KR102241872B1 - Method and apparatus for detecting unusual behavior in satellite - Google Patents

Abstract

Various embodiments of the present invention provide a method for detecting an unusual behavior of a satellite. The method for detecting an unusual behavior of a satellite comprises the following steps: receiving satellite telemetry data, which is performed by a computing device; determining p related parameters having relevance with each other among the parameters of the satellite telemetry data; generating m telemetry data subject to be analyzed from the satellite telemetry data, based on the p related parameters, wherein each of the m telemetry data to be analyzed has n data values for each of the p related parameters; performing principal component analysis for each of the m telemetry data to be analyzed, thereby generating m sets of first principal component coefficients and m sets of second principal component coefficients, wherein each of the m sets of the first principal component coefficients is composed of p first principal component coefficients and each of the m sets of the second principal component coefficients is composed of p second principal component coefficients; performing clustering of the m telemetry data to be analyzed, based on the m sets of the first principal component coefficients and the m sets of the second principal component coefficients, thereby generating a plurality of clusters; and determining unusual telemetry data, based on the m telemetry data to be analyzed, which are included in the clusters.

Description

Method and apparatus for detecting unusual behavior in satellite

The present invention relates to a method and apparatus for detecting a singular motion of a satellite, and more particularly, a method for detecting a singular motion using a principal component analysis of satellite telemetry data received from a satellite, a computer program, And to an apparatus.

Telemetry is a technology that performs various observations at a point separated from the object to be observed and acquires the data. When there is a physical, economic, or safety problem because it resides at the observation point, or if the object to be observed is moving. It is mainly used. For example, it is used to determine the state of an orbiting satellite, such as a low-orbit satellite.

Since LEO satellites have limitations in the time and number of times they can communicate with a ground station, satellite telemetry data while not communicating with a ground station is stored in a large-capacity memory. Satellite telemetry data accumulated over a long period of time is transmitted to the ground station during the time available for communication with the satellite and the ground station. Satellite telemetry data is investigated by personnel with expertise to identify trends.

Since there are so many types of data included in satellite telemetry data, personnel with various expertise must be put in to conduct such an investigation, and a high level of expertise is required to analyze trends. It takes a lot of time to analyze satellite telemetry data accumulated over a long period of time. In the case of a low-orbit satellite, since the time to communicate with the ground station is limited, a prompt response is required.

Conventionally, a threshold value is designated for each telemetry data, and an alarm is generated for the corresponding telemetry data when the threshold value is exceeded. However, since satellites perform various tasks in various environments, there are numerous operating states. However, when one threshold value is used, it is difficult to optimize the threshold value for each of the various operating states, and only the worst case is assumed to be assigned as the threshold value. In addition, an abnormal operation may have occurred even if the actual value does not exceed the threshold value.

In addition, since individual threshold values or rules are defined with expertise for each telemetry data, it cannot be extended to other satellites or systems, and threshold values or rules must be determined individually for each satellite.

An object of the present invention is to provide a method, a computer program, and an apparatus for detecting an unusual motion of a satellite by using principal component analysis of satellite telemetry data received from a satellite.

Another problem to be solved by the present invention is a method, a computer program, and an apparatus for determining a fuzzy rule using principal component analysis of satellite telemetry data received from a satellite, and detecting a singular motion of a satellite using the fuzzy rule. To provide.

Another problem to be solved by the present invention is to provide a method, a computer program, and an apparatus for detecting detailed singular motion of a satellite that determines an abnormal operation section and an abnormal operation parameter by using principal component analysis of telemetry data to be precisely analyzed. It is to do.

As a technical means for achieving the above-described technical problems, a method for detecting a singular motion of a satellite performed by a computing device according to an aspect of the present invention includes: receiving satellite telemetry data; Determining p related parameters that are related to each other from among the parameters of the satellite telemetry data; Generating m telemetry data to be analyzed from the satellite telemetry data based on the p related parameters, each of the m telemetry data to be analyzed, each of the p related parameters Having n data values for n; Performing principal component analysis on each of the m telemetry data to be analyzed to generate m first principal component coefficient sets and m second principal component coefficient sets, wherein each of the m first principal component coefficient sets consisting of p first principal component coefficients, each of the m second principal component coefficient sets consisting of p second principal component coefficients; Generating a plurality of clusters by clustering the m telemetry data to be analyzed based on the m first principal component coefficient sets and the m second principal component coefficient sets; And determining specific telemetry data based on the m telemetry data to be analyzed belonging to each of the plurality of clusters.

The p related parameters may be determined based on a keyword search result for the names of the parameters of the satellite telemetry data.

The p related parameters may be previously selected by a satellite expert.

The generating of the m pieces of telemetry data to be analyzed includes generating m pieces of telemetry data corresponding to the p related parameters among the m pieces of satellite telemetry data, wherein the m pieces of telemetry data Each of the related telemetry data having n original data values for each of the p related parameters; And pre-processing the m related telemetry data.

The preprocessing of the m related telemetry data may include normalizing the n original data values of each of the m related telemetry data for each of the p related parameters.

The preprocessing of the m related telemetry data includes filtering the n raw data values for each of the p related parameters of each of the m related telemetry data using a moving average filter. Can include.

The number of clusters may be k determined through gap statistics for the m first principal component coefficient sets and the m second principal component coefficient sets.

The generating of the plurality of clusters includes generating the k clusters by applying a k-means clustering method to the m first principal component coefficient sets and the m second principal component coefficient sets. It may include.

When only one analysis target telemetry data belongs to any one of the plurality of clusters, the one analysis target telemetry data may be determined as the specific telemetry data.

Generating the plurality of clusters may include determining properties of each of the plurality of clusters. The attributes of each of the plurality of clusters are i) the number of telemetry data to be analyzed belonging to each cluster, ii) the center position of each cluster, iii) the analysis target belonging to each cluster at the center position of each cluster The telemetry data may include a maximum distance among distances to each location, and iv) a margin distance of each cluster.

The method of detecting the singular motion of the satellite includes: receiving new satellite telemetry data; Generating new telemetry data to be analyzed from the new satellite telemetry data based on the p related parameters; Performing principal component analysis on the newly analyzed telemetry data to generate a third principal component coefficient set and a new fourth principal component coefficient set; Calculating a location of the new analysis target telemetry data based on the third principal component coefficient set and the fourth principal component coefficient set; And determining a closest cluster from among the plurality of clusters to the location of the newly analyzed telemetry data.

The method for detecting the singular motion of the satellite is when the distance from the center position of the nearest cluster to the position of the telemetry data to be analyzed is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the The method may further include determining telemetry data to be analyzed as the specific telemetry data.

When the distance from the center position of the nearest cluster to the position of the telemetry data to be analyzed is less than or equal to the sum of the maximum distance and the margin distance of the nearest cluster, the method for detecting the singular motion of the satellite is It may further include the step of updating the attribute of the contact cluster.

The margin distance of each cluster may be determined as a value obtained by dividing the maximum distance of each cluster by the number of telemetry data to be analyzed belonging to each cluster.

In another aspect of the present invention, a method for detecting singular motion of a satellite performed by a computing device is the step of generating a plurality of clusters each having an attribute, wherein the attribute of each cluster is telemetry data to be analyzed belonging to the cluster Of the number of, the central position of each cluster, the maximum distance among distances from the central position of each cluster to the positions of each of the analysis target telemetry data belonging to each cluster, and a margin distance of each cluster Comprising a; Receiving new satellite telemetry data; generating new telemetry data to be analyzed from the new satellite telemetry data based on p related parameters; Performing principal component analysis on the newly analyzed telemetry data to generate a first principal component coefficient set and a second principal component coefficient set; Calculating a position of the new analysis target telemetry data based on the first principal component coefficient set and the second principal component coefficient set; And determining a closest cluster from among the plurality of clusters to a location of the newly analyzed telemetry data.

The method for detecting the singular motion of the satellite is when the distance from the center position of the nearest cluster to the position of the telemetry data to be analyzed is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the The method may further include determining telemetry data to be analyzed as the specific telemetry data.

When the distance from the center position of the nearest cluster to the position of the telemetry data to be analyzed is less than or equal to the sum of the maximum distance and the margin distance of the nearest cluster, the method for detecting the singular motion of the satellite is It may further include the step of updating the attribute of the contact cluster.

The margin distance of each cluster may be determined as a value obtained by dividing the maximum distance of each cluster by the number of telemetry data to be analyzed belonging to each cluster.

According to another aspect of the present invention, a computer program stored in a medium is provided in order to execute a method of detecting a singular motion of a satellite by using a computing device.

According to another aspect of the present invention, an apparatus for detecting an unusual motion of a satellite includes a memory for storing satellite telemetry data; And determining p related parameters that are related to each other among the parameters of the satellite telemetry data, and based on the p related parameters, from the satellite telemetry data, each of the p related parameters M pieces of telemetry data to be analyzed having n data values are generated, and principal component analysis is performed on each of the m pieces of telemetry data to be analyzed. A first principal component coefficient set, and m second principal component coefficient sets, each consisting of p second principal component coefficients, are generated, and based on the m first principal component coefficient sets and the m second principal component coefficient sets, the m At least one, configured to generate a plurality of clusters by clustering the telemetry data to be analyzed, and to determine specific telemetry data based on the m telemetry data to be analyzed belonging to each of the plurality of clusters Including the processor of.

In another aspect of the present invention, a method for detecting singular motion of a satellite performed by a computing device is a step of receiving c telemetry data to be analyzed, wherein each of the c telemetry data to be analyzed is p related. Having n data values for each of the parameters; Performing principal component analysis on each of the c analysis target telemetry data to generate c first principal component coefficient sets each consisting of p first principal component coefficients; By multiplying each of the c telemetry data to be analyzed by the p first principal component coefficients of the corresponding first principal component coefficient set among the c first principal component coefficient sets, each of the c first principal component coefficients consisting of n first values. 1 calculating score data; Determining whether each of the c first score data has periodicity; Determining first and second periodic characteristics when each of the c first score data has periodicity; Calculating first and second periodic property values corresponding to first and second periodic properties for each of the c first score data; Determining a fuzzy rule based on the first and second periodic characteristic values of the c first score data; And determining specific telemetry data using the fuzzy rule.

The method of detecting the singular motion of the satellite includes receiving satellite telemetry data; Determining the p related parameters; Generating m telemetry data to be analyzed from the satellite telemetry data based on the p related parameters, each of the m telemetry data to be analyzed, each of the p related parameters Having n data values for n; Performing principal component analysis on each of the m telemetry data to be analyzed to generate m first principal component coefficient sets and m second principal component coefficient sets, wherein each of the m first principal component coefficient sets consisting of p first principal component coefficients, each of the m second principal component coefficient sets consisting of p second principal component coefficients; And generating a plurality of clusters by clustering the m telemetry data to be analyzed based on the m first principal component coefficient sets and the m second principal component coefficient sets. The c analysis target telemetry data may be analysis target telemetry data belonging to one of the plurality of clusters among the m analysis target telemetry data.

