Spacecraft telemetry parameter anomaly detection method and device

Abstract

The invention provides a spacecraft telemetry parameter anomaly detection method and device, and the method comprises the steps: carrying out the parameter classification of telemetry parameter data, and generating a parameter type; determining a corresponding target fitting algorithm according to the parameter type through the fitting algorithm distribution table; through the threshold value fitting result of the target fitting algorithm and the algorithm threshold value weight, anomaly detection is carried out on the telemetering parameter data, and a parameter detection result is generated, so that the whole process automation from threshold value fitting to optimization is realized, the operation and maintenance cost is reduced, and the operation efficiency of the system is improved; the threshold value is dynamically adjusted according to the characteristics of different types of data, the adaptability is high, the method is suitable for various complex scenes, the dependence on artificial experience is reduced, the system flexibility is enhanced, and the telemetry parameter monitoring coverage area is increased; by automatically optimizing the threshold parameter, the problem of misjudgment or missing detection possibly brought by a traditional fixed threshold method is avoided, and the accuracy of anomaly detection is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of space TT&C, and particularly to a method and device for detecting anomalies in spacecraft telemetry parameters. Background Art

[0002] In space TT&C missions, the ground TT&C system needs to timely detect possible abnormal states and discover potential problems through various types of telemetry data transmitted by the spacecraft to ensure the safe operation of the current spacecraft. To improve the performance of telemetry processing, the ground TT&C system usually uses a distributed parallel processing method for telemetry data parsing and processing. When the ground receives the telemetry data transmitted by the spacecraft, the telemetry source packet is extracted from the data frame according to the data format, and then the data is allocated according to the characteristics such as the subsystems and source packets of the telemetry, and multiple processing nodes complete the parallel processing of the data.

[0003] In related technologies, a parameter threshold judgment method based on expert knowledge is adopted. The ground TT&C system usually judges whether a spacecraft has a fault by determining whether multiple telemetry parameters exceed a specific range. Usually, a fault criterion is designed for a certain type of equipment failure, and then judgment rules are written, for example: A > 5 and B < 10, etc. Whether the spacecraft has a fault is determined according to the judgment result of the parameter value. This method requires experts from the manufacturing department to participate and provide relevant knowledge. However, as the division of labor in the spacecraft manufacturing department becomes more and more detailed, it is difficult to gather experts in various fields to provide all the knowledge. Moreover, this knowledge is very limited, only detecting key parts, and very few telemetry parameters are used, which cannot cover tens of thousands of telemetry parameters, resulting in a large number of monitoring blind spots. Summary of the Invention

[0004] An object of the present invention is to provide a method for detecting anomalies in spacecraft telemetry parameters, which realizes the full-process automation from threshold fitting to optimization, reduces the operation and maintenance costs, and improves the operation efficiency of the system; dynamically adjusts the threshold according to the characteristics of different types of data, has strong adaptability, is applicable to various complex scenarios, reduces the dependence on manual experience, enhances the flexibility of the system, and increases the coverage area of telemetry parameter monitoring; avoids the possible misjudgment or missed detection problems brought by the traditional fixed threshold method by automatically optimizing the threshold parameters, and significantly improves the accuracy of anomaly detection. Another object of the present invention is to provide a device for detecting anomalies in spacecraft telemetry parameters. Still another object of the present invention is to provide a computer-readable medium. Yet another object of the present invention is to provide a computer device.

[0005] To achieve the above object, on the one hand, the present invention discloses a method for detecting anomalies in spacecraft telemetry parameters, including:

[0006] Obtaining the telemetry parameter data transmitted by the spacecraft;

[0007] Classify the telemetry parameter data to generate parameter types;

[0008] Determine the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm allocation table;

[0009] Perform anomaly detection on the telemetry parameter data through the threshold fitting result and algorithm threshold weight of the target fitting algorithm to generate a parameter detection result. The threshold fitting result is obtained by pre-training the target fitting algorithm.

[0010] Preferably, classifying the telemetry parameter data to generate parameter types includes:

[0011] Judge whether the telemetry values in the telemetry parameter data are all integers;

[0012] If they are all integers, judge whether the number of different telemetry values in the telemetry parameter data is less than or equal to a preset number threshold;

[0013] If so, determine that the parameter type is a status quantity;

[0014] If not, perform trend analysis on the telemetry parameter data. If the telemetry parameter data is on an upward trend, determine that the parameter type is a counter;

[0015] If the telemetry parameter data is not on an upward trend, determine that the parameter type is an enumeration;

[0016] If they are not all integers, judge whether the standard deviation of the telemetry parameter data is less than a preset standard deviation threshold;

[0017] If so, determine that the parameter type is a stable pseudo-period;

[0018] If not, determine that the parameter type is a pseudo-period.

[0019] Preferably, the target fitting algorithm is an independent selection algorithm. The independent selection algorithm includes one of the quartile method, percentile method, sigma criterion, and self-help clustering algorithm. The threshold fitting result is an independent selection threshold interval, and the independent selection threshold interval includes one of the quartile threshold interval, percentile threshold interval, sigma threshold interval, and self-help clustering threshold interval;

[0020] Perform anomaly detection on the telemetry parameter data through the threshold fitting result and algorithm threshold weight of the target fitting algorithm to generate a parameter detection result, including:

[0021] Perform anomaly detection on the telemetry parameter data according to the independent selection threshold interval, and count the independent anomaly count;

[0022] Judge whether the independent anomaly count is greater than a preset independent anomaly count threshold;

[0023] If so, determine that the parameter detection result is abnormal;

[0024] If not, determine that the parameter detection result is normal.

