Satellite telemetry parameter anomaly detection system and method based on neural contrast enhanced auto-encoder
The neural contrast enhancement autoencoder system addresses temporal feature modeling and representativeness issues by using data augmentation and LSTM networks for satellite telemetry parameter anomaly detection, achieving superior performance in anomaly detection tasks.
Patent Information
- Application Number
- CN202411082706.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-08-08
AI Technical Summary
Existing satellite telemetry parameter anomaly detection methods face challenges in temporal feature modeling and lack of representativeness in feature extraction due to limited normal sample training, especially for time context anomalies.
A neural contrast enhancement autoencoder system utilizing a neural converter for data enhancement and LSTM autoencoders for feature extraction, combined with deterministic contrastive loss and reconstruction loss for end-to-end training, to generate diverse training samples and enhance feature representation.
The system effectively detects satellite telemetry parameter anomalies with improved accuracy, achieving superior F1* scores on benchmark datasets, demonstrating enhanced generalization and representation learning capabilities.
Smart Images

Figure CN119089301A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the field of satellite status anomaly detection, and specifically relates to a satellite telemetry parameter anomaly detection system and method based on a neural contrast enhancement autoencoder. Background Art
[0002] As a precision spacecraft, artificial satellites play a vital role in national defense security, economic development, ecological protection, and national life. However, due to factors such as adverse space environment and component aging, satellite status may be abnormal and inconsistent with expectations, posing a threat to the safety of satellite in-orbit operation and the reliability of target mission execution. Therefore, it is crucial to monitor the satellite in-orbit status. Anomaly detection can provide strong decision-making information support for ground control personnel and help ensure the stable and reliable operation of satellites. It has important academic research significance and engineering application value.
[0003] According to the manifestation, satellite on-orbit status anomalies can be divided into three categories: single point anomalies, correlation anomalies, and time context anomalies. The most complex one is the time context anomaly, which refers to the abnormal change of telemetry parameters in the time dimension. Multiple consecutive single points violate the law of the original sequence and show a trend or morphological abnormality that does not exceed the threshold. Time context anomalies may be caused by changes in the satellite's operating status or events that exceed the predetermined behavior range, such as the continuous slow change of current caused by component aging.
[0004] As a direct reflection of the satellite's working mode and operating status, telemetry parameters are the key basis for ground control personnel to evaluate satellite performance and health. In recent years, with the increasing sophistication of satellite instrument components and the rapid increase in the number of satellites in orbit, the dimensions and data volume of telemetry parameters have also increased significantly, which has put forward higher requirements for the detection of satellite operating status anomalies. Satellite on-orbit status anomalies are characterized by diverse phenomena, complex correlations between parameters, and time-dependent information superposition. Traditional threshold-based, model-based, and rule-based anomaly detection methods can no longer meet the needs of current satellite anomaly detection tasks due to the lack of modeling capabilities for the complex characteristics of telemetry parameters.
[0005] Compared with traditional methods, data-driven methods do not rely on prior knowledge and expert experience, can automatically mine potential patterns and laws in historical telemetry parameters, and have powerful representation learning and modeling capabilities, with significant advantages. Deep learning is one of the commonly used data-driven methods and has been widely used in the field of telemetry parameter anomaly detection. However, the time context anomaly detection method based on deep learning still faces problems such as insufficient temporal feature modeling and lack of representativeness in feature extraction. Satellite telemetry parameters have complex temporal characteristics. In order to mine time-dependent information, normal time series windows are usually used as input for semi-supervised learning to extract important representations. However, compared with abnormal situations that may appear in any form, the normal samples used for training are relatively limited, which makes it difficult for machine learning methods to learn representative generalized features, limiting the generalization ability of the model and thus affecting the effect of anomaly detection. In-depth research and solutions to these problems will help to improve the existing telemetry parameter monitoring methods and further realize the intelligent and automated detection of satellite abnormal conditions. Summary of the invention
[0006] The purpose of this application is to overcome the defects of the prior art in that the temporal feature modeling is insufficient and the feature extraction lacks representativeness.