The generating of the m pieces of telemetry data to be analyzed includes generating m pieces of telemetry data corresponding to the p related parameters among the m pieces of satellite telemetry data, wherein the m pieces of telemetry data Each of the related telemetry data having n original data values for each of the p related parameters; And pre-processing the m related telemetry data.

The preprocessing of the m related telemetry data may include normalizing the n original data values of each of the m related telemetry data for each of the p related parameters.

The preprocessing of the m related telemetry data includes filtering the n raw data values for each of the p related parameters of each of the m related telemetry data using a moving average filter. Can include.

Determining whether or not the periodicity has the periodicity may include calculating at least one highest point and at least one lowest point separated by a set interval or more from the n first values; And a distance between the at least one highest point, a distance between the at least one lowest point, and a number of the n first values based on the number of the at least one highest point and the number of the at least one lowest point. 1 It may comprise the step of determining whether the score data has periodicity.

The determining of whether or not the periodicity has a periodicity further comprises determining that the first score data consisting of the n first values has periodicity when the at least one highest point and the at least one lowest point alternately appear. Can include.

The determining of the first and second periodic characteristics includes, for each of the c first score data, i) a distance from the lowest point to the next highest point, ii) a distance from the highest point to the next lowest point, iii) The minimum value, which is the value of the lowest point, iv) the maximum value, which is the value of the highest point, v) the difference between the value of the lowest point and the value of the next highest value, vi) the difference between the value of the highest point and the value of the next lowest value, vii) the lowest point Calculating at least two characteristics of a slope from the highest point to the next highest point, and viii) a slope from the highest point to the next lowest point; Determining a characteristic having the smallest variance for the at least two characteristics calculated for each of the c first score data as the first periodic characteristic; And determining a characteristic having the lowest linear dependence on the first periodic characteristic among the at least two characteristics as the second periodic characteristic.

The determining of the fuzzy rule may include determining first input intervals for the first periodic characteristic values; Determining second input intervals for the second periodic characteristic values; Determining a fuzzy output period; The first input sections and the second input section based on the number of the first periodic feature values belonging to each of the first input sections and the number of the second periodic feature values belonging to each of the second input sections Calculating a distribution value of each of the input regions defined by a combination of them; And determining the fuzzy rule based on the fuzzy output section and the distribution values.

The method of detecting the singular motion of the satellite includes: receiving new satellite telemetry data; Generating new telemetry data to be analyzed from the new satellite telemetry data based on the p related parameters; Calculating a fuzzy output value according to a center of sums method by applying new analysis target telemetry data to the fuzzy rule; And determining whether the new satellite telemetry data is singular telemetry data based on the fuzzy output value.

In the method of detecting the singular motion of the satellite, when each of the c first score data does not have periodicity, a principal component analysis is performed on each of the c analysis target telemetry data, and p second principal components respectively. Generating c second principal component coefficient sets consisting of coefficients; By multiplying each of the c telemetry data to be analyzed by p second principal component coefficients of a corresponding second principal component coefficient set among the c second principal component coefficient sets, each of the c second principal component coefficients consisting of n second values. 2 calculating score data; Determining whether each of the c second score data has periodicity; Determining third and fourth periodic characteristics when each of the c second score data has periodicity; Calculating third and fourth periodic feature values corresponding to third and fourth periodic features for each of the c second score data; And determining a fuzzy rule based on the third and fourth periodic characteristic values of the c second score data.

The method of detecting the singular motion of the satellite includes, when each of the c first score data does not have periodicity, for each of the c first score data, for n first values of each first score data. Calculating a plurality of statistical values; Determining a position of each of the c first score data in a multidimensional space based on the plurality of statistical values calculated for each of the c first score data; Calculating a center position based on the distribution of the c first score data in the multidimensional space; Calculating a center separation distance of each of the c first score data based on a difference between the center position and the positions of the c first score data; And determining the singular telemetry data based on the center separation distances.

The plurality of statistical values for the n first values of each of the first score data may include at least two of a minimum value, a maximum value, a median value, and a standard deviation of the n first values of each first score data. have.

The determining of the singular telemetry data based on the center separation distances may include calculating an average separation distance and a standard deviation of the center separation distances; Determining a lower limit and an upper limit based on the average separation distance and a preset multiple of the standard deviation; And when the center separation distance of any one of the first score data among the c first score data is out of the normal range defined by the lower limit and the upper limit, the analysis target telemetry corresponding to the one center separation distance. And determining the tree data as the singular telemetry data.

According to another aspect of the present invention, a computer program stored in a medium is provided in order to execute a method of detecting a singular motion of a satellite by using a computing device.

An apparatus for detecting a singular motion of a satellite according to another aspect of the present invention includes a memory for storing c telemetry data to be analyzed having n data values for each of p related parameters; And performing principal component analysis on each of the c analysis target telemetry data to generate c first principal component coefficient sets each consisting of p first principal component coefficients, and the c analysis target telemetry data Each of the c first principal component coefficient sets is multiplied by the p first principal component coefficients of the corresponding first principal component coefficient set to calculate c first score data consisting of n first values, and the c It is determined whether or not each of the first score data has periodicity, and when each of the c first score data has periodicity, first and second periodic characteristics are determined, and the c first score data First and second periodic property values corresponding to the first and second periodic properties are calculated for each of the first and second periodic properties, and a fuzzy rule is calculated based on the first and second periodic property values of the c first score data. And at least one processor configured to determine and determine outlier telemetry data using the fuzzy rule.

A method for detecting detailed singular motion of a satellite performed by a computing device according to another aspect of the present invention includes the steps of: receiving telemetry data to be precise analysis having n data values for each of p related parameters; Generating d telemetry data for detailed analysis by dividing the telemetry data for precision analysis into a preset time unit; Performing principal component analysis on each of the d detailed analysis target telemetry data to generate d principal component coefficient sets each consisting of p principal component coefficients; And determining an abnormal operation interval and a normal operation interval based on the p principal component coefficients of each of the d principal component coefficient sets.

The determining of the abnormal operation period and the normal operation period may include mapping the d principal component coefficient sets to d positions in a p-dimensional space, respectively; Calculating center points of the d positions; Calculating d center separation distances between the center point and the d locations, respectively; Calculating an average separation distance and a standard deviation of the d center separation distances; Determining an upper limit based on the average separation distance and a preset multiple of the standard deviation; And determining the abnormal operation section and the normal operation section based on a result of comparing each of the d center separation distances and the upper limit.

The step of determining the abnormal operation section and the normal operation section based on the comparison result includes the detailed analysis target telemetry data corresponding to a center separation distance greater than the upper limit among the d center separation distances. Determining that it corresponds to the section; And determining that the detailed analysis target telemetry data corresponding to a center separation distance smaller than the upper limit among the d center separation distances corresponds to the normal operation period.

The detailed singular motion detection method of the satellite comprises: determining first principal component coefficients based on the p principal component coefficients of a principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the normal operating period. ; Determining second principal component coefficients based on the p principal component coefficients of a principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the abnormal operation period; And determining an abnormal operation parameter based on a comparison result between the first principal component coefficients and the second principal component coefficients.

The determining of the first principal component coefficients includes calculating a coefficient average of the absolute values of the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the normal operation period. ; And principal component coefficients having a magnitude greater than the coefficient average among the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the normal operation period as the first principal component coefficients. It may include the step of determining.

The determining of the second principal component coefficients includes calculating a coefficient average of the absolute values of the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the abnormal operation section. ; And principal component coefficients having a magnitude greater than the coefficient average among the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the abnormal operation period as the second principal component coefficients. It may include the step of determining.

The determining of the abnormal operation parameter may include: determining first principal parameters corresponding to the first principal component coefficients; Determining second principal parameters corresponding to the second principal component coefficients; And determining a parameter corresponding to the second main parameters but not the first main parameters as the abnormal operation parameter.

The determining of the abnormal operation parameter may further include determining a parameter corresponding to the first main parameters but not the second main parameters as the abnormal operation parameter.

The detailed singular motion detection method of the satellite includes: receiving satellite telemetry data; Determining the p related parameters; Generating m telemetry data to be analyzed from the satellite telemetry data based on the p related parameters, each of the m telemetry data to be analyzed, each of the p related parameters Having n data values for n; Performing principal component analysis on each of the m telemetry data to be analyzed to generate m first principal component coefficient sets and m second principal component coefficient sets, wherein each of the m first principal component coefficient sets consisting of p first principal component coefficients, each of the m second principal component coefficient sets consisting of p second principal component coefficients; Generating a plurality of clusters by clustering the m telemetry data to be analyzed based on the m first principal component coefficient sets and the m second principal component coefficient sets; And determining the precise analysis target telemetry data based on the m analysis target telemetry data belonging to each of the plurality of clusters.

When only one analysis target telemetry data belongs to any one of the plurality of clusters, the one analysis target telemetry data may be determined as the precise analysis target telemetry data.

Generating the plurality of clusters may include determining properties of each of the plurality of clusters. The attributes of each of the plurality of clusters are i) the number of telemetry data to be analyzed belonging to each cluster, ii) the center position of each cluster, iii) the analysis target belonging to each cluster at the center position of each cluster The telemetry data may include a maximum distance among distances to each location, and iv) a margin distance of each cluster.

The detailed singular motion detection method of the satellite includes: receiving new satellite telemetry data; Generating new telemetry data to be analyzed from the new satellite telemetry data based on the p related parameters; Performing principal component analysis on the newly analyzed telemetry data to generate a third principal component coefficient set and a new fourth principal component coefficient set; Calculating a location of the new analysis target telemetry data based on the third principal component coefficient set and the fourth principal component coefficient set; Determining a closest cluster from among the plurality of clusters to a location of the newly analyzed telemetry data; And when the distance from the center position of the nearest cluster to the position of the new analysis target telemetry data is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the new analysis target telemetry data. It may further include determining as the telemetry data to be precisely analyzed.

The detailed singular motion detection method of the satellite is the step of receiving c analysis target telemetry data belonging to one of the plurality of clusters among the m analysis target telemetry data, wherein the c analysis targets Each of the telemetry data having n data values for each of the p related parameters; Performing principal component analysis on each of the c analysis target telemetry data to generate c first principal component coefficient sets each consisting of p first principal component coefficients; By multiplying each of the c telemetry data to be analyzed by the p first principal component coefficients of the corresponding first principal component coefficient set among the c first principal component coefficient sets, each of the c first principal component coefficients consisting of n first values. 1 calculating score data; Determining whether each of the c first score data has periodicity; Determining first and second periodic characteristics when each of the c first score data has periodicity; Calculating first and second periodic property values corresponding to first and second periodic properties for each of the c first score data; Determining a fuzzy rule based on the first and second periodic characteristic values of the c first score data; And determining the telemetry data to be precise analysis using the fuzzy rule.

According to another aspect of the present invention, there is provided a computer program stored in a medium in order to execute a method of detecting detailed singular motion of a satellite using a computing device.