[0025] Preferably, the target fitting algorithm is a combination selection algorithm, and the combination selection algorithm includes any combination of the quartile method, percentile method, sigma criterion, and self-organizing clustering algorithm. The threshold fitting result is a combination selection threshold interval, and the combination selection threshold interval includes any combination of the quartile threshold interval, percentile threshold interval, sigma threshold interval, and self-organizing clustering threshold interval;

[0026] Perform anomaly detection on the telemetry parameter data through the threshold fitting result of the target fitting algorithm and the algorithm threshold weight, and generate a parameter detection result, including:

[0027] According to the combination selection threshold interval, perform anomaly detection on the telemetry parameter data respectively, and count multiple independent anomaly counts;

[0028] According to the preset independent anomaly count threshold, determine the anomaly results of multiple independent anomaly counts respectively, and generate multiple independent detection results;

[0029] Judge whether multiple independent detection results are consistent;

[0030] If they are consistent, determine the independent detection result as the parameter detection result;

[0031] If they are inconsistent, perform anomaly voting according to the normal dictionary weight structure, abnormal dictionary weight structure, algorithm threshold weight, and multiple independent detection results in the combination selection algorithm to generate a parameter detection result.

[0032] Preferably, the method further includes:

[0033] If the parameter detection result is abnormal, determine whether to trigger a risk level warning according to the target fitting algorithm and the telemetry parameter data;

[0034] If it is determined to trigger a risk level warning, generate a threshold achievement duration according to the telemetry parameter data and the obtained expert threshold parameters through a pre-trained regression model;

[0035] Perform a risk warning according to the threshold achievement duration and the parameter detection result.

[0036] Preferably, the method further includes:

[0037] Obtain a training telemetry value list and expert threshold parameters;

[0038] Classify the training telemetry value list to generate training parameter types;

[0039] Determine the corresponding pre-trained fitting algorithm according to the training parameter type through the fitting algorithm allocation table;

[0040] If the training parameter type is a state quantity or a counter, perform threshold fitting on the pre-trained fitting algorithm to generate a threshold fitting result;

[0041] If the training parameter type is a stable pseudo-period, pseudo-period or enumeration, perform threshold fitting on the pre-trained fitting algorithm according to the expert threshold parameter to generate a threshold fitting result;

[0042] Perform threshold weight verification and threshold weight optimization according to the training telemetry value list to generate the algorithm threshold weight.

[0043] The present invention also discloses a spacecraft telemetry parameter anomaly detection device, including:

[0044] An acquisition unit for acquiring telemetry parameter data transmitted by the spacecraft;

[0045] A parameter classification unit for classifying the telemetry parameter data to generate a parameter type;

[0046] An algorithm determination unit for determining the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm allocation table;

[0047] An anomaly detection unit for performing anomaly detection on the telemetry parameter data through the threshold fitting result of the target fitting algorithm to generate a parameter detection result, and the threshold fitting result is obtained by pre-training the target fitting algorithm.

[0048] The present invention also discloses a computer-readable medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned method is implemented.

[0049] The present invention also discloses a computer device, including a memory and a processor, the memory is used for storing information including program instructions, the processor is used for controlling the execution of the program instructions, and when the processor executes the program, the above-mentioned method is implemented.

[0050] The present invention also discloses a computer program product, including computer program / instructions, and when the computer program / instructions are executed by a processor, the above-mentioned method is implemented.

[0051] The present invention obtains the telemetry parameter data transmitted by a spacecraft; classifies the telemetry parameter data to generate parameter types; determines the corresponding target fitting algorithm according to the parameter types through a preset fitting algorithm allocation table; performs anomaly detection on the telemetry parameter data through the threshold fitting result and algorithm threshold weight of the target fitting algorithm, and the threshold fitting result is obtained by pre-training the target fitting algorithm, realizing the full-process automation from threshold fitting to optimization, reducing the operation and maintenance cost, and improving the operation efficiency of the system; dynamically adjusts the threshold according to the characteristics of different categories of data, has strong adaptability, is applicable to a variety of complex scenarios, reduces the dependence on manual experience, enhances the flexibility of the system, and improves the coverage area of telemetry parameter monitoring; avoids the misjudgment or missed detection problems that may be caused by the traditional fixed threshold method by automatically optimizing the threshold parameters, and significantly improves the accuracy of anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0053] Figure 1 It is a flowchart of a method for anomaly detection of spacecraft telemetry parameters provided by an embodiment of the present invention;

[0054] Figure 2 It is a flowchart of another method for anomaly detection of spacecraft telemetry parameters provided by an embodiment of the present invention;

[0055] Figure 3 It is a schematic structural diagram of a device for anomaly detection of spacecraft telemetry parameters provided by an embodiment of the present invention;

[0056] Figure 4 It is a schematic structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0058] To facilitate the understanding of the technical solution provided in this application, the relevant content of the technical solution of this application will be described below. With the rapid development of the aerospace industry, the design of spacecraft has become increasingly complex. The telemetry parameters of the space station reach up to the order of 100,000. Manually setting the parameter anomaly detection rules is extremely cumbersome. How to effectively and conveniently monitor such a huge amount of telemetry data has become an urgent problem to be solved by the flight control center. The technical problem to be solved by this invention is to provide a method and device for analyzing anomalies in spacecraft telemetry parameters, which can realize solutions for parameter classification, threshold fitting, threshold weight optimization, detection of anomalies, risk level warning, and anomaly voting of telemetry parameters, so as to provide a mode for detecting telemetry parameters by means of data analysis.