[0007] In order to achieve the above objectives, the present application proposes a satellite telemetry parameter anomaly detection system based on a neural contrast enhancement autoencoder, characterized in that the system comprises:
[0008] A neural converter data enhancement module, including a neural converter, is used to convert the original data through the neural converter, and then enhance the converted data to obtain multiple enhanced samples of the original sample, and output a sample set including the original subsequence and the sequence synthesized by data enhancement; and
[0009] The autoencoder feature module includes two symmetrical LSTM networks, an encoder and a decoder, which are used to map samples to feature space using the encoder, and then reconstruct the feature representation through the decoder to obtain the anomaly labels at each time point in the sequence to be tested.
[0010] As an improvement to the above system, the neural converter is a recurrent neural network or a Transformer network.
[0011] As an improvement on the above system, the neural transformer is trained using a deterministic contrastive loss method.
[0012] As an improvement of the above system, the method for enhancing the converted data includes: directly outputting the converted samples as enhanced samples, multiplying the converted samples with the original samples to obtain enhanced sample output, and adding the converted samples with the original samples to obtain enhanced sample output.
[0013] As an improvement to the above system, the training of the autoencoder feature module adopts reconstruction error as the loss function.
[0014] As an improvement of the above system, the training of the system adopts an end-to-end method, and the training process includes:
[0015] Input the historical normal telemetry sequence, perform preprocessing and sliding window segmentation, and obtain a subsequence set for training;
[0016] Inputting the subsequence set into the neural transformer data enhancement module to generate multiple enhanced samples of the original samples, thereby obtaining a sample set including the original subsequence and the sequence synthesized by data enhancement;
[0017] The sample set is input into the autoencoder feature module, the encoder maps the sample to the feature space, and then the feature representation is reconstructed by the decoder for feature learning;
[0018] Contrastive loss is used to distinguish samples from different neural converters in the feature space, and reconstruction loss is used to measure the feature representation ability. By jointly optimizing the contrastive loss and reconstruction loss, the network parameters of the neural converter data enhancement module and the autoencoder feature module are iteratively updated.
[0019] The present application also proposes a method for detecting satellite telemetry parameter anomalies based on a neural contrast enhanced autoencoder, which is implemented based on the above system. The method includes:
[0020] The satellite telemetry parameters are preprocessed and segmented and then input into the trained satellite telemetry parameter anomaly detection system based on neural contrast enhancement autoencoder to obtain the anomaly labels of the satellite telemetry parameters at each time point.
[0021] Compared with the prior art, the advantages of this application are:
[0022] By conducting comparative experiments on two public telemetry parameter datasets SMAP and MSL, which are commonly used in the field of spacecraft anomaly detection, the superior performance of the NCAAE method proposed in this paper is verified, and its anomaly detection effect is better than the current mainstream deep anomaly detection method. The proposed method obtained the best F1* scores of 0.955 and 0.962 on the SMAP and MSL datasets, respectively, which are 4.83% and 0.94% higher than the suboptimal baseline methods. At the same time, by removing the neural converter and conducting ablation experiments, the effectiveness of self-supervised learning through data augmentation in the proposed method is verified. The results of a series of experiments prove that the proposed method has strong practical ability and can effectively detect telemetry parameter time context anomalies. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1The figure shows the structure diagram of the satellite telemetry parameter anomaly detection system based on neural contrast enhancement autoencoder;
[0024] Figure 2 Shown is a stacked diagram of the evaluation index matrix of the results of each method on the SMAP dataset;
[0025] Figure 3 Shown is a stacked chart of the evaluation indicator matrix of the results of each method on the MSL dataset. DETAILED DESCRIPTION
[0026] The technical solution of the present application is described in detail below with reference to the accompanying drawings.