According to another aspect of the present invention, an apparatus for detecting detailed singular motion of a satellite includes: a memory for storing telemetry data to be precisely analyzed having n data values for each of p related parameters; And dividing the telemetry data for precision analysis into a preset time unit to generate d telemetry data for detailed analysis, and performing principal component analysis on each of the d telemetry data for detailed analysis, At least one, configured to generate d principal component coefficient sets each consisting of p principal component coefficients, and determine an abnormal operation interval and a normal operation interval based on the p principal component coefficients of each of the d principal component coefficient sets. Includes a processor.

According to an embodiment of the present invention, a specific motion of a satellite may be detected using principal component analysis of satellite telemetry data received from a satellite. Several clusters can be built through principal component analysis to represent the various operating states of the satellite. The specific motion data can be classified by using the statistical characteristics of the score data corresponding to the telemetry data to be analyzed in each cluster.

According to an embodiment of the present invention, a large amount of data can be processed in a short time by reducing the data dimension through principal component analysis. By using statistical values that change adaptively according to various operating states, not simply preset threshold values, the Upson's unusual behavior, which can be confirmed, can be detected by using the threshold value. By building clusters for various operating states, it is possible to detect unusual motions in consideration of various operating states and characteristics of the satellite. According to an embodiment of the present invention, if several experts with specialized knowledge are not required, it can be extended and applied to various systems.

According to another embodiment of the present invention, a fuzzy rule may be determined using principal component analysis of satellite telemetry data received from a satellite, and a singular motion of a satellite may be detected using the fuzzy rule. Since satellites orbit around the earth according to a certain orbital period, satellite telemetry data can be automatically analyzed and abnormal motions of satellites can be automatically detected by using the characteristic that satellite telemetry data often exhibits periodic characteristics. I can.

According to still another embodiment of the present invention, by determining an abnormal operation section and an abnormal operation parameter using principal component analysis of telemetry data to be precisely analyzed, detailed singular motion of a satellite may be detected. According to embodiments of the present invention, singular telemetry data corresponding to the singular operation may be determined. The peculiar telemetry data is telemetry data to be analyzed for precision, and telemetry data to be analyzed may be generated by dividing into meaningful intervals (eg, 90 minutes or an orbiting period of a satellite). By performing principal component analysis on them, it is possible to determine how much each of the parameters contributed to generating common information. By grasping which parameter data are mainly used in each detailed analysis target telemetry data, an abnormal data section and an abnormal parameter can be detected.

According to another embodiment of the present invention, large data can be processed in a short time by reducing the data dimension through principal component analysis, and can be used without the help of several experts with expertise, and can be applied to various systems. can do.

1 schematically shows a satellite system according to the invention.
2 schematically shows an internal configuration of a computing device according to the present invention.
3 is a flowchart illustrating a method of detecting a singular motion of a satellite according to an exemplary embodiment.
4 is an exemplary flow chart for explaining in more detail step (S130) of a method for detecting a singular motion of a satellite according to an embodiment of the present invention.
5 is a flowchart illustrating a method of detecting a singular motion of a satellite according to another embodiment.
6 is a flowchart illustrating a method of detecting a singular motion of a satellite according to another embodiment.
7 is an exemplary flow chart for explaining in more detail step (S340) of the method for detecting a singular motion of a satellite according to another embodiment of the present invention.
8 is an exemplary flow chart for explaining in more detail step S380 of a method for detecting a singular motion of a satellite according to another embodiment of the present invention.
9 is an exemplary flowchart illustrating an operation when first score data do not have periodicity in step S340 of a method of detecting a singular motion of a satellite according to another embodiment of the present invention.
10 is an exemplary flow chart for explaining an operation according to another example of a case in which first score data do not have periodicity in step S340 of a method of detecting a singular motion of a satellite according to another embodiment of the present invention.
11 is a flowchart illustrating a method of detecting a detailed singular motion of a satellite according to another embodiment.

Hereinafter, various embodiments will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present disclosure. However, the technical idea of the present disclosure is not limited to the embodiments described in the present specification because it may be modified and implemented in various forms. In describing the embodiments disclosed in the present specification, when it is determined that a detailed description of a related known technology may obscure the gist of the technical idea of the present disclosure, a detailed description of the known technology will be omitted. The same or similar components are given the same reference numerals, and redundant descriptions thereof will be omitted.

In the present specification, when an element is described as being "connected" with another element, this includes not only the case of being "directly connected" but also the case of being "indirectly connected" with another element interposed therebetween. When a certain element "includes" another element, it means that another element may be included in addition to the other element, not excluded, unless specifically stated to the contrary.

Some embodiments may be described with functional block configurations and various processing steps. Some or all of these functional blocks may be implemented with various numbers of hardware and/or software components that perform a specific function. For example, the functional blocks of the present disclosure may be implemented by one or more microprocessors, or may be implemented by circuit configurations for a predetermined function. Functional blocks of the present disclosure may be implemented in various programming or scripting languages. The functional blocks of the present disclosure may be implemented as an algorithm executed on one or more processors. Functions performed by a function block of the present disclosure may be performed by a plurality of function blocks, or functions performed by a plurality of function blocks in the present disclosure may be performed by a single function block. In addition, the present disclosure may employ conventional techniques for electronic environment setting, signal processing, and/or data processing.

1 schematically shows a satellite system according to the invention.

Referring to FIG. 1, the satellite 10 may fly around the earth according to a preset orbit. After the satellite 10 is launched, it is extremely difficult for a person to directly check the condition or repair a malfunction. Accordingly, the satellite 10 periodically monitors its own condition and transmits it to the earth's ground station in the form of satellite telemetry data (STD). Experts at ground stations can analyze satellite telemetry data (STD) to determine the state of the satellite 10 and take necessary actions.

In the case where the satellite 10 is a stationary satellite capable of always communicating with a ground station, since satellite telemetry data (STD) is received in real time, the state of the satellite 10 can be immediately grasped. However, when the satellite 10 is a low-orbit satellite, the satellite 10 orbits the earth once in approximately 90 minutes, and the time to communicate with the ground station is less than 10 minutes. Further, depending on the orbit of the satellite 10, the ground station may not be able to communicate with the satellite 10 every 90 minutes, but may communicate with the satellite 10 approximately every 12 hours.

The satellite 10 stores data that monitored each of the various parameters of the satellite 10 in a large-capacity memory, and transmits the data stored in the large-capacity memory as satellite telemetry data (STD) to the ground station when it can communicate with the ground station. I can. In this specification, satellite telemetry data (STD) refers to a large amount of original data transmitted when the satellite 10 is capable of communicating with a ground station. Satellite telemetry data (STD) includes periodically measured data values for all kinds of parameters.

Satellite telemetry data (STD) may include all information on the satellite 10. For example, satellite telemetry data (STD) is the position of the satellite 10, the speed of movement, the pause (direction), the operating state of the processor, the processor temperature, the type of mission being performed, the battery charging and discharging state, the battery It may include data values for all kinds of parameters such as charging/discharging current, battery voltage, battery temperature, solar cell status, and the like.

In this specification, a parameter refers to an item included in satellite telemetry data (STD) and periodically monitored. The data value means the value measured or monitored for that parameter. The data value may be an analog value, for example. However, the data values of some parameters may be digital values or digital codes.

The data collection period may be different for each parameter. For example, for important parameters, data values may be collected every 1 second, for less important parameters, data values may be collected every 8 seconds. The satellite telemetry data STD may include not only data values collected every second, but also data values collected every 2 seconds, and data values collected every 4 and 8 seconds.

In the present specification, as an example, it is assumed that the satellite 10 transmits satellite telemetry data (STD) to a ground station approximately every 12 hours. However, 12 hours is only exemplary, and satellite telemetry data (STD) may be transmitted in a shorter or longer time period. In this specification, it is assumed that the satellite telemetry data (STD) includes data values corresponding to approximately 12 hours.

The computing device 100 receives satellite telemetry data (STD) from the satellite 10, a method of detecting a singular motion of a satellite according to an embodiment, a method of detecting a singular motion of a satellite according to another embodiment, and another method. An apparatus that performs at least one of the detailed singular motion detection methods of a satellite according to an embodiment. The computing device 100 may be configured with a plurality of computing devices. The computing device 100 may also be referred to as a device for detecting a specific motion of a satellite or a device for detecting a detailed specific motion of a satellite.

The computing device 100 may be installed in a ground station that communicates with the satellite 10 or may be connected to the ground station through communication so as to receive satellite telemetry data (STD) through the ground station.

2 schematically shows an internal configuration of a computing device according to the present invention.

Referring to FIG. 2, the computing device 100 may include a control unit 110 and a memory 120. The controller 110 and the memory 120 may exchange data with each other through a bus. The computing device 100 may further include other components, such as a communication module, in addition to the components illustrated in FIG. 2.

The controller 110 typically controls the overall operation of the computing device 100. The controller 110 may perform basic arithmetic, logic, and input/output operations, and may execute, for example, program code stored in the memory 120.

The control unit 110 may include at least one processor. The control unit 110 may include a plurality of processors. The control unit 110 may be referred to as a processor.

The memory 120 is a recording medium that can be read by the processor 110 and may include a permanent mass storage device such as RAM, ROM, and a disk drive. The memory 120 may store an operating system and at least one program or application code. Satellite telemetry data (STD) transmitted by the satellite 10 may be stored in the memory 120. A plurality of satellite telemetry data STDs may be stored in the memory 120. Satellite telemetry data (STD) stored in the memory 120 may be collected for days, weeks, or months.

The computing device 100 may receive new satellite telemetry data (STD) transmitted by the satellite 10 and store it in the memory 120.

According to an embodiment, the controller 110 may determine p related parameters that are related to each other among all parameters of the satellite telemetry data STD. The controller 110 may generate m telemetry data to be analyzed from the satellite telemetry data STD based on the p related parameters. Each of the telemetry data to be analyzed may have n data values for each of the p related parameters. The controller 110 may perform principal component analysis on each of the m analysis target telemetry data to generate m first principal component coefficient sets and m second principal component coefficient sets. Each of the m first principal component coefficient sets may consist of p first principal component coefficients, and each of m second principal component coefficient sets may consist of p second principal component coefficients. The controller 110 may generate a plurality of clusters by clustering m telemetry data to be analyzed based on m first principal component coefficient sets and m second principal component coefficient sets. The controller 110 may determine the specific telemetry data based on m pieces of telemetry data to be analyzed belonging to each of the plurality of clusters.

According to another embodiment, the controller 110 may create a plurality of clusters. Each of the plurality of clusters is selected from among the number of telemetry data to be analyzed belonging to each cluster, the center position of each cluster, and distances from the center position of each cluster to the position of each of the telemetry data to be analyzed belonging to each cluster. It may have properties such as a maximum distance and a margin distance of each cluster. The controller 110 may receive new satellite telemetry data. The controller 110 may generate new analysis target telemetry data from new satellite telemetry data based on p related parameters. The controller 110 may generate a first principal component coefficient set and a second principal component coefficient set by performing principal component analysis on the newly analyzed telemetry data. The controller 110 may calculate the position of the new analysis target telemetry data based on the first principal component coefficient set and the second principal component coefficient set. The control unit 110 may determine a closest cluster from among the plurality of clusters to the location of the telemetry data to be analyzed. When the distance from the center position of the nearest cluster to the position of the new analysis target telemetry data is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the controller 110 selects the new analysis target telemetry data. It can be determined by telemetry data.