[0059] To solve the above technical problems, this invention provides a training device and a detection device. The training device includes modules such as parameter classification, threshold fitting, and weight optimization. The detection device includes modules such as detection of anomalies, risk level warning, and anomaly voting, realizing solutions for parameter classification, threshold fitting, threshold weight optimization, detection of anomalies, risk level warning, and anomaly voting of telemetry parameters. This invention realizes the automatic training of parameter classification, threshold fitting, and threshold weight optimization through the training device for analyzing anomalies in telemetry parameters, and realizes the analysis of the whole process of training the anomaly analysis model of telemetry parameters from parameter classification, threshold fitting algorithm allocation, threshold fitting, threshold interval optimization to threshold weight optimization, expanding the analysis method of telemetry parameters; through the detection device for analyzing anomalies in telemetry parameters, it realizes the detection forms of threshold detection of anomalies, risk level warning, and anomaly voting, expanding the anomaly detection method of telemetry parameters. In particular, anomaly voting makes up for the false alarms and missed detections in the results of threshold detection of anomalies, improving the accuracy of anomaly detection.

[0060] Taking the spacecraft telemetry parameter anomaly detection device as the execution subject as an example below, the implementation process of the spacecraft telemetry parameter anomaly detection method provided in the embodiments of this invention will be described. It can be understood that the execution subject of the spacecraft telemetry parameter anomaly detection method provided in the embodiments of this invention includes but is not limited to the spacecraft telemetry parameter anomaly detection device.

[0061] Figure 1 It is a flowchart of a spacecraft telemetry parameter anomaly detection method provided in the embodiments of this invention. As Figure 1 shown, the method includes:

[0062] Step 101, obtain the telemetry parameter data transmitted by the spacecraft.

[0063] In the embodiments of this invention, the telemetry parameter data is various types of telemetry data transmitted by the spacecraft in the downlink. The downlink of the spacecraft refers to the process of transmitting data from the spacecraft in space to the ground station through wireless communication. This process is usually called the downlink.

[0064] In the embodiments of the present invention, the telemetry parameter data includes multiple telemetry values. As an alternative, the telemetry parameter data is a list of telemetry values.

[0065] Step 102: Classify the telemetry parameter data to generate parameter types.

[0066] In the embodiments of the present invention, according to the analysis of the characteristics of the telemetry parameter data, the parameter types of the telemetry parameter data include status quantities, counters, stable pseudo-periods, pseudo-periods, and enumerations. The definitions of each parameter type are shown in Table 1:

[0067] Table 1

[0068]

[0069] For the specific process of parameter classification, please refer to Step 209 for details and will not be repeated here.

[0070] Step 103: Determine the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm allocation table.

[0071] In the embodiments of the present invention, the fitting algorithm allocation table is preset, and the embodiments of the present invention do not limit this. As an alternative, the fitting algorithm allocation table is shown in Table 2:

[0072] Table 2

[0073] Parameter type Target fitting algorithm State variable State variable algorithm Counter Counter algorithm Stable pseudo-period Quartile method, percentile method, sigma criterion, Bootstrap algorithm Pseudo-period Quartile method, percentile method, sigma criterion, Bootstrap algorithm Enumeration Quartile method, percentile method, sigma criterion, Bootstrap algorithm

[0074] Step 104: Perform anomaly detection on the telemetry parameter data through the threshold fitting result of the target fitting algorithm and the algorithm threshold weight to generate a parameter detection result.

[0075] In the embodiments of the present invention, the threshold fitting result is obtained by pre-training the target fitting algorithm.

[0076] In the technical solution provided by the embodiments of the present invention, the telemetry parameter data transmitted by the spacecraft is obtained; the telemetry parameter data is classified to generate parameter types; the corresponding target fitting algorithm is determined according to the parameter type through a preset fitting algorithm allocation table; anomaly detection is performed on the telemetry parameter data through the threshold fitting result of the target fitting algorithm and the algorithm threshold weight to generate a parameter detection result, and the threshold fitting result is obtained by pre-training the target fitting algorithm, realizing the full-process automation from threshold fitting to optimization, reducing the operation and maintenance cost, and improving the operation efficiency of the system; dynamically adjusting the threshold according to the characteristics of different types of data, with strong adaptability, applicable to a variety of complex scenarios, reducing the dependence on manual experience, enhancing the flexibility of the system, and increasing the monitoring coverage area of telemetry parameters; avoiding the misjudgment or missed detection problems that may be brought by the traditional fixed threshold method by automatically optimizing the threshold parameters, and significantly improving the accuracy of anomaly detection.

[0077] Figure 2 This is a flowchart of another spacecraft telemetry parameter anomaly detection method provided by an embodiment of the present invention. As Figure 2 shown, the method includes:

[0078] Step 201: Obtain a training telemetry value list and an expert threshold parameter.

[0079] In an embodiment of the present invention, the training telemetry value list can be obtained from historical data for training the fitting algorithm. The expert threshold parameter can be obtained from authoritative materials in the industry.

[0080] Step 202: Classify the training telemetry value list to generate training parameter types.

[0081] In an embodiment of the present invention, according to the analysis of the data characteristics of the telemetry parameters, the training parameter types of the training telemetry value list include status variables, counters, stable pseudo-periods, pseudo-periods, and enumerations. The definitions of each parameter type are as shown in Table 1 of Step 102.

[0082] It should be noted that the difference between this step and Step 209 is that the data in this step is the telemetry value list used during training, and the data in Step 209 is the telemetry parameter data (telemetry value list) used during actual application. The specific process of parameter classification and the specific parameter types are the same, and will not be repeated here.

[0083] Step 203: Determine the corresponding pre-training fitting algorithm according to the training parameter type through the fitting algorithm allocation table.