[0027] This application proposes a satellite telemetry parameter anomaly detection system and method based on a neural contrast enhancement autoencoder for the temporal context anomaly of satellite telemetry parameters. In order to solve the problem that the learned features lack representativeness due to limited application of normal sample training, neural conversion learning is introduced, and a neural contrast enhancement autoencoder is proposed. This model uses neural networks to learn different conversion methods and perform data enhancement, which expands the diversity of training samples, thereby prompting the autoencoder to learn more generalized and rich representations. At the same time, the model uses enhanced samples from different sources in the separated feature space as auxiliary tasks for self-supervised learning, solving the problem of lack of supervision information in semi-supervised learning. By jointly optimizing the neural converter and the autoencoder network, the learned model can reconstruct the normal sequence with high quality, so that the abnormal subsequence can be distinguished due to the large reconstruction error, and the detection of temporal context anomalies is realized.
[0028] The present invention aims at the problem that the learned representation lacks representativeness due to limited training samples in satellite telemetry parameter time context anomaly detection, and proposes a satellite telemetry parameter anomaly detection system and method based on neural contrast enhancement autoencoder. The method uses a neural network to learn different conversion methods and generates a variety of enhanced samples based on the original samples, thereby enriching the diversity of training samples. The neural converter for data enhancement and the LSTM autoencoder for extracting time series features together constitute the neural contrast enhancement autoencoder. By jointly optimizing the parameters of the two networks, the proposed method solves the problems of lack of supervision information and single training samples, prompting the autoencoder to learn more generalized and more important representations, so that the model can reconstruct normal sequences with high quality and detect abnormal subsequences through high reconstruction errors.
[0029] In the context of the continuous development of automation and intelligence, this paper aims to build an anomaly detection model based on deep learning to enhance the ability to mine the deep information and rules contained in telemetry parameters and improve the accuracy of anomaly detection. The proposed method provides a theoretical methodological basis for ground control personnel to monitor the temporal context anomalies of satellite telemetry parameters, which helps to improve the existing telemetry parameter monitoring method, thereby effectively identifying the potential risks faced by satellites.
[0030] The satellite telemetry parameter anomaly detection system and method based on neural contrast enhancement autoencoder proposed in the present invention consists of two stages: detection model training and abnormal state determination. Figure 1 As shown in the figure, the system consists of two main components: a neural converter data enhancement module and an autoencoder feature extraction module. It is called a neural contrast enhanced autoencoder, which is an end-to-end deep self-supervised anomaly detection model. Through the detection model training, it is possible to learn the potential rules in the normal telemetry sequence and build a reconstruction-based model for abnormal state determination. For the sequence to be tested, the data rules of normal samples are similar to the original training data and can be well reconstructed by the model, while the errors of abnormal samples reconstructed by the model are large, so they can be distinguished.
[0031] The training process of the model is summarized as follows: first, the historical normal telemetry sequence is input, and preprocessed and sliding window split to obtain a set of subsequences for training; next, these subsequences are input into the neural converter to generate multiple enhanced samples of the original samples, and a sample set including the original subsequences and sequences synthesized by data enhancement is obtained; then, the sample set is input into the LSTM autoencoder, the encoder maps the samples to the feature space, and then the feature representation is reconstructed by the decoder for feature learning; in the above process, the contrast loss is used to distinguish samples from different neural converters in the feature space, and the reconstruction loss is used to measure the feature representation ability; finally, by jointly optimizing the contrast loss and the reconstruction loss, the network parameters of the neural converter and the autoencoder are iteratively updated to improve the performance of the overall model.
[0032] The steps of anomaly detection based on the trained model are summarized as follows: First, the sequence to be tested is preprocessed in the same way as the training set, and the original sequence is divided into a set of subsequences through a sliding window; then, these subsequences are input into the LSTM autoencoder to obtain the reconstructed sequence, and the reconstruction error at each time point in the sequence is calculated as the anomaly score at that time point; next, the subsequences are spliced according to the segmentation method to obtain the anomaly score sequence corresponding to the original sequence to be tested; finally, the anomaly score is compared with the preset judgment threshold to obtain the anomaly label of each time point in the sequence to be tested. Specifically, if the anomaly score is greater than the threshold, it is judged as abnormal, otherwise it is judged as normal.