According to another embodiment, the controller 110 may generate c first principal component coefficient sets by performing principal component analysis on each of the c analysis target telemetry data. Each of the c analysis target telemetry data has n data values for each of the p related parameters, and each of the c first principal component coefficient sets may consist of p first principal component coefficients. The control unit 110 multiplies each of the c analysis target telemetry data by the p first principal component coefficients of the corresponding first principal component coefficient set among the c first principal component coefficient sets, thereby each consisting of n first values. The c first score data may be calculated. The controller 110 may determine whether each of the c first score data has periodicity. When each of the c first score data has periodicity, the control unit 110 determines first and second periodic characteristics, and corresponds to the first and second periodic characteristics for each of the c first score data First and second periodic characteristic values to be calculated may be calculated. The controller 110 may determine a fuzzy rule based on the first and second periodic characteristic values of the c first score data. The controller 110 may determine the specific telemetry data using the fuzzy rule.

According to another embodiment, the control unit 110 divides telemetry data to be precise analysis having n data values for each of the p related parameters into a preset time unit to provide d telemetry data to be analyzed. Can be created. The controller 110 may perform principal component analysis on each of the d detailed analysis target telemetry data to generate d principal component coefficient sets each consisting of p principal component coefficients. The controller 110 may determine an abnormal operation period and a normal operation period based on the p main component coefficients of each of the d main component coefficient sets.

The operation of the control unit 110 according to various embodiments will be described in more detail below.

3 is a flowchart illustrating a method of detecting a singular motion of a satellite according to an exemplary embodiment.

Referring to FIG. 3, the controller 110 may receive satellite telemetry data (S110). Satellite telemetry data is data stored in a large-capacity memory of the satellite 10 and may be received by the computing device 100 through a ground station. The satellite telemetry data includes raw data values periodically monitored by the satellite 10, and the raw data values may be monitored values for each of all parameters of the satellite 10.

For example, the satellite telemetry data may include first raw data values corresponding to the first parameter and second raw data values corresponding to the second parameter. The first raw data values may be values collected every first period (eg, 1 second), and the second raw data values may be values collected every second period (eg, 8 seconds). Satellite telemetry data may be approximately 12 hours of data.

The controller 110 may receive satellite telemetry data collected in different periods. For example, the satellite telemetry data may include first satellite telemetry data for a first period and second satellite telemetry data for a second period.

The controller 110 may determine p related parameters that are related to each other among all parameters of the satellite telemetry data (S120). The p related parameters can be selected by the user. p may be a natural number of 2 or more. p denotes the number of related parameters determined in step S120.

According to an embodiment, the p related parameters may be determined based on a keyword search result for the names of all parameters of satellite telemetry data. For example, the user may select p related parameters by searching for a keyword selected by the user for the names of all parameters. For example, keywords may be battery, temperature, voltage, and current. For example, when the user inputs'battery' as a search keyword, p related parameters may include accumulated battery charge amount, accumulated battery discharge amount, battery charging current, battery discharge current, battery temperature, battery voltage, and the like. .

According to another embodiment, p related parameters may be preselected by a satellite expert. Satellite experts can pre-select related parameters that are related to each other. Since at least most of the parameters are common to all satellite systems, related parameters previously selected by a satellite expert can be equally applied to other satellite systems.

The controller 110 may generate m telemetry data to be analyzed from satellite telemetry data based on the p related parameters (S130). m may be a natural number of 2 or more. m denotes the number of telemetry data to be analyzed generated in step S130. The number of satellite telemetry data received in step S110 may also be m.

Each of the m telemetry data to be analyzed may have n data values for each of the p related parameters. n may be determined corresponding to the collection period of each satellite telemetry data. For example, if the longest collection period among the periods for collecting raw data values of satellite telemetry data is 8 seconds and the collection period of satellite telemetry data is 12 hours, n is 12 hours divided by 8 seconds. May be 5400.

4 is an exemplary flow chart for explaining in more detail step (S130) of a method for detecting a singular motion of a satellite according to an embodiment of the present invention.

Referring to FIG. 4, the controller 110 may generate m related telemetry data corresponding to p related parameters among m satellite telemetry data (S131 ). Satellite telemetry data includes raw data values of each of a number of parameters greater than p. In step S131, related telemetry data corresponding to p related parameters may be selected from among the satellite telemetry data.

According to an embodiment, raw data values of satellite telemetry data may be values collected at different periods. In step S131, some of the raw data values may be removed so as to have one raw data value every preset collection period, and a collection time of some may be changed. For example, when the preset collection period is 8 seconds, when there are 8 raw data values collected every second, only one raw data value remains and 7 raw data values may be removed. If the collection time set according to the preset collection period is 00: 00: 16 seconds, if the raw data values collected in response to the first parameter were collected at 00: 00: 15 and 00: 00: 19, 00 The collection time is changed from the raw data values collected at 00:15 to 00:00:16, and the raw data values collected at 00:00:19 can be removed.

Each of the m related telemetry data may have n original data values for each of the p related parameters.

The controller 110 may pre-process m related telemetry data (S132). The range of raw data values may be different depending on the parameters of the related telemetry data. For example, a range of raw data values corresponding to a battery voltage and a range of raw data values corresponding to a battery current may be different from each other. Also, noise may be included in the raw data values.

According to an embodiment, the controller 110 may normalize n raw data values of each of m related telemetry data for each of p related parameters.

According to an example, (m x n) raw data values may be normalized for each of p related parameters. The (m x n) raw data values may mean n raw data values of each of m related telemetry data.

According to an example, by calculating the mean and standard deviation of (mxn) raw data values, and converting (mxn) raw data values into (mxn) standardized scores (Z-scores), (mxn) The raw data values can be normalized.

According to another example, after calculating the maximum and minimum values of (m x n) raw data values, (m x n) raw data values may be normalized by applying a min-max algorithm.

According to another example, by converting (mxn) raw data values into (mxn) standardized scores (Z-scores) and then applying a min-max algorithm to (mxn) standardized scores, (mxn) ) Raw data values can be normalized.

According to another example, n raw data values may be normalized for each of p related parameters of each of m related telemetry data.

According to another embodiment, the controller 110 may filter n raw data values for each of p related parameters of each of m related telemetry data using a moving average filter in step S132. . The controller 110 may generate n moving averaged data values by inputting n raw data values into the moving average filter.

According to another embodiment, the control unit 110 filters n raw data values for each of p related parameters of m related telemetry data using a moving average filter, Generate n moving averaged data values for each of the p related parameters of each of the tree data, and n moving averaged data values of each of m related telemetry data for each of the p related parameters Can be normalized. Normalization can be achieved by applying a standardized score and/or a min-max algorithm to (m x n) moving averaged data values.

The controller 110 may generate m telemetry data to be analyzed using the m related telemetry data preprocessed in step S132 (S133). Each of the m telemetry data to be analyzed may have n data values for each of the p related parameters. According to an example, the data value may be a value obtained by normalizing a moving average data value generated by applying a raw data value to a moving average filter.

Referring back to FIG. 3, the controller 110 may perform principal component analysis on each of the m telemetry data to be analyzed (S140). The control unit 110 performs principal component analysis on n data values corresponding to each of the p parameters, that is, (nxp) data values, for each of the m telemetry data to be analyzed, that is, m times. I can.

Principal component analysis is a statistical technique that extracts principal components that concisely express patterns of variance (variance and covariance) of many variables as a linear combination of original variables (average points for weight). In other words, if there are p variables, the information obtained there is summarized into a variables that are significantly smaller than p. In other words, it is possible to find a few new axes that best reflect the variance of many variables by rotating the axes of the p-dimensional space.

When principal component analysis is performed on (n x p) data values, (p x p) coefficients can be generated. In this embodiment, among (p x p) coefficients generated by performing principal component analysis on (n x p) data values, (p x 2) coefficients may be used. The (px 2) coefficients are the most important p coefficients (hereinafter referred to as'p first principal component coefficients') and the second most important p coefficients (hereinafter'p second principal component coefficients'). . The p first principal component coefficients are referred to as a first principal component coefficient set, and the p second principal component coefficients are referred to as a second principal component coefficient set.

In step S140, since principal component analysis for (nxp) data values is performed for each of the m telemetry data to be analyzed, that is, m times, m first principal component coefficient sets and m second principal component coefficients Sets can be created.

The i-th first principal component coefficient set and the i-th second principal component coefficient set are generated as a result of performing principal component analysis on (n x p) data values of the i-th analysis target telemetry data. i is a natural number of 1 or more and m or less.

The first principal component coefficient set consists of p first principal component coefficients, and the second principal component coefficient set consists of p second principal component coefficients. The first principal component coefficient set may correspond to a position in the p-dimensional space, and the coordinates of this position may be expressed by p first principal component coefficients. The second principal component coefficient set may also correspond to a position in the p-dimensional space, and the coordinates of this position may be expressed by p second principal component coefficients.

The controller 110 may generate a plurality of clusters by clustering m telemetry data to be analyzed based on m first principal component coefficient sets and m second principal component coefficient sets (S150).

The i-th analysis target telemetry data may correspond to a position in the 2p-dimensional space based on the i-th first principal component coefficient set and the i-th second principal component coefficient set. The coordinates of this position may be expressed using p i-th first principal component coefficients and p i-th second principal component coefficients.

Each of the m telemetry data to be analyzed may correspond to a position in a 2p-dimensional space based on a corresponding first principal component coefficient set and a second principal component coefficient set. It may be clustered into a plurality of clusters based on a position in a 2p-dimensional space corresponding to each of the m telemetry data to be analyzed.

The number of clusters may be k. k may be a natural number of 2 or more and less than m. k may be determined through gap statistics for m first principal component coefficient sets and m second principal component coefficient sets. Gap statistics can be used to determine the optimal number of clusters for m principal component coefficient data. k may correspond to the number of distinct operating states of the satellite 10.

K clusters may be generated by applying a k-means clustering method to m first principal component coefficient sets and m second principal component coefficient sets. The k-means clustering method may be applied to m locations in a 2p-dimensional space corresponding to m telemetry data to be analyzed.

The k clusters include first to kth clusters. Each of the m telemetry data to be analyzed may belong to one of k clusters. Some of the telemetry data to be analyzed may belong to the j-th cluster among the m telemetry data to be analyzed.

The controller 110 may determine specific telemetry data based on m pieces of analysis target telemetry data belonging to each of the plurality of clusters (S160). According to an embodiment, when only one analysis target telemetry data (eg, i-th analysis target telemetry data) belongs to any one of k clusters (eg, j-th cluster), only one cluster One analysis target telemetry data belonging to (eg, i-th analysis target telemetry data) may be determined as specific telemetry data. The singular telemetry data can be the basis for detecting the singular motion of the satellite. For example, the satellite may be estimated to have performed an unusual operation during a collection period in which the i th analysis target telemetry data determined as the singular telemetry data is collected.