[0084] In an embodiment of the present invention, the fitting algorithm allocation table is pre-set, and the present invention does not limit this. As an optional solution, the fitting algorithm allocation table is shown in Table 3:

[0085] Table 3

[0086]

[0087]

[0088] That is: if the training parameter type is a status variable, the corresponding pre-training fitting algorithm is the status variable algorithm; if the training parameter type is a counter, the corresponding pre-training fitting algorithm is the counter algorithm; if the training parameter type is a stable pseudo-period, pseudo-period, or enumeration, the corresponding pre-training fitting algorithm is one or any combination of the quartile method, percentile method, sigma criterion, and Bootstrap algorithm.

[0089] In the embodiments of the present invention, in an application scenario, for telemetry parameters of the stable pseudo-period type, pseudo-period type, and enumeration type, any one of the quartile method, percentile method, sigma criterion, and Bootstrap algorithm can be selected for allocation. The embodiments of the present invention describe the fitting process based on the default allocation of four pre-trained fitting algorithms.

[0090] Step 204: Determine whether the training parameter type is a status quantity or a counter. If so, execute Step 205; if not, execute Step 206.

[0091] In the embodiments of the present invention, if the training parameter type is a status quantity or a counter, threshold fitting of the corresponding pre-trained fitting algorithm can be performed to obtain the final threshold fitting result, and then continue to execute Step 205; if the training parameter type is stable pseudo-period, pseudo-period, or enumeration, threshold fitting and optimization of the corresponding pre-trained fitting algorithm are required to obtain the final threshold fitting result, and then continue to execute Step 206.

[0092] Step 205: Perform threshold fitting on the pre-trained fitting algorithm to generate a threshold fitting result, and then continue to execute Step 208.

[0093] In the embodiments of the present invention, if the training parameter type is a status quantity, the pre-trained fitting algorithm is a status quantity algorithm. Specifically, the fitting process of the status quantity algorithm is to remove duplicates from the list of telemetry values to obtain a threshold set, that is, the threshold fitting result.

[0094] In the embodiments of the present invention, if the training parameter type is a counter, the pre-trained fitting algorithm is a counter algorithm. Specifically, first traverse the list of training telemetry values, divide each telemetry value by the integer α, use the quotient as the frequency band of the threshold interval, and the remainder as the count value within the frequency band. Continuously update the minimum and maximum values of the frequency band and the list of count values. Calculate the threshold interval by adding the product of the frequency band multiplied by the integer α to the minimum and maximum values of the frequency band. Then calculate the growth step between the count values from the list of count values of the frequency band, and statistically select the step with the most occurrences as the threshold step; use the obtained threshold interval and threshold step as the threshold result, and put the threshold results calculated for each frequency band into a list, and this list is used as the final threshold fitting result.

[0095] It should be noted that the value of the integer α can be set according to actual needs, and the embodiments of the present invention do not limit this.

[0096] Step 206: Perform threshold fitting on the pre-trained fitting algorithm according to the expert threshold parameter to generate a threshold fitting result.

[0097] In the embodiments of the present invention, the quartile method, the percentile method, the sigma criterion, and the Bootstrap algorithm rely on expert thresholds to filter the list of telemetry values, and after filtering, a list of values a within the range of the expert thresholds is obtained. The expert threshold parameters can be obtained from authoritative reference materials; when the expert threshold parameters are empty, there is no need to filter the training list of telemetry values, and the unfiltered list of values a is obtained, that is: the training list of telemetry values; if the expert threshold parameters are not empty, it is necessary to filter the training list of telemetry values according to the range of the expert threshold parameters.

[0098] In the embodiments of the present invention, if there are multiple threshold intervals for the expert threshold parameters corresponding to the telemetry parameters, sub-lists of values that meet each interval are generated according to each threshold interval, such as sub-list of values a1, sub-list of values a2, etc. The a mentioned in the description of the fitting process of the four algorithms (quartile method, percentile method, sigma criterion, Bootstrap algorithm) can also refer to each sub-list of values. The four algorithms fit the threshold intervals according to each sub-list of values, and the obtained threshold intervals are also multiple intervals.

[0099] It should be noted that for the convenience of understanding the formula in the subsequent threshold fitting process, both the list of values after filtering the training telemetry values and the unfiltered list of training telemetry values are represented as the list of values a, which is used as the input for the subsequent threshold fitting.

[0100] The following specifically introduces the threshold fitting process of the four algorithms:

[0101] For the threshold fitting process of the quartile method: Sort the list of values a from small to large, and calculate the value x1 at the 25% position and the value x2 at the 75% position of the sorted list of values a respectively; Calculate the threshold lower limit low and the threshold upper limit high according to the following formula; Determine the threshold fitting result of the quartile method according to the threshold lower limit low and the threshold upper limit high, that is: the quartile threshold interval:

[0102] low = x1 - 1.5×(x2 - x1)

[0103] high = x2 + 1.5×(x2 - x1)

[0104] Wherein, low is the threshold lower limit, high is the threshold upper limit, x1 is the value at the 25% position of the sorted list of values a, and x2 is the value at the 75% position of the sorted list of values a.

[0105] For the threshold fitting process of the percentile method: Sort the numerical list a from smallest to largest, and calculate the value x1 at the 0.5% position and the value x2 at the 99.5% position of the sorted numerical list a respectively. Determine the threshold lower limit low as the value x1 at the 0.5% position, and determine the threshold upper limit high as the value x2 at the 99.5% position; Determine the threshold fitting result of the percentile method according to the threshold lower limit low and the threshold upper limit high, that is: the percentile threshold interval.

[0106] For the threshold fitting process of the sigma criterion: According to the telemetry data of the numerical list a, calculate the mean b and the standard deviation c of the numerical list a; Calculate the threshold lower limit low and the threshold upper limit high respectively according to the following formula; Determine the threshold fitting result of the sigma criterion according to the threshold lower limit low and the threshold upper limit high, that is: the sigma threshold interval:

[0107] low = b - 3×c

[0108] high = b + 3×c

[0109] Where, low is the threshold lower limit, high is the threshold upper limit, b is the mean of the numerical list a, and c is the standard deviation of the numerical list a.

