power modeling method and system based on hardware structure perception | |
| CN114390563A (en) | 5G cell residence capacity evaluation method and device and electronic equipment | |
| CN114356235A (en) | Data standardization processing method and device, electronic equipment and storage medium | |
| JP5373967B2 (en) | Sphere detector that performs depth-first search until finished | |
| CN111191827A (en) | Method and device for predicting data fluctuation trend | |
| CN115002002B (en) | Equipment system information communication capability assessment method, device, equipment and medium | |
| CN115269245B (en) | Memory fault processing method and computing device | |
| CN113378977B (en) | Recording data processing method and device |
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 |