In the following, it will be described in more detail how the satellite can estimate which unusual motion during which detailed collection period based on the i-th analysis target telemetry data determined as the singular telemetry data.

5 is a flowchart illustrating a method of detecting a singular motion of a satellite according to another embodiment.

Referring to FIG. 5, the controller 110 may determine specific telemetry data based on m pieces of analysis target telemetry data belonging to each of a plurality of clusters (S210 ). Step S210 may be performed through steps S110 to S160 described with reference to FIG. 3. In the following, it is assumed that the number of clusters is k.

In step S210, an attribute of each of the k clusters may be determined. The attributes of each of the clusters are i) the number of telemetry data to be analyzed belonging to each cluster, ii) the center position of each cluster, iii) each of the telemetry data to be analyzed belonging to each cluster at the center position of each cluster. Among the distances to the location, the maximum distance, and iv) may include a margin distance of each cluster.

As described above, the telemetry data to be analyzed may correspond to a position in a 2p-dimensional space. Clusters can also be expressed in a 2p-dimensional space. For example, when the first to third analysis target telemetry data belong to the first cluster, the center position of the first cluster is defined in a 2p-dimensional space, and corresponds to the first to third analysis target telemetry data, respectively. It may be an average of the first to third positions in the 2p-dimensional space. In this case, the maximum distance, which is an attribute of the first cluster, may be determined as the farthest distance from the center position in the 2p-dimensional space to the first to third positions in the 2p-dimensional space. The margin distance, which is an attribute of the first cluster, may be determined as a value divided by the number (eg, 3) of telemetry data to be analyzed belonging to the first cluster from the maximum distance of the first cluster.

For example, when only one analysis target telemetry data (eg, fourth analysis target telemetry data) belongs to the second cluster, the center position of the second cluster is the first corresponding to the fourth analysis target telemetry data. It is determined by 4 positions. According to an example, the maximum distance of the second cluster may be determined as 1/2 of the largest maximum distance among the maximum distances of k clusters. In this case, the margin distance of the second cluster may be determined equal to the maximum distance.

The controller 110 may receive new satellite telemetry data (S220). The new satellite telemetry data may be satellite telemetry data that has not been used to create clusters in step S210. The new satellite telemetry data may be satellite telemetry data that has not been received in step S110.

The controller 110 may generate new analysis target telemetry data from new satellite telemetry data based on the p related parameters (S230). Step S230 may be substantially the same as the process of generating m telemetry data to be analyzed from m satellite telemetry data in step S130. The telemetry data to be analyzed includes n data values for each of the p parameters.

The controller 110 may generate a third principal component coefficient set and a new fourth principal component coefficient set by performing principal component analysis on the newly analyzed telemetry data (S240). Step S240 may be substantially the same as the process of performing principal component analysis on each of the m telemetry data to be analyzed in step S140.

The controller 110 may perform principal component analysis on n data values, that is, (n x p) data values corresponding to each of the p parameters of the newly analyzed telemetry data.

When principal component analysis is performed on (n x p) data values, (p x p) coefficients can be generated. In this embodiment, among (p x p) coefficients, (p x 2) coefficients may be used. The (px 2) coefficients are the most important p coefficients (hereinafter referred to as'p third principal component coefficients') and the second most important p coefficients (hereinafter referred to as'p fourth principal component coefficients'). . The p third principal component coefficients are referred to as a third principal component coefficient set, and the p fourth principal component coefficients are referred to as a fourth principal component coefficient set. The third principal component coefficient set consists of p third principal component coefficients, and the fourth principal component coefficient set consists of p fourth principal component coefficients.

The controller 110 may calculate the position of the new analysis target telemetry data based on the third principal component coefficient set and the fourth principal component coefficient set (S250). The telemetry data to be analyzed may correspond to a position in a 2p-dimensional space, and the coordinates of this position may be expressed by p third principal component coefficients and p fourth principal component coefficients.

The control unit 110 may determine the nearest cluster closest to the location of the telemetry data to be analyzed from among the k clusters (S260). The controller 110 may determine the nearest cluster closest to the location of the telemetry data to be analyzed based on the central location, which is an attribute of the k clusters.

The controller 110 may determine whether or not the new analysis target telemetry data is included in the nearest cluster (S270). The controller 110 may calculate a distance between a location corresponding to the new analysis target telemetry data and a center location of the nearest cluster. The controller 110 may determine whether telemetry data to be analyzed is included in the nearest cluster by comparing the calculated separation distance with the sum of the maximum distance and the margin distance of the nearest cluster.

When the newly analyzed telemetry data is included in the nearest cluster, that is, when the calculated separation distance is less than or equal to the sum of the maximum distance and the margin distance of the nearest cluster, the controller 110 may update the property of the nearest cluster. Can be (S280). The number of telemetry data to be analyzed belonging to the nearest cluster can be increased by 1, and the center position, maximum distance, and margin distance of the nearest cluster can be updated by additionally considering the location of the newly analyzed telemetry data. . Telemetry data to be analyzed may newly belong to the nearest cluster.

When the new analysis target telemetry data is not included in the nearest cluster, that is, when the calculated separation distance is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the controller 110 The data may be determined as singular telemetry data (S290). The singular telemetry data can be the basis for detecting the singular motion of the satellite. For example, the satellite may be estimated to have performed an unusual operation during a collection period in which the i th analysis target telemetry data determined as the singular telemetry data is collected.

6 is a flowchart illustrating a method of detecting a singular motion of a satellite according to another embodiment.

Referring to FIG. 6, the controller 110 may receive c pieces of analysis target telemetry data (S310). Each of the c analysis target telemetry data may have n data values for each of the p related parameters. Here, c, p, and n may be 2 or more natural numbers. c, p, and n are independent of each other. c is a natural number less than m.

The c analysis target telemetry data may be analysis target telemetry data belonging to any one of k clusters generated by performing the steps S110 to S150 described with reference to FIG. 3. As described above, k is a natural number of 2 or more and less than m.

The controller 110 may perform principal component analysis on each of the c analysis target telemetry data (S320). The control unit 110 may perform principal component analysis on each of the m analysis target telemetry data including c analysis target telemetry data in step S140 of FIG. 3, and store the result of the principal component analysis. Can be doing. In this case, step S320 may be omitted.

The control unit 110 performs principal component analysis on n data values corresponding to each of the p parameters, that is, (nxp) data values, for each of the c analysis target telemetry data, that is, c times. I can.

When principal component analysis is performed on (n x p) data values, (p x p) coefficients can be generated. In the present embodiment, among (p x p) coefficients generated by performing principal component analysis on (n x p) data values, (p x 1) coefficients may be used. The (p x 1) coefficients are the most important p coefficients (hereinafter referred to as “p first principal component coefficients”). The p first principal component coefficients are referred to as a first principal component coefficient set. The first principal component coefficient set consists of p first principal component coefficients.

The first principal component coefficient set may correspond to a position in the p-dimensional space, and the coordinates of this position may be expressed by p first principal component coefficients.

In step S320, since principal component analysis on (n x p) data values is performed on each of the c analysis target telemetry data, that is, c times, c first principal component coefficient sets may be generated.

The controller 110 may calculate c first score data as a result of the principal component analysis in step S320 (S330). Each of the c first score data may consist of n first values.

The control unit 110 multiplies each of the c analysis target telemetry data by the p first principal component coefficients of the corresponding first principal component coefficient set among the c first principal component coefficient sets, thereby each consisting of n first values. The c first score data may be calculated. For example, the control unit 110 multiplies p data values having the same collection time among (nxp) data values of the i-th analysis target telemetry data by the p i-th first principal component coefficients, and then accumulates them. By performing n times, it is possible to calculate the i-th first score data consisting of n first values.

The controller 110 may determine whether the c pieces of first score data have periodicity (S340).

7 is an exemplary flow chart for explaining in more detail step (S340) of the method for detecting a singular motion of a satellite according to another embodiment of the present invention.

Referring to FIG. 7, the controller 110 may set a minimum interval (S341). The minimum interval may be set to an interval corresponding to, for example, 10 minutes. For example, the minimum interval may be set to approximately n/100 based on the number n of first values of the first score data. The minimum interval set in step S341 may be changed to a different value if it is estimated in step S343 that the first score data do not have periodicity.

The controller 110 may calculate at least one highest point and at least one lowest point from n first values of the first score data (S342). Among the highest points, two adjacent highest points may be separated by more than the minimum interval set in step S341. Among the lowest points, two adjacent lowest points may fall by more than the minimum interval set in step S341.

For example, the control unit 110 may detect a peak point that is separated from n first values by a minimum interval or more and is greater than or equal to a first specific value. Here, the first specific value may be set based on an average of the first values and a maximum value of the first values. For example, the first specific value may be set to 1/2 of the sum of the average of the first values and the maximum value of the first values.

The controller 110 may detect a lowest point that is separated by more than a minimum interval from the n first values and is less than or equal to a second specific value. Here, the second specific value may be set based on an average of the first values and a minimum value of the first values. For example, the specific value may be set to 1/2 of the sum of the average of the first values and the minimum value of the first values.

In step S342, the controller 110 may calculate a plurality of highest points and a plurality of lowest points from n first values of the first score data. For example, it is assumed that two highest points and three lowest points are calculated. Since the first values located at both ends of the n first values may not be the actual highest and lowest points, they may be excluded from the highest and lowest points calculated in step S342.

The control unit 110 may review the periodicity of the first score data (S343). For example, the control unit 110 includes n first values based on the distance between at least one highest point, the distance between at least one lowest point, and the number of at least one highest point and the number of at least one lowest point. It is possible to determine whether the first score data consisting of has periodicity.

For example, the control unit 110 may i) the maximum value of the distance between adjacent highest points is less than twice the minimum value of the distance between adjacent highest points, and ii) the maximum value of the distance between adjacent lowest points is adjacent lowest points. If it is less than twice the minimum value of the distance between and c) the difference between the number of lowest points and the number of highest points calculated in step S342 is less than or equal to the reference value, it may be determined that the first score data has periodicity. Here, the reference value may be 2. The reference value may be 1.

When it is estimated in step S343 that the first score data does not have periodicity, after changing the minimum interval in step S341, steps S341-S343 may be repeated. Even if it is estimated that the first score data does not have periodicity even if the minimum interval of step S341 is changed and repeated a predetermined number of times, it may be determined that the first score data does not have periodicity.

The controller 110 may determine the periodicity of the first score data (S344). For example, when at least one highest point and at least one lowest point alternately appear, the controller 110 may determine that first score data consisting of n first values has periodicity. When at least one highest point and at least one lowest point do not appear alternately, it may be determined that first score data consisting of n first values does not have periodicity.

Referring back to FIG. 6, when it is determined in step S340 that the first score data has periodicity, the controller 110 may determine first and second periodic characteristics (S350).