[0110] For the threshold fitting process of the Bootstrap algorithm: Calculate the lower index index_low and the upper index index_high of the numerical list a according to the following formula; Sort the numerical list a from smallest to largest, and take the values corresponding to the lower index and the upper index as the threshold lower limit low and the threshold upper limit high respectively; Determine the threshold fitting result of the Bootstrap algorithm according to the threshold lower limit low and the threshold upper limit high, that is: the Bootstrap threshold interval:

[0111] index_low = len_a×0.05 / 2

[0112] index_high = len_a×(1 - 0.05 / 2)

[0113] low = sort_a[index_low]

[0114] high = sort_a[index_high]

[0115] Where, index_low is the lower index, index_high is the upper index, low is the threshold lower limit, and high is the threshold upper limit.

[0116] Further, optimize the threshold intervals for the quartile method, percentile method, sigma criterion, and Bootstrap algorithm respectively. If multiple threshold intervals are fitted in the algorithm, optimize the threshold intervals for each segment separately.

[0117] Specifically, for each of the above four algorithms, count the minimum and maximum values of the numerical list a, and determine whether the upper threshold is equal to the lower threshold and whether the minimum value of the numerical list a is equal to the maximum value. If both are yes, it means that all the telemetry values in the list are the same. When the telemetry values are all the same, the upper and lower limits fitted according to the provided threshold fitting formula are equal, and the threshold interval cannot be optimized, so stop adjusting the threshold; otherwise, the threshold interval can be adjusted, and the new upper and lower thresholds can be obtained according to the following formula:

[0118] new_low = bak_low - unit_len × expand_rate

[0119] new_high = bak_high + unit_len × expand_rate

[0120] Among them, unit_len is half of the threshold interval length, expand_rate is the specified multiple, new_low is the new lower threshold, bak_low is the original lower threshold, new_high is the new upper threshold, and bak_high is the original upper threshold.

[0121] It should be noted that the specified multiple can be set according to actual needs, and the embodiments of the present invention do not limit this.

[0122] Further, if the new lower threshold is less than the minimum value of the numerical list a, then expand_rate is decremented by the specified change multiple a_rate, that is: expand_rate - a_rate, to obtain the updated expand_rate; use the updated expand_rate to calculate the new lower threshold according to the above formula new_low = bak_low - unit_len × expand_rate; repeat the judgment of whether the new lower threshold is less than the minimum value of the numerical list a. If so, repeat the step of updating expand_rate until the calculated new lower threshold is greater than or equal to the minimum value of the numerical list a, and the threshold interval optimization ends.

[0123] In the embodiments of the present invention, since the lower threshold obtained at the end of the optimization is greater than or equal to the minimum value of the numerical list a, there is a false alarm anomaly when detecting telemetry data. Therefore, use the expand_rate of the penultimate time to execute the above formula new_low = bak_low - unit_len × expand_rate to obtain the optimal lower threshold and update the lower threshold.

[0124] Further, if the new upper threshold is greater than the maximum value of the numerical list a, then expand_rate is decremented by a specified change multiple a_rate, i.e., expand_rate - a_rate, to obtain the updated expand_rate; the updated expand_rate is used to calculate the new upper threshold according to the above formula new_high = bak_high + unit_len × expand_rate; repeat the judgment on whether the new upper threshold is still greater than the maximum value of the numerical list a; if so, repeat the step of updating expand_rate until the calculated new upper threshold is less than or equal to the maximum value of the numerical list a, and the threshold interval optimization ends.

[0125] In the embodiment of the present invention, since the upper threshold obtained at the end of the optimization is less than or equal to the maximum value of the numerical list a, there is a false alarm anomaly when detecting telemetry data. Therefore, the expand_rate of the penultimate time is used to execute the above formula new_high = bak_high + unit_len × expand_rate to obtain the optimal upper threshold and update the upper threshold.

[0126] The present invention first performs a preliminary positioning of the threshold interval and then optimizes the threshold interval, realizing the automatic optimization of the threshold interval. At the same time, in order to improve the accuracy of anomaly detection, an anomaly voting session is also set up to perform weighted tuning on each algorithm, and this process can be extended to the model construction and tuning of other methods.

[0127] Step 207: Perform threshold weight verification and threshold weight optimization according to the training telemetry value list to generate algorithm threshold weights.

[0128] In the embodiment of the present invention, the value range of the threshold weight of the fitting algorithm is a floating-point number in the interval [0, 1], and the larger the weight value, the more trustworthy the fitting algorithm.

[0129] In the embodiment of the present invention, during the threshold weight verification process, the quartile method, percentile method, sigma criterion, and Bootstrap algorithm are defaultly assigned the same weight initially and the training telemetry value list is used for verification. Traverse the training telemetry value list, and the traversed telemetry values are respectively input into the four fitting algorithms for verification to obtain verification results. If any of the verification results shows an abnormal state, the anomaly count is incremented by 1.

[0130] When the verification result of a part of the algorithm shows an abnormal state and the verification results of the remaining algorithms show a normal state, the threshold weights of the normal state verification results are collected into the normal dictionary, and the threshold weights of the abnormal state verification results are collected into the abnormal dictionary. The weights of the normal dictionary and the abnormal dictionary are used as the input for abnormal voting. Abnormal voting is performed. If an abnormal result is voted out, the weights of the normal dictionary and the abnormal dictionary are used as the input for threshold weight optimization, and the threshold weight optimization process is executed. After the weights are optimized, other telemetry values are traversed continuously. When the traversal of the telemetry value list ends, the abnormal rate is counted once. The abnormal rate = the number of abnormal counts / the length of the telemetry value list, and the verification execution count is incremented by 1. When the abnormal rate is not greater than the specified abnormal rate threshold or the verification execution count is not less than the specified maximum verification execution count, the threshold weight verification is stopped; otherwise, the threshold weight verification is repeated.