[0033] (1) Neural Transformer Data Augmentation
[0034] The anomaly detection method based on self-supervised learning promotes feature representation learning through constructed auxiliary tasks, and builds anomaly discriminators based on the extracted features to detect anomalies. However, for other types of data besides images, such as time series data and tabular data, it is difficult to design manual transformation methods for sample enhancement. Therefore, for satellite telemetry sequences, the present invention proposes a data enhancement method using neural network learning transformation, called a neural transformer. This method replaces manual transformation with neural network transformation to learn different transformation methods, so that the transformed samples share semantic information with the original samples, and the samples generated by different transformation methods are easy to distinguish. The neural network model can be a recurrent neural network or a Transformer network, etc.
[0035] The pseudo code of the specific method of data enhancement using neural converter is shown in Table 1. The main part of data enhancement is the proposed neural converter. After the original sample is input, it is converted by the neural network to generate intermediate samples, which are then processed by different data processing methods to obtain the final enhanced samples. There are three common processing methods: (1) directly outputting the intermediate samples as enhanced samples; (2) multiplying the intermediate samples with the original samples to obtain the enhanced sample output; (3) adding the intermediate samples to the original samples to obtain the enhanced sample output.
[0036] Table 1 Pseudocode for neural transformer learning data augmentation
[0037]
[0038]
[0039] On this basis, the idea of contrastive learning is used to combine neural converters and feature learning methods for self-supervised anomaly detection. First, the neural converter is trained to promote learning different conversion methods. Then, multiple enhanced samples are generated using the original samples through the learned different converters, and a set of samples with category pseudo-labels is obtained. Next, feature representation learning is performed on the samples to map the samples into the feature space and to distinguish samples from different sources. Finally, the neural converter and feature learning are jointly trained to iteratively optimize the network parameters to obtain an anomaly detection model.
[0040] Different from the way of extracting negative samples from the noise distribution for enhancement, the neural converter proposed in this application deterministically constructs enhanced samples from the original samples. Deterministic Contrastive Loss (DCL) is used for training. DCL is a non-random target and does not require additional regularization. It encourages the neural converter to learn different transformation methods to generate different enhanced samples, while each sample shares semantic information with the original sample.
[0041] Assume that there are N samples in the original data space χ There are K conversion modes in the neural converter {T1, T2, …, T K |T k :χ→χ}. DCL encourages the sample x generated by the kth transformer k =T k (x) is similar to its original sample x, but the sample x generated by other transformers is similar to the original sample x. l =T l (x), l≠k are different. Define the conversion score between two enhanced samples as h(x k x l ):
[0042]
[0043] Where τ is the temperature parameter, f φ (·): χ→γ is a feature extractor that maps samples from the original space χ to the feature space γ. The similarity between enhanced samples is measured by the cosine similarity of the samples in the feature space:
[0044]
[0045] Among them, ||·|| means taking the modulus of a vector.
[0046] Define the deterministic contrast loss as
[0047]
[0048] In the formula, Denotes the mathematical expectation E. The numerator promotes the alignment of the enhanced samples generated by the neural converter with the original samples in the feature space, while the denominator promotes the separation of enhanced samples generated by different neural converters for the same original sample. During the training process of the neural converter, the network parameters are optimized by minimizing formula (3).
[0049] (2) Autoencoder feature learning
[0050] Autoencoder is one of the commonly used feature learning models. It aims to learn the most important representation of raw data through neural networks and try to reconstruct the raw data using low-dimensional encoding in feature space. Semi-supervised anomaly detection based on autoencoder is based on the assumption that normal samples can be better reconstructed from feature space than abnormal samples. By training autoencoders and their variants with normal data, it is possible to learn the hidden patterns and regularities in normal data and establish a model suitable for normal data reconstruction, and then detect anomalies based on the reconstruction error of the test data through the model.
[0051] Different neural network structures are usually used for different types of data. LSTM can capture the long-term dependency of the sequence and is suitable for processing time series data. Therefore, for the satellite telemetry sequence, the present invention uses LSTM autoencoder for feature learning. The encoder and decoder of the LSTM autoencoder are two symmetrical LSTM networks. After the original sequence is input, it is mapped to a vector representation of a fixed dimension in the feature space by the encoder, and then the decoder generates a reconstructed sequence.