According to an example, the controller 110 may calculate at least two characteristics for each of the c pieces of first score data. For example, for each of the c first score data, the control unit 110 includes: i) the distance from the lowest point to the next highest point, ii) the distance from the highest point to the next lowest point, iii) the minimum value, which is the value of the lowest point, iv) the highest point. The maximum value, v) the difference between the value of the lowest point and the value of the next highest value, vi) the difference between the value of the highest point and the value of the next lowest value, vii) the slope from the lowest point to the next highest point, and viii) from the highest point to the next lowest point. It is possible to calculate at least two characteristics among the slopes of. The controller 110 may calculate all of the above-described eight characteristics for each of the c pieces of first score data. In the following, it is assumed that the control unit 110 calculates all of the above-described eight characteristics for each of the c pieces of first score data.

The number of eight features calculated for each of the c first score data may be the same. When two highest points and three lowest points are calculated for the first score data, i) to viii) characteristics may be calculated by two.

The control unit 110 may determine a characteristic with the smallest variance for the eight characteristics calculated for each of the c first score data as the first periodic characteristic. For example, when two of eight features are calculated for each of the first score data, there are a total of 2c data for each of the eight features. A variance value may be calculated for 2c pieces of data of each of the eight characteristics. Eight variance values may be calculated corresponding to each of the eight characteristics. However, this is exemplary, and the same number of feature values may not be calculated for each of the c first score data. For example, although two of eight characteristics are calculated for the a-th first score data, one eight characteristics may be calculated for each of the b-th first score data. However, in the following, it is assumed that, for example, two of each of the eight characteristics are calculated for each of the c first score data.

Before calculating the variance value for 2c pieces of data of each of the 8 characteristics, the control unit 110 normalizes the min-max for 2c pieces of data of each of the 8 characteristics, since the range of data values may be different according to the eight characteristics. I can make it. The control unit 110 may calculate a variance value for 2c normalized data of each of the eight characteristics.

The control unit 110 may determine a characteristic having the smallest variance value among the eight variance values as the first periodic characteristic.

The control unit 110 may determine a first periodic characteristic and a characteristic having the lowest linear dependence among the eight characteristics as the second periodic characteristic. Correlation analysis may be performed on 2c data of each of the eight characteristics. 2c pieces of data of the first periodic characteristic and 2c pieces of data having the smallest correlation coefficient may be determined. A characteristic corresponding to the 2c pieces of data having the smallest correlation coefficient may be determined as the second periodic characteristic.

When the first and second periodic characteristics are determined in step S350, the control unit 110 provides first and second periodic characteristic values corresponding to the first and second periodic characteristics for each of the c first score data. Can be calculated (S360).

For example, first periodic property values may be calculated based on 2c pieces of data of a first periodic property, and second periodic property values may be calculated based on 2c pieces of data of a second periodic property.

The 2c data values of the first periodic characteristic and the 2c data values of the second periodic characteristic may have different ranges. The controller 110 may normalize 2c data values of the first periodic characteristic and 2c data values of the second periodic characteristic to values in a preset range (eg, 0 or more and 1 or less). For example, a standardized score (Z-score) may be used.

For example, 2c data values of the first periodic characteristic may be converted into 2c standardized score values. In this case, the average and standard deviation of 2c data values of the first periodic characteristic may be used. By adding 1 to a value obtained by dividing the 2c standardized score values by the scale factor value and multiplying by 1/2, the 2c data values of the first periodic characteristic may be normalized to a value of 0 or more and 1 or less. The 2c data values of the second periodic characteristic may also be normalized to values of 0 or more and 1 or less in the same manner. The scale factor value may be preset as the largest value among the standardized scores calculated in each case of the clusters.

In the above example, the controller 110 may calculate 2c first periodic property values and 2c second periodic property values. The first periodic property values and the second periodic property values are values normalized to values within a preset range.

The controller 110 may determine a fuzzy rule based on the first and second periodic characteristic values of the c first score data (S370). In the above example, each of the first periodic property values and the second periodic property values is 2c.

According to an example, the controller 110 may determine first input sections for first periodic characteristic values. The number of first input sections may be, for example, four. In this case, four first input reference values may be used to classify the first input sections. For example, the four first input reference values may be 0 of ZO (zero), 1/3 of PS (Positive Small), 2/3 of PM (Positive Medium), and 1 PB (Positive Big). The first input sections may include a first section ZO, a second section PS, a third section PM, and a fourth section PB.

The controller 110 may calculate the number of first periodic characteristic values belonging to each of the first input sections. The number of first periodic feature values belonging to the first period ZO may be calculated as the number of first periodic feature values that are greater than ZO(0) and less than PS(1/3) among 2c first periodic feature values. The number of first periodic feature values belonging to the second section PS is 1/2 of the number of first periodic feature values that are greater than ZO (1/3) and less than PM (1/3) among 2c first periodic feature values Can be calculated as The number of first periodic feature values belonging to the third section PM is calculated as 1/2 of the number of first periodic feature values that are greater than PS(0) and less than PB(1/3) among 2c first periodic feature values. Can be. The number of first periodic feature values belonging to the fourth section PB may be calculated as the number of first periodic feature values that are greater than PM(0) and less than PB(1/3) among 2c first periodic feature values.

The controller 110 may determine second input sections for the second periodic characteristic values. The number of second input sections may be, for example, four. Four second input reference values may be used to distinguish the second input sections. For example, the four second input reference values may be 0 ZO (zero), 1/3 PS (Positive Small), 2/3 PM (Positive Medium), and 1 PB (Positive Big). The second input sections may include a first section ZO, a second section PS, a third section PM, and a fourth section PB.

The controller 110 may calculate the number of second periodic characteristic values belonging to each of the second input sections. The number of second periodic property values belonging to the first period ZO may be calculated as the number of second periodic property values that are greater than ZO(0) and less than PS(1/3) among 2c second periodic property values. The number of second periodic feature values belonging to the second section PS is 1/2 of the number of second periodic feature values that are greater than ZO (1/3) and less than PM (1/3) among 2c second periodic feature values Can be calculated as The number of second periodic feature values belonging to the third section (PM) is calculated as 1/2 of the number of second periodic feature values that are greater than PS(0) and less than PB(1/3) among 2c second periodic feature values. Can be. The number of second periodic feature values belonging to the fourth section PB may be calculated as the number of second periodic feature values that are greater than PM(0) and less than PB(1/3) among 2c second periodic feature values.

The controller 110 may determine a fuzzy output section. The fuzzy output section may be sections classified according to a distribution value. The number of fuzzy output sections may be 4, for example. Four fuzzy output sections are ZO (zero) when the distribution value is more than 0 and less than 1/4, PS (positive small) when the distribution value is more than 1/4 and less than 1/2, and the distribution value is more than 1/2 3/ It may be PM (Positive Medium) when it is less than 4, and PB (Positive Big) when the distribution value is 3/4 or more and 1 or less.

The controller 110 is based on the number of first periodic characteristic values belonging to each of the first input intervals and the number of second periodic characteristic values belonging to each of the second input intervals, the first input intervals and the second input Distribution values of each of the input regions defined by a combination of sections may be calculated.

When there are 4 first input sections and 4 second input sections, there are 16 input areas defined by a combination of the first input sections and the second input sections. For example, if a first periodic feature values belong to the first section ZO of the first input section, and b second periodic feature values belong to the first section ZO of the second input section, A data value of (a+b) is determined in the first input area defined by the combination of the first section ZO of the first input section and the first section ZO of the second input section. In this way, the data values of each of the 16 input regions may be determined based on the number of first periodic characteristic values belonging to each of the first input intervals and the number of second periodic characteristic values belonging to each of the second input intervals. have.

The distribution value of the input area may be defined as a value obtained by dividing a data value of a corresponding input area by a maximum value among data values of 16 input areas. For example, if the maximum value of the data values of the 16 input areas is c, the distribution value of the first input area may be determined as (a+b)/c.

The controller 110 may determine a fuzzy rule based on the fuzzy output section and distribution values of the input regions. The controller 110 may define distribution values of the input regions according to the previously determined fuzzy output section. For example, when (a+b)/c is 0.2, the first input region may be determined as ZO according to the previously determined purge output section. In this case, when the first periodic characteristic is ZO and the second periodic characteristic is ZO, a first fuzzy rule that outputs ZO according to the first input area may be generated. As described above, first to sixteenth fuzzy rules may be generated in correspondence to the first to sixteenth input regions.

The controller 110 may determine the specific telemetry data using the fuzzy rule determined in step S370 (S380). For example, the controller 110 may determine the specific telemetry data using the first to sixteenth fuzzy rules.

8 is an exemplary flow chart for explaining in more detail step S380 of a method for detecting a singular motion of a satellite according to another embodiment of the present invention.

Referring to FIG. 8, the controller 110 may receive new satellite telemetry data (S410). The new satellite telemetry data may be satellite telemetry data that has not been received in step S110.

The controller 110 may generate new analysis target telemetry data from new satellite telemetry data based on the p related parameters (S420). Step S420 may be substantially the same as the process of generating m telemetry data to be analyzed from m satellite telemetry data in step S130. The telemetry data to be analyzed includes n data values for each of the p parameters.

The controller 110 may calculate a fuzzy output value according to the center of sums method by applying the new analysis target telemetry data to the fuzzy rule determined in step S370 (S440).

For example, by performing principal component analysis on telemetry data to be analyzed, first score data having n first values may be calculated. First and second periodic characteristic values may be calculated for the first score data of the newly analyzed telemetry data. By applying the first and second periodic characteristic values to the fuzzy rule, a fuzzy output value may be calculated. The fuzzy output value can be calculated according to the center of sums method, for example.

The controller 110 may determine whether the new satellite telemetry data is unusual telemetry data based on the fuzzy output value of the new analysis target telemetry data (S450).

According to an example, the control unit 110 may compare the fuzzy output value of the newly analyzed telemetry data with a reference value. If the fuzzy output value of the newly analyzed telemetry data is less than the reference value, the new analysis target It can be determined that the telemetry data is peculiar telemetry data. If the fuzzy output value of the new analysis target telemetry data is greater than or equal to the reference, the new analysis target telemetry data may be determined rather than the specific telemetry data. The reference value varies according to the fuzzy rule, but may be preset, for example, to 0.8.

9 is an exemplary flowchart illustrating an operation when first score data do not have periodicity in step S340 of a method of detecting a singular motion of a satellite according to another embodiment of the present invention.

Referring to FIG. 9, if it is determined in step S340 of FIG. 6 that the first score data do not have periodicity, the controller 110 performs principal component analysis on each of the c analysis target telemetry data, Two second principal component coefficient sets may be generated (S341). Each of the c second principal component coefficient sets may consist of p second principal component coefficients.

The control unit 110 may perform principal component analysis on each of the m analysis target telemetry data including c analysis target telemetry data in step S140 of FIG. 3, and store the result of the principal component analysis. Can be doing. In this case, step S341 may be omitted.

The control unit 110 performs principal component analysis on n data values corresponding to each of the p parameters, that is, (nxp) data values, for each of the c analysis target telemetry data, that is, c times. I can.