[0131] It should be noted that the process of abnormal voting is shown in detail in step 5111 and will not be elaborated here.

[0132] In the embodiments of the present invention, the weight structures of the normal dictionary and the abnormal dictionary are as follows:

[0133] [{0:{algorithm 1 label: algorithm 1 weight, algorithm 2 label: algorithm 2 weight}}, {1:{algorithm 3 label: algorithm 3 weight, algorithm 4 label: algorithm 4 weight}}]

[0134] Among them, 0 represents normal and 1 represents abnormal. Each telemetry value corresponds to a weight structure of a normal dictionary and an abnormal dictionary.

[0135] During the process of threshold weight verification, the weight structures of the normal dictionary and the abnormal dictionary are converted, and the converted structure is as follows:

[0136] {0:{'weight':[algorithm 1 weight, algorithm 2 weight], 'algrith':[algorithm 1 label, algorithm 2 label], 'average': average weight}, 1:{'weight':[algorithm 3 weight, algorithm 4 weight], 'algrith':[algorithm 3 label, algorithm 4 label], 'average': average weight}}

[0137] Among them, 'weight' represents weight, 'algrith' represents algorithm, 'average' represents average weight, 0 represents normal, and 1 represents abnormal.

[0138] The threshold weight optimization process is divided into five types, and the process is as follows:

[0139] (a) When there is only an abnormal situation in the converted structure. If the number M of abnormal algorithms is only 1 (one of the selected fitting algorithms is the quartile method, percentile method, sigma criterion, Bootstrap algorithm), the weight is corrected to qb; if the number M of abnormal algorithms is multiple and the weights of abnormal algorithms are not less than the threshold qa, the weight is corrected to qb.

[0140] (b) When the normal weight average value NA is equal to the abnormal weight average value MA in the converted structure. If the weight of a normal algorithm is less than the threshold qa, the algorithm weight is incremented by the weight value qc, that is: algorithm weight value + qc; if the weight of an abnormal algorithm is not less than the threshold qa, the algorithm weight is decremented by the weight value qc, that is: algorithm weight value - qc; if the weight of an abnormal algorithm is less than the threshold qa, the algorithm weight is decremented by the weight value qd.

[0141] (c) When the normal weight average value NA is not equal to the abnormal weight average value MA in the converted structure, and the number N of normal algorithms is greater than the number M of abnormal algorithms. If the weight of a normal algorithm is less than the threshold qa, the algorithm weight is incremented by the weight value qc; if the weight of a normal algorithm is not less than the threshold qa, the algorithm weight is incremented by the weight value qd; the weights of all abnormal algorithms are decremented by the weight value qc.

[0142] (d) When the normal weight average value NA is not equal to the abnormal weight average value MA in the converted structure, and the number N of normal algorithms is less than or equal to the number M of abnormal algorithms. If the weight of a normal algorithm is not greater than the threshold qa, the algorithm weight is incremented by the weight value qc; if the weight of a normal algorithm is greater than the threshold qa, the algorithm weight is incremented by the weight value qd; if the weight of an abnormal algorithm is not less than the threshold qa, the algorithm weight is decremented by the weight value qc; if the weight of an abnormal algorithm is less than the threshold qa, the algorithm weight is decremented by the weight value qd.

[0143] (e) When the normal weight average value NA is not equal to the abnormal weight average value MA in the converted structure, and the number N of normal algorithms is equal to the number M of abnormal algorithms. If the weight of a normal algorithm is less than the threshold qa, the algorithm weight is incremented by the weight value qc; the weights of all abnormal algorithms are decremented by the weight value qc.

[0144] It should be noted that the values of qa, qb, qc, and qd can be preset according to actual needs, and the embodiments of the present invention do not limit this. For the situations not mentioned in the above (a)-(e), it can be considered that the algorithm weights are appropriate and do not need to be updated.

[0145] In the embodiments of the present invention, the verified and optimized algorithm weights are determined as the final algorithm threshold weights.

[0146] Step 208, obtain the telemetry parameter data transmitted by the spacecraft.

[0147] In the embodiments of the present invention, the telemetry parameter data are various types of telemetry data transmitted from the spacecraft. The telemetry parameter data include multiple telemetry values. As an alternative, the telemetry parameter data is a list of telemetry values.

[0148] Step 209: Classify the telemetry parameter data to generate a parameter type.

[0149] In the embodiments of the present invention, step 209 specifically includes:

[0150] Step 2091: Determine whether all the telemetry values in the telemetry parameter data are integers. If they are all integers, execute step 2092; if they are not all integers, execute step 2097.

[0151] In the embodiments of the present invention, based on the parameter classification definition, if all the telemetry values in the telemetry parameter data are integers, it may be a status quantity, a counter, or an enumeration; continue to execute step 2092; if the telemetry values in the telemetry parameter data are not all integers, it may be a stable pseudo-period or a pseudo-period, and continue to execute step 2097.

[0152] It should be noted that the parameter classification definition can refer to Table 1 in step 102.

[0153] Step 2092: Determine whether the number of different telemetry values in the telemetry parameter data is less than or equal to a preset quantity threshold. If it is, execute step 2093; if not, execute step 2094.

[0154] In the embodiments of the present invention, the quantity threshold can be set according to actual requirements, and the embodiments of the present invention do not limit this.