[0052] During the training of the autoencoder, the reconstruction error (RL) is used as the loss function. The reconstruction error measures the difference between the reconstructed data and the original data. By minimizing the reconstruction error to optimize the network parameters, the autoencoder can be forced to learn and extract the most important features in the data.
[0053] The reconstruction loss is defined as
[0054]
[0055] In the formula, χ train is a training set consisting of normal sequences. During the LSTM autoencoder training process, the network parameters are optimized by minimizing formula (4).
[0056] (3) Neural Contrast Enhanced Autoencoder
[0057] In semi-supervised anomaly detection based on autoencoders, only normal samples are included in the training set. Compared with the large number of potential anomalies that can be represented in any form, the normal samples used as training data are relatively limited, which may cause the trained anomaly detection model to overfit. In addition, since the diversity of normal samples cannot be maintained, the learned representation may lack representativeness for complex time series data, resulting in limited model generalization ability.
[0058] Based on the above analysis, the present invention combines neural converter data enhancement and LSTM autoencoder feature learning, and proposes a telemetry sequence temporal context anomaly detection method (NCAAE) based on neural contrast enhanced autoencoder. NCAAE is an end-to-end full depth anomaly detection method that belongs to self-supervised learning. During the training process, the original samples are data enhanced using neural converters, and samples from different converters in the separated feature space are used as auxiliary tasks for feature learning to help the autoencoder learn more generalized and rich representations, thereby improving the performance of the model.
[0059] Taking into account the deterministic contrast loss of the neural converter and the reconstruction loss of the LSTM autoencoder, a single objective function is proposed to perform end-to-end training and jointly optimize the network parameters. The loss function of NCAAE is defined as L:
[0060] L=L RL +λL DCL (5)
[0061] Where λ∈[0,1] is the balanced deterministic contrast loss L DCL and the reconstruction loss L RL By unifying these two parts of loss, the optimization of network parameters can be jointly guided to promote the neural converter to learn different transformation methods while promoting the autoencoder to extract the most important representation.
[0062] The pseudo code of the training process of NCAAE is shown in Table 2. Among them, the samples in the training set are a set of subsequences obtained by randomly sampling and segmenting multiple normal time series through sliding windows. First, the training samples are input into the neural converter to generate multiple enhanced samples using the original samples, and a set containing the original samples and enhanced samples is obtained. Then, the sample set is input into the LSTM autoencoder, the encoder first maps the samples to the feature space, and then the decoder uses the extracted features to reconstruct the input sequence. Finally, by minimizing formula (5), the neural converter and LSTM autoencoder are jointly trained, and the network parameters are iteratively optimized to obtain the anomaly detection model.
[0063] Table 2 Pseudocode for neural contrast enhancement autoencoder training
[0064]
[0065] In order to verify the performance of the proposed method, two public datasets commonly used in the field of spacecraft anomaly detection were selected for experiments, namely, the Curiosity Mars rover of NASA Mars Science Laboratory (MSL) and the telemetry parameter dataset of the Soil Moisture Active Passive (SMAP). On the SMAP and MSL datasets, the precision, recall and F1* scores of the proposed NCAAE method and nine advanced deep anomaly detection benchmark methods are compared, and the results are shown in Table 3. For each algorithm, the hyperparameters are determined by grid search, and the experimental results recorded in the table are all optimal values.
[0066] Table 3 Performance comparison results of different methods
[0067]
[0068]
[0069] In Table 3, the best F1* score is in bold, and the suboptimal F1* score is underlined. It can be seen that the proposed NCAAE method achieves the best F1* score on both the SMAP and MSL datasets, and its anomaly detection capability is superior to all the benchmark methods. Specifically, on the SMAP dataset, the F1* score of the NCAAE method is 0.955, which is 4.83% higher than the suboptimal F1* score of 0.911 obtained by the LSTM-NDT method; on the MSL dataset, the F1* score of the NCAAE method is 0.962, which is 0.94% higher than the suboptimal F1* score of 0.953 obtained by the DAEMON method, proving the effectiveness of the proposed method.