When principal component analysis is performed on (n x p) data values, (p x p) coefficients can be generated. In this embodiment, among (pxp) coefficients generated by performing principal component analysis on (nxp) data values, (px 2 coefficients may be used. (px 2) coefficients are the most important p coefficients) (Hereinafter referred to as “p first principal component coefficients”) and second most important p coefficients (hereinafter referred to as “p second principal component coefficients”). Are referred to as a second principal component coefficient set, the second principal component coefficient set consisting of p second principal component coefficients.

The second principal component coefficient set may correspond to a position in the p-dimensional space, and the coordinates of this position may be expressed by p second principal component coefficients.

In step S341, since principal component analysis for (n x p) data values is performed on each of the c analysis target telemetry data, that is, c times, c second principal component coefficient sets may be generated.

The controller 110 may calculate c pieces of second score data as a result of the principal component analysis in step S341 (S342). Each of the c pieces of second score data may consist of n number of second values.

The control unit 110 multiplies each of the c analysis target telemetry data by p second principal component coefficients of the corresponding second principal component coefficient set among the c second principal component coefficient sets, thereby each consisting of n second values. It is possible to calculate c pieces of second score data. For example, the control unit 110 multiplies p data values having the same collection time among (nxp) data values of the i-th analysis target telemetry data by p number of i-th second principal component coefficients, and then accumulates it. By performing n times, it is possible to calculate the i-th second score data consisting of n second values.

The controller 110 may determine whether the c pieces of second score data have periodicity (S343). The step S343 may be performed in substantially the same manner as the method of determining whether the c first score data of the step S340 of FIGS. 6 and 7 have periodicity. A method for the control unit 110 to determine whether the c pieces of second score data have periodicity is not described repeatedly.

If it is determined in step S343 that the second score data has periodicity, the controller 110 may determine third and fourth periodic characteristics (S344). Step S344 may be performed in substantially the same manner as the method of determining first and second periodic characteristics for the first score data in step S350 of FIG. 6. In step S344, the third periodic characteristic corresponds to the first periodic characteristic of step S350, and in step S344 the fourth periodic characteristic corresponds to the second periodic characteristic of step S350.

The controller 110 may calculate third and fourth periodic characteristic values corresponding to the third and fourth periodic characteristics for each of the c pieces of second score data (S345). Step S345 is substantially the same as the method of calculating first and second periodic feature values corresponding to the first and second periodic features for each of the c first score data in step S360 of FIG. 6. It can be done in a way. In step S345, the third periodic feature values correspond to the first periodic feature values in step S360, and in step S345, the fourth periodic feature values correspond to the second periodic feature values in step S360.

The controller 110 may determine a fuzzy rule based on the third and fourth periodic characteristic values of the c second score data (S346). Step S346 may be performed in substantially the same manner as a method of determining a fuzzy rule based on the first and second periodic characteristic values of the c first score data in step S370 of FIG. 6. A method of determining a fuzzy rule based on the third and fourth periodic characteristic values of the c second score data will not be described repeatedly.

In step S380, the controller 110 may determine the specific telemetry data using the fuzzy rule determined in step S346. The controller 110 may perform the steps S410 to S450 described with reference to FIG. 8 by using the fuzzy rule determined in step S346.

10 is an exemplary flow chart for explaining an operation according to another example of a case in which first score data do not have periodicity in step S340 of a method of detecting a singular motion of a satellite according to another embodiment of the present invention.

Referring to FIG. 10, for each of the c pieces of first score data, the controller 110 may calculate a plurality of statistical values for n first values of each of the first score data (S510).

According to an embodiment, a plurality of statistical values for n first values of each first score data are at least two of a minimum value, a maximum value, a median value, and a standard deviation for n first values of each first score data. It may include.

For example, a plurality of statistical values for n first values of each first score data may include a minimum value, a maximum value, a median value, and a standard deviation for the n first values of each first score data. . In this case, corresponding to the number of first score data, c minimum values, c maximum values, c median values, and c standard deviations may be calculated. Below, it is assumed that the statistical values include four statistical values: the minimum value, the maximum value, the median value, and the standard deviation.

According to an example, c minimum values, c maximum values, c median values, and c standard deviations may be normalized to, for example, a standardized score (Z-score). In this case, the statistical values may include a normalized value of a minimum value, a normalized value of a maximum value, a normalized value of a median, and a normalized value of a standard deviation.

The controller 110 may determine a position of each of the c first score data in a multidimensional space based on a plurality of statistical values calculated for each of the c first score data (S520).

If the statistical values contain four statistical values of minimum, maximum, median, and standard deviation, each of the c first score data is based on the four statistical values of its n first values in a four-dimensional space. It can be expressed as the position of the image. The coordinates of the location in this four-dimensional space can be expressed using four statistical values. In step S520, c positions in the 4D space may be determined corresponding to each of the c first score data.

The controller 110 may calculate the center position based on the distribution of the c first score data in the multidimensional space (S530). For example, the control unit 110 may determine a center position based on c positions in a 4D space.

The control unit 110 may calculate a center separation distance of each of the c first score data based on a difference between the center position and the positions of the c first score data (S540). For example, the controller 110 may calculate c center separation distances based on a difference between the center location and each of the c locations in the 4D space.

The controller 110 may determine the specific telemetry data based on the center separation distances (S550). According to an embodiment, the controller 110 may calculate an average separation distance and a standard deviation of the center separation distances. The controller 110 may determine a lower limit and an upper limit based on a preset multiple of the average separation distance and the standard deviation. The preset multiple may be 2, for example. The multiple may be preset to a value other than 2. The upper limit may be determined as a value obtained by adding a preset multiple of the standard deviation to the average separation distance, and the upper limit may be determined as a value obtained by subtracting a preset multiple of the standard deviation from the average separation distance.

The controller 110 may define a normal range based on a lower limit and an upper limit. When the center separation distance of any one of the first score data among the c first score data is out of the normal range, the control unit 110 transmits the analysis target telemetry data corresponding to the one center separation distance to a specific telemetry. It can be determined by tree data. The singular telemetry data can be the basis for detecting the singular motion of the satellite. For example, the satellite may be estimated to have performed an unusual operation during a collection period in which the i th analysis target telemetry data determined as the singular telemetry data is collected.

In FIG. 10, an operation according to another example of the case in which the first score data does not have periodicity in step S340 is described, but the steps of FIG. 10 even when the second score data does not have periodicity in step S343 S510-550) may be performed.

11 is a flowchart illustrating a method of detecting a detailed singular motion of a satellite according to another embodiment.

Referring to FIG. 11, the control unit 110 may receive telemetry data for precision analysis (S610). Telemetry data to be precise analysis may have n data values for each of the p related parameters.

According to an example, the telemetry data to be precise analysis may be telemetry data to be analyzed determined as specific telemetry data in step S160 of FIG. 3. For example, the telemetry data to be precise analysis may be telemetry data to be analyzed that belong to any one cluster alone among a plurality of clusters generated in step S150 of FIG. 3.

According to another example, the telemetry data to be precisely analyzed may be telemetry data to be analyzed newly determined as the specific telemetry data in step S290 of FIG. 5.

According to another example, the telemetry data to be precise analysis may be telemetry data to be analyzed determined as specific telemetry data in step S380 of FIG. 6.

According to another example, the telemetry data to be precise analysis may be telemetry data to be analyzed newly determined as specific telemetry data in step S450 of FIG. 8.

According to another example, the telemetry data to be precise analysis may be telemetry data to be analyzed determined as specific telemetry data in step S550 of FIG. 10.

The controller 110 may generate d telemetry data to be analyzed in detail (S620). The d telemetry data for detailed analysis may be generated by dividing the telemetry data for detailed analysis into preset time units. The preset time unit may correspond to the orbiting period of the satellite 10. The preset time unit may be 90 minutes, for example.

Each of the d detailed analysis target telemetry data may have s data values for each of the p related parameters. s may be a natural number, or may be a natural number approximating n/d. When n is 5400, d may be 8, in this case, s may be 675.

The controller 110 may perform principal component analysis on each of the d telemetry data to be analyzed in detail (S630). The control unit 110 performs principal component analysis on s data values corresponding to each of the p parameters, that is, (sxp) data values, for each of the d telemetry data to be analyzed, that is, d times. can do.

When principal component analysis is performed on (s x p) data values, (p x p) coefficients may be generated. In this embodiment, among (p x p) coefficients generated by performing principal component analysis on (s x p) data values, (p x 1) coefficients may be used. The (p x 1) coefficients are the most important p coefficients (hereinafter referred to as “p principal component coefficients”). The p principal component coefficients are referred to as the principal component coefficient set. The principal component coefficient set consists of p principal component coefficients.

The principal component coefficient set may correspond to a position in a p-dimensional space, and the coordinates of this position may be expressed by p principal component coefficients.

In step S630, since principal component analysis for (s x p) data values is performed on each of the d telemetry data to be analyzed, that is, d times, d principal component coefficient sets may be generated.

The controller 110 may determine an abnormal operation period and a normal operation period based on the p main component coefficients of each of the d main component coefficient sets (S640).

According to an embodiment, the control unit 110 may associate d principal component coefficient sets to d positions in a p-dimensional space, respectively. The control unit 110 may calculate center points of d positions in the p-dimensional space. The controller 110 may calculate d center separation distances between a center point and d locations in the p-dimensional space, respectively. The controller 110 may calculate an average separation distance and a standard deviation of the d center separation distances.

The control unit 110 may determine the upper limit based on the average separation distance and a preset multiple of the standard deviation. The preset multiple may be 1, for example. The preset multiple may be 2, for example. In addition to 1 and 2, the preset multiple may be a real number greater than 0. The upper limit may be determined as a value obtained by adding a predetermined multiple of the standard deviation to the average separation distance.

The control unit 110 may compare the upper limit with each of the d center separation distances. The controller 110 may determine an abnormal operation section and a normal operation section based on the comparison result. Among the d center separation distances, the detailed analysis target telemetry data corresponding to the center separation distance less than or equal to the upper limit may be determined to correspond to a normal operation section. Among the d center separation distances, the detailed analysis target telemetry data corresponding to the center separation distance exceeding the upper limit may be determined to correspond to the abnormal operation section.

For example, assuming that the center separation distances corresponding to the d detailed analysis target telemetry data are 0.0184, 0.0224, 0.03, 0.03, 0.006, 0.013, 0.1014, and 0.0189, respectively, the average of the center separation distances is 0.0300, and the center separation. The standard deviation of the distances is 0.0299. If the preset multiple is set to 1, the upper limit is 0.0599, and the center separation distance of 0.1014 corresponding to the seventh detailed analysis target telemetry data exceeds the upper limit. In this case, an operation section corresponding to the seventh detailed analysis target telemetry data corresponds to an abnormal operation section. In other words, it may be estimated that the satellite 10 has abnormally operated during the period in which telemetry data to be analyzed for the seventh detailed analysis is collected.

The controller 110 may determine an abnormal operation parameter based on principal component coefficients of telemetry data to be analyzed in detail corresponding to the abnormal operation section (S650).

According to an embodiment, the controller 110 may determine the first principal component coefficients based on p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the normal operation section. .