[0155] In the embodiments of the present invention, based on the definition of the status quantity in the parameter classification definition, if the number of different telemetry values is less than or equal to the preset quantity threshold, it is determined as a status quantity, and continue to execute step 2093; if the number of different telemetry values is greater than the quantity threshold, further judgment is required, and continue to execute step 2094.

[0156] Step 2093: Determine that the parameter type is a status quantity, and this step ends.

[0157] Step 2094: Perform a trend analysis on the telemetry parameter data. If the telemetry parameter data shows an upward trend, execute step 2095; if the telemetry parameter data does not show an upward trend, execute step 2096.

[0158] In the embodiment of the present invention, the telemetry value list of the telemetry parameter data is evenly divided into two parts, the front and the back. The values at the same index positions of the two parts are compared in sequence. If the value of the latter part is greater than that of the former part, the up - count is incremented by 1; if the value of the latter part is less than that of the former part, the down - count is incremented by 1. The cumulative probability density of the binomial distribution is used to calculate the cumulative probability density of the down - count, and it is judged whether the down - count is less than the up - count and whether the cumulative probability density of the down - count is less than the specified threshold. If so, it indicates that the trend of the main body of the telemetry value list is an upward trend, that is: the telemetry parameter data is in an upward trend, and step 2095 is continued; if not, a section of data is randomly intercepted from the telemetry value list, and the up - count of the values in the intercepted data is counted. If the up - count is greater than the intercepted length multiplied by the specified multiple, it is a counter, and step 2095 is continued; if the up - count is less than or equal to the intercepted length multiplied by the specified multiple, step 2096 is continued.

[0159] The cumulative probability density function of the binomial distribution is:

[0160]

[0161] Where y is the cumulative probability density, k is the up - count, and n is all up - counts and down - counts.

[0162] It should be noted that the specified multiple can be set according to actual needs, and the embodiment of the present invention does not limit this.

[0163] Step 2095: Determine that the parameter type is a counter, and this step ends.

[0164] In the embodiment of the present invention, if the trend of the main body of the telemetry value list is an upward trend, the parameter type is a counter.

[0165] Step 2096: Determine that the parameter type is an enumeration, and this step ends.

[0166] In the embodiment of the present invention, if the trend of the main body of the telemetry value list is not an upward trend, the parameter type is an enumeration.

[0167] Step 2097: Judge whether the standard deviation of the telemetry parameter data is less than the preset standard deviation threshold. If so, execute step 2098; if not, execute step 2099.

[0168] In the embodiment of the present invention, the standard deviation threshold can be set according to actual needs, and the embodiment of the present invention does not limit this.

[0169] In the embodiment of the present invention, if the telemetry value list is not all integers, it is judged whether the standard deviation of the telemetry value list is less than the standard deviation threshold. If so, it is a stable pseudo - period, and step 2098 is continued; if not, it is a pseudo - period, and step 2099 is continued.

[0170] Step 2098: Determine that the parameter type is stable pseudo-period, and this step ends.

[0171] Step 2099: Determine that the parameter type is pseudo-period, and this step ends.

[0172] The method of the present invention classifies parameters into several categories according to the change trend and data type of historical data of telemetry parameters, automatically completes parameter classification, and then selects the algorithm corresponding to the classification to model it, realizing the function of automatic modeling of massive telemetry parameters. This classification method can further expand parameter classification to achieve refined parameter modeling. This method has strong universality and can be effectively applied to other fields of massive data processing, simplifying complex problems.

[0173] Step 210: Determine the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm allocation table.

[0174] In the embodiment of the present invention, the fitting algorithm allocation table is shown in Table 2 of Step 103 and will not be elaborated here.

[0175] The anomaly processing framework based on parameter classification learning of the present invention divides a large number of complex telemetry parameters into several categories, designs model training algorithms for different types of telemetry parameters to complete parameter modeling, and then uses the model to detect the real-time data of the parameters. Without the premise of expert knowledge, it effectively expands the number of real-time monitored telemetry parameters. This framework uses different algorithms to model and detect different types of parameters, effectively solving the problem that one algorithm cannot adapt to all parameter changes.

[0176] Step 211: Perform anomaly detection on the telemetry parameter data through the threshold fitting result and algorithm threshold weight of the target fitting algorithm to generate a parameter detection result.

[0177] In the embodiment of the present invention, the threshold fitting result is obtained by pre-training the target fitting algorithm, and the algorithm threshold weight is optimized by performing anomaly detection based on the training telemetry value list.

[0178] In the embodiment of the present invention, if the target fitting algorithm is a status quantity algorithm, the threshold fitting result is a set of thresholds. Step 211 specifically includes:

[0179] Step 2111: Perform anomaly detection on the telemetry parameter data according to the set of thresholds of the status quantity algorithm, and count the status quantity anomaly count.

[0180] In the embodiment of the present invention, the detection design of the status quantity algorithm is to determine whether each telemetry value in the telemetry parameter data exceeds the set of thresholds. If it exceeds the set of thresholds n times, the anomaly count is incremented by n, and finally the status quantity anomaly count is statistically obtained.

[0181] Step 2112: Determine whether the abnormal count of the status quantity is greater than the preset abnormal count threshold of the status quantity. If so, execute Step 2113; if not, execute Step 2114.

[0182] In the embodiments of the present invention, the abnormal count threshold of the status quantity can be set according to actual needs, and the embodiments of the present invention do not limit this.

[0183] In the embodiments of the present invention, if the abnormal count of the status quantity is greater than the preset abnormal count threshold of the status quantity, it indicates that there are too many abnormal telemetry values, and continue to execute Step 2113; if the abnormal count of the status quantity is less than or equal to the preset abnormal count threshold of the status quantity, it indicates that the number of abnormal telemetry values is within the normal range, and continue to execute Step 2114.