[0070] For anomaly detection methods, if the precision obtained on the dataset is much higher than the recall rate, it means that the model is more inclined to judge uncertain samples as normal, resulting in a high missed detection rate. On the contrary, if the recall rate is much higher than the precision rate, it means that the model is more inclined to judge uncertain samples as abnormal, resulting in a high false alarm rate. The matrix stacking diagram of the three evaluation indicators of each method on the SMAP and MSL datasets is shown in Figure 2. Figure 2 and Figure 3 Shown
[0071] It can be seen that the NCAAE method proposed in the present invention is significantly better than the benchmark methods. At the same time, the precision and recall rates obtained by the NCAAE method are not much different, indicating that the model can better separate normal samples from abnormal samples, thereby finding a better discrimination threshold dividing line. In addition, the precision and recall rates obtained by the NCAAE method on the SMAP dataset are both greater than 0.93, and the precision and recall rates obtained on the MSL dataset are both greater than 0.96, which strongly proves the robustness and robustness of the NCAAE method.
[0072] Experimental results show that compared with the baseline method, the NCAAE method proposed in this invention uses the LSTM network to extract the temporal features of the data, takes into account the long-term dependency information, and promotes the autoencoder to learn more generalized and rich representations through joint optimization with the neural converter for data enhancement, thereby increasing the distinction between normal and abnormal sequences after model reconstruction, and effectively improving the ability of anomaly detection.
[0073] In order to verify the effectiveness of data augmentation by introducing neural converters in the NCAAE method, this module was removed from the algorithm for ablation experiments. The comparison method only contains the LSTM autoencoder part of the NCAAE method, which is called LSTM-AE. During the training process, LSTM-AE optimizes the network parameters by minimizing the reconstruction error between the input subsequence and the reconstructed subsequence. After the model training is completed, the calculation method of the anomaly score in the test set is the same as that of the NCAAE method. On the SMAP and MSL datasets, the optimal precision, recall, and F1* scores of the NCAAE method and the LSTM-AE method are compared, and the results are shown in Table 4.
[0074] Table 4. The impact of data augmentation on anomaly detection results
[0075]
[0076] As can be seen from Table 4, when there is no significant difference in recall rate, the precision obtained by the NCAAE method is significantly improved compared with the LSTM-AE method without a data enhancement module. On the SMAP dataset, the F1* score obtained by the NCAAE method is 0.74% higher than that of the LSTM-AE method; on the MSL dataset, the F1* score obtained by the NCAAE method is 0.73% higher than that of the LSTM-AE method, proving the effectiveness of introducing neural converters for data enhancement.
[0077] Aiming at the problem that the learned representations are not representative due to the limited normal samples used for training in semi-supervised learning, the present invention introduces neural transformation learning for data enhancement, uses neural networks to learn different transformation methods, and generates enhanced samples with category pseudo-labels, thereby expanding the diversity of training samples. At the same time, the LSTM autoencoder is used to extract the time series features in the telemetry parameters. Through the joint optimization of the training process, the detection capability of the temporal context anomalies of the satellite telemetry parameters is improved.
[0078] The present application may also provide a computer device, comprising: at least one processor, a memory, at least one network interface and a user interface. The various components in the device are coupled together through a bus system. It is understood that the bus system is used to achieve connection and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus and a status signal bus.
[0079] The user interface may include a display, a keyboard or a pointing device, such as a mouse, a trackball, a touch pad or a touch screen.
[0080] It is understood that the memory in the embodiments disclosed in the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used 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 dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). The memories described herein are intended to include, but are not limited to, these and any other suitable types of memories.
[0081] In some embodiments, the memory stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof: an operating system and applications.
[0082] The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application includes various application programs, such as a media player (Media Player), a browser (Browser), etc., which are used to implement various application services. The program for implementing the method of the embodiment of the present disclosure can be included in the application.