For example, the controller 110 may calculate a coefficient average of the absolute values of the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the normal operation period. The control unit 10 converts the principal component coefficients having a magnitude greater than the coefficient average among the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the normal operation section as first principal component coefficients. You can decide.

The controller 110 may determine the second principal component coefficients based on the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the abnormal operation period.

For example, the controller 110 may calculate a coefficient average of the absolute values of the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the abnormal operation section. The control unit 110 converts the principal component coefficients having a magnitude greater than the coefficient average among the p principal component coefficients of the principal component coefficient set corresponding to the detailed analysis target telemetry data determined to correspond to the abnormal operation section as second principal component coefficients. You can decide.

The controller 110 may determine an abnormal operation parameter based on a result of comparison between the first principal component coefficients and the second principal component coefficients. For example, the controller 110 may determine first main parameters corresponding to the first main component coefficients and determine second main parameters corresponding to the second main component coefficients. For example, it is assumed that the identification numbers of the first main parameters are 13, 18, 19, 25, 32, and the identification numbers of the second main parameters are 9, 10, 13, 18, 19, 25, 33.

The controller 110 may determine a parameter that corresponds to the second main parameters but does not correspond to the first main parameters as the abnormal operation parameter. In the above example, parameters having identification numbers 9, 10, and 33 may be determined as abnormal operation parameters.

The controller 110 may determine a parameter that corresponds to the first major parameters but does not correspond to the second major parameters as the abnormal operation parameter. In the above example, a parameter with an identification number of 32 may also be determined as an abnormal operation parameter.

In the above example, it may be estimated that the satellite 10 has performed abnormal operation in parameters having identification numbers 9, 10, 32, and 33 during the period in which the seventh detailed analysis target telemetry data is collected.

Accordingly, it is possible to detect an abnormal motion of the satellite 10 without the assistance of several experts with expertise, and specifically, it is possible to detect when and which parameter an abnormal motion has occurred. Since this method can be applied to various satellite systems as it is, it can have wide scalability.

The various embodiments described above are exemplary, and are not to be performed independently as they are distinguished from each other. The embodiments described in the present specification may be implemented in combination with each other.

The various embodiments described above may be implemented in the form of a computer program that can be executed through various components on a computer, and such a computer program may be recorded in a computer-readable medium. In this case, the medium may be one that continuously stores a program executable by a computer, or temporarily stores a program for execution or download. In addition, the medium may be a variety of recording means or storage means in a form in which a single piece of hardware or several pieces of hardware are combined. The medium is not limited to a medium directly connected to a computer system, but may be distributed on a network. Examples of media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, And there may be ones configured to store program instructions, including ROM, RAM, flash memory, and the like. In addition, examples of other media include an app store that distributes applications, a site that supplies or distributes various software, and a recording medium or a storage medium managed by a server.

In the present specification, a "unit", a "module", etc. may be a hardware component such as a processor or a circuit, and/or a software component executed by a hardware configuration such as a processor. For example, "unit", "module", etc. are components such as software components, object-oriented software components, class components and task components, processes, functions, properties, It can be implemented by procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays and variables.

The above description of the present invention is for illustrative purposes only, and those of ordinary skill in the art to which the present invention pertains will be able to understand that other specific forms can be easily modified without changing the technical spirit or essential features of the present invention. will be. Therefore, it should be understood that the embodiments described above are illustrative in all respects and are not limiting. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as being distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims to be described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

Claims (20)

In the method performed by the computing device,
Receiving satellite telemetry data;
Determining p related parameters that are related to each other from among the parameters of the satellite telemetry data;
Generating m telemetry data to be analyzed from the satellite telemetry data based on the p related parameters, each of the m telemetry data to be analyzed, each of the p related parameters Having n data values for n;
Performing principal component analysis on each of the m telemetry data to be analyzed to generate m first principal component coefficient sets and m second principal component coefficient sets, wherein each of the m first principal component coefficient sets consisting of p first principal component coefficients, each of the m second principal component coefficient sets consisting of p second principal component coefficients;
Clustering the m telemetry data to be analyzed based on the m first principal component coefficient sets and the m second principal component coefficient sets to generate a plurality of clusters; And
Including the step of determining specific telemetry data based on the m telemetry data to be analyzed belonging to each of the plurality of clusters,
And the number of clusters is k determined through gap statistics for the m first principal component coefficient sets and the m second principal component coefficient sets. The method of claim 1,
And the p related parameters are determined based on a result of a keyword search for the names of the parameters of the satellite telemetry data. delete In the method performed by the computing device,
Receiving satellite telemetry data;
Determining p related parameters that are related to each other from among the parameters of the satellite telemetry data;
Generating m telemetry data to be analyzed from the satellite telemetry data based on the p related parameters, each of the m telemetry data to be analyzed, each of the p related parameters Having n data values for n;
Performing principal component analysis on each of the m telemetry data to be analyzed to generate m first principal component coefficient sets and m second principal component coefficient sets, wherein each of the m first principal component coefficient sets consisting of p first principal component coefficients, each of the m second principal component coefficient sets consisting of p second principal component coefficients;
Clustering the m telemetry data to be analyzed based on the m first principal component coefficient sets and the m second principal component coefficient sets to generate a plurality of clusters; And
Including the step of determining specific telemetry data based on the m telemetry data to be analyzed belonging to each of the plurality of clusters,
Generating the m number of analysis target telemetry data,
generating m related telemetry data corresponding to the p related parameters among m satellite telemetry data, wherein each of the m related telemetry data is each of the p related parameters Having n original data values for; And
And pre-processing the m related telemetry data. The method of claim 4,
The preprocessing of the m related telemetry data comprises normalizing the n raw data values of each of the m related telemetry data for each of the p related parameters. A method of detecting an unusual motion of a satellite. The method of claim 4,
The preprocessing of the m related telemetry data includes filtering the n raw data values for each of the p related parameters of each of the m related telemetry data using a moving average filter. A method for detecting a singular motion of a satellite, comprising: a. delete The method of claim 1,
The generating of the plurality of clusters includes generating the k clusters by applying a k-means clustering method to the m first principal component coefficient sets and the m second principal component coefficient sets. A method for detecting a singular motion of a satellite, comprising: a. The method of claim 1,
When only one telemetry data to be analyzed belongs to any one of the plurality of clusters, the one telemetry data to be analyzed is determined as the singular telemetry data. . In the method performed by the computing device,
Receiving satellite telemetry data;
Determining p related parameters that are related to each other from among the parameters of the satellite telemetry data;
Generating m telemetry data to be analyzed from the satellite telemetry data based on the p related parameters, each of the m telemetry data to be analyzed, each of the p related parameters Having n data values for n;
Performing principal component analysis on each of the m telemetry data to be analyzed to generate m first principal component coefficient sets and m second principal component coefficient sets, wherein each of the m first principal component coefficient sets consisting of p first principal component coefficients, each of the m second principal component coefficient sets consisting of p second principal component coefficients;
Clustering the m telemetry data to be analyzed based on the m first principal component coefficient sets and the m second principal component coefficient sets to generate a plurality of clusters; And
Including the step of determining specific telemetry data based on the m telemetry data to be analyzed belonging to each of the plurality of clusters,
The step of creating the plurality of clusters includes determining an attribute of each of the plurality of clusters,
The attributes of each of the plurality of clusters,
i) The number of telemetry data to be analyzed belonging to each cluster,
ii) the central position of each cluster,
iii) the maximum distance among the distances from the center position of each cluster to each position of the analysis target telemetry data belonging to each cluster, and
iv) A method for detecting a singular motion of a satellite, comprising the margin distance of each cluster. The method of claim 10,
Receiving new satellite telemetry data;
Generating new telemetry data to be analyzed from the new satellite telemetry data based on the p related parameters;
Performing principal component analysis on the newly analyzed telemetry data to generate a third principal component coefficient set and a new fourth principal component coefficient set;
Calculating a location of the new analysis target telemetry data based on the third principal component coefficient set and the fourth principal component coefficient set; And
And determining a nearest cluster from among the plurality of clusters that is closest to a location of the newly analyzed telemetry data. The method of claim 11,
When the distance from the center position of the nearest cluster to the position of the new analysis target telemetry data is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the new analysis target telemetry data And determining the singular telemetry data as the singular telemetry data. The method of claim 11,
If the distance from the center position of the nearest cluster to the position of the newly analyzed telemetry data is less than or equal to the sum of the maximum distance and the margin distance of the nearest cluster, updating an attribute of the nearest cluster A method for detecting a singular motion of a satellite, characterized in that it further comprises. The method of claim 10,
The margin distance of each cluster is determined as a value obtained by dividing the maximum distance of each cluster by the number of telemetry data to be analyzed belonging to each cluster. In the method performed by the computing device,
Generating a plurality of clusters each having an attribute, wherein the attribute of each cluster is the number of telemetry data to be analyzed belonging to each cluster, a central position of each cluster, and each cluster at a central position of each cluster. Including a maximum distance and a margin distance of each cluster from among distances to positions of each of the telemetry data to be analyzed belonging to;
Receiving new satellite telemetry data;
generating new telemetry data to be analyzed from the new satellite telemetry data based on p related parameters;
Performing principal component analysis on the newly analyzed telemetry data to generate a first principal component coefficient set and a second principal component coefficient set;
Calculating a position of the new analysis target telemetry data based on the first principal component coefficient set and the second principal component coefficient set; And
And determining a nearest cluster from among the plurality of clusters that is closest to a location of the newly analyzed telemetry data. The method of claim 15,
When the distance from the center position of the nearest cluster to the position of the new analysis target telemetry data is greater than the sum of the maximum distance and the margin distance of the nearest cluster, the new analysis target telemetry data And determining the singular telemetry data as the singular telemetry data. The method of claim 15,
If the distance from the center position of the nearest cluster to the position of the newly analyzed telemetry data is less than or equal to the sum of the maximum distance and the margin distance of the nearest cluster, updating an attribute of the nearest cluster A method for detecting a singular motion of a satellite, characterized in that it further comprises. The method of claim 15,
The margin distance of each cluster is determined as a value obtained by dividing the maximum distance of each cluster by the number of telemetry data to be analyzed belonging to each cluster. A computer program stored on a medium for executing the method of any one of claims 1, 2, 4 to 6, and 8 to 18 using a computing device. A memory for storing satellite telemetry data; And
Determine p related parameters that are related to each other among the parameters of the satellite telemetry data, and for each of the p related parameters from the satellite telemetry data based on the p related parameters m number of analysis target telemetry data having n data values are generated, and principal component analysis is performed on each of the m number of analysis target telemetry data. One principal component coefficient set, and m second principal component coefficient sets, each consisting of p second principal component coefficients, are generated, and based on the m first principal component coefficient sets and the m second principal component coefficient sets, the m number of At least one, configured to cluster telemetry data to be analyzed to create a plurality of clusters, and to determine specific telemetry data based on the m telemetry data to be analyzed belonging to each of the plurality of clusters. Including a processor,
And the number of clusters is k determined through gap statistics for the m first principal component coefficient sets and the m second principal component coefficient sets.

Images