[0184] Step 2113: Determine that the parameter detection result is abnormal, and this step ends.

[0185] In the embodiments of the present invention, if there are too many abnormal telemetry values, determine that the parameter detection result is abnormal.

[0186] Step 2114: Determine that the parameter detection result is normal, and this step ends.

[0187] In the embodiments of the present invention, if the number of abnormal telemetry values is within the normal range, determine that the parameter detection result is normal.

[0188] In the embodiments of the present invention, if the target fitting algorithm is the counter algorithm, the threshold fitting result is the threshold interval and the threshold step size. Step 211 specifically includes:

[0189] Step 3111: Determine whether two consecutive telemetry values in the telemetry parameter data are within the threshold interval. If so, execute Step 3112; if not, execute Step 3113.

[0190] In the embodiments of the present invention, the detection design of the counter algorithm is to first determine whether two consecutive telemetry values are within the threshold interval. If two consecutive telemetry values are within the threshold interval, it indicates that the range of the telemetry value is normal, and continue to execute Step 3112; if two consecutive telemetry values exceed the threshold interval, it indicates that the range of the telemetry value is abnormal, and continue to execute Step 3113.

[0191] Step 3112: According to the threshold step size, perform abnormal detection on the change step size of the two consecutive telemetry values, count the counter abnormal count, and continue to execute Step 3114.

[0192] In the embodiments of the present invention, if it does not exceed the threshold interval, determine whether the change step size of the telemetry value is greater than the product of the threshold step size and the jump threshold (threshold step size × jump threshold). If it is greater, the abnormal count is incremented by 1, and finally the counter abnormal count is counted, and continue to execute Step 3114.

[0193] It should be noted that the threshold step size is the threshold fitting result of the counter algorithm, and the jump threshold can be set according to actual requirements. The embodiments of the present invention do not limit this.

[0194] Step 3113: Count the number of times not in the threshold interval, and determine the counted number as the counter abnormal count.

[0195] In the embodiments of the present invention, count the number of times that two consecutive telemetry values in the telemetry parameter data are not in the threshold interval. If it exceeds the threshold interval n times, the abnormal count increments by n, and finally the counter abnormal count is counted.

[0196] Step 3114: Determine whether the counter abnormal count is greater than the preset counter abnormal count threshold. If so, execute step 3115; if not, execute step 3116.

[0197] In the embodiments of the present invention, the counter abnormal count threshold can be set according to actual requirements. The embodiments of the present invention do not limit this.

[0198] In the embodiments of the present invention, if the counter abnormal count is greater than the counter abnormal count threshold, it indicates that there are too many abnormal telemetry values, and continue to execute step 3115; if the counter abnormal count is less than or equal to the counter abnormal count threshold, it indicates that the number of abnormal telemetry values is within the normal range, and continue to execute step 3116.

[0199] Step 3115: Determine that the parameter detection result is abnormal, and this step ends.

[0200] In the embodiments of the present invention, if there are too many abnormal telemetry values, determine that the parameter detection result is abnormal.

[0201] Step 3116: Determine that the parameter detection result is normal, and this step ends.

[0202] In the embodiments of the present invention, if the number of abnormal telemetry values is within the normal range, determine that the parameter detection result is normal.

[0203] In the embodiments of the present invention, if the target fitting algorithm is an independent selection algorithm, the independent selection algorithm includes one of the quartile method, percentile method, sigma criterion, and self-help clustering algorithm, the threshold fitting result is an independent selection threshold interval, and the independent selection threshold interval includes one of the quartile threshold interval, percentile threshold interval, sigma threshold interval, and self-help clustering threshold interval. Step 211 specifically includes:

[0204] Step 4111: Perform abnormal detection on the telemetry parameter data according to the independent selection threshold interval, and count the independent abnormal count.

[0205] In the embodiments of the present invention, the detection designs of the quartile method, percentile method, sigma criterion, and Bootstrap algorithm are the same; by independently selecting the independent selection threshold interval (one of the quartile threshold interval, percentile threshold interval, sigma threshold interval, and self-clustering threshold interval) of the algorithm (one of the quartile method, percentile method, sigma criterion, and Bootstrap algorithm), the telemetry value of the telemetry parameter data is detected for anomalies, and it is determined whether the telemetry value exceeds the fitted threshold interval. If it exceeds the threshold interval n times, the anomaly count is incremented by n, and finally the independent anomaly count is statistically obtained.

[0206] Step 4112: Determine whether the independent anomaly count is greater than the preset independent anomaly count threshold. If so, execute Step 4113; if not, execute Step 4114.

[0207] In the embodiments of the present invention, the independent anomaly count threshold can be set according to actual requirements, and the embodiments of the present invention do not limit this.

[0208] In the embodiments of the present invention, if the independent anomaly count is greater than the preset independent anomaly count threshold, it indicates that there are too many abnormal telemetry values, and continue to execute Step 4113; if the independent anomaly count is less than or equal to the preset independent anomaly count threshold, it indicates that the number of abnormal telemetry values is within the normal range, and continue to execute Step 4114.

[0209] Step 4113: Determine that the parameter detection result is abnormal, and this step ends.

[0210] In the embodiments of the present invention, if there are too many abnormal telemetry values, determine that the parameter detection result is abnormal.

[0211] Step 4114: Determine that the parameter detection result is normal, and this step ends.

[0212] In the embodiments of the present invention, if the number of abnormal telemetry values is within the normal range, determine that the parameter detection result is normal.

[0213] In the embodiments of the present invention, if the target fitting algorithm is a combined selection algorithm, the combined selection algorithm includes any combination of the quartile method, percentile method, sigma criterion, and self-clustering algorithm, and the threshold fitting result is a combined selection threshold