[0083] In the above embodiment, the processor may also call a program or instruction stored in the memory, specifically, a program or instruction stored in an application program, and is used to:
[0084] Execute the steps of the above method.
[0085] The above method can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The above-disclosed methods, steps and logic block diagrams can be implemented or executed. The general processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the above-disclosed method can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined to execute. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.
[0086] It is understood that the embodiments described in the present application can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application or a combination thereof.
[0087] For software implementation, the technology of the present application can be implemented by executing the functional modules (such as procedures, functions, etc.) of the present application. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.
[0088] The present application may also provide a non-volatile storage medium for storing a computer program. When the computer program is executed by a processor, each step in the above method embodiment can be implemented.
[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application is described in detail with reference to the embodiments, a person skilled in the art should understand that any modification or equivalent replacement of the technical solution of the present application does not depart from the spirit and scope of the technical solution of the present application and should be included in the scope of the claims of the present application.
Claims
1. A satellite telemetry parameter anomaly detection system based on neural contrast enhanced autoencoder, It is characterized in that The system comprises: A neural converter data enhancement module, including a neural converter, is used to convert the original data through the neural converter, and then enhance the converted data to obtain multiple enhanced samples of the original sample, and output a sample set including the original subsequence and the sequence synthesized by data enhancement; and The autoencoder feature module includes two symmetrical LSTM networks, an encoder and a decoder, which are used to map samples to feature space using the encoder, and then reconstruct the feature representation through the decoder to obtain the anomaly labels at each time point in the sequence to be tested.
2. The satellite telemetry parameter anomaly detection system based on neural contrast enhanced autoencoder according to claim 1, It is characterized in that The neural converter is a recurrent neural network or a Transformer network.
3. The satellite telemetry parameter anomaly detection system based on neural contrast enhanced autoencoder according to claim 1, It is characterized in that The Neural Transformer is trained using a deterministic contrastive loss method.
4. The satellite telemetry parameter anomaly detection system based on neural contrast enhanced autoencoder according to claim 1, It is characterized in that The method for enhancing the converted data includes: Directly output the converted sample as an enhanced sample, multiply the converted sample with the original sample to obtain the enhanced sample output, and add the converted sample to the original sample to obtain the enhanced sample output.
5. The satellite telemetry parameter anomaly detection system based on neural contrast enhanced autoencoder according to claim 1, It is characterized in that The training of the autoencoder feature module uses reconstruction error as the loss function.
6. The satellite telemetry parameter anomaly detection system based on neural contrast enhanced autoencoder according to claim 1, It is characterized in that The training of the system adopts an end-to-end approach, and the training process includes: Input the historical normal telemetry sequence, perform preprocessing and sliding window segmentation, and obtain a subsequence set for training; Inputting the subsequence set into the neural transformer data enhancement module to generate multiple enhanced samples of the original samples, thereby obtaining a sample set including the original subsequence and the sequence synthesized by data enhancement; The sample set is input into the autoencoder feature module, the encoder maps the sample to the feature space, and then the feature representation is reconstructed by the decoder for feature learning; Contrastive loss is used to distinguish samples from different neural converters in the feature space, and reconstruction loss is used to measure the feature representation ability. By jointly optimizing the contrastive loss and reconstruction loss, the network parameters of the neural converter data enhancement module and the autoencoder feature module are iteratively updated.
7. A method for detecting satellite telemetry parameter anomalies based on a neural contrast enhanced autoencoder, implemented based on the system described in any one of claims 1 to 6, wherein the method include: The satellite telemetry parameters are preprocessed and segmented and then input into the trained satellite telemetry parameter anomaly detection system based on neural contrast enhancement autoencoder to obtain the anomaly labels of the satellite telemetry parameters at each time point.
Citation Information
Patent Citations
A satellite anomaly detection method of an adversarial network autoencoder
CN109948117A
Comparative generative adversarial network-based multivariate time sequence anomaly detection method
CN117493930A
Differential time sequence reconstruction spacecraft anomaly detection method based on auto-encoder and model construction method
CN117574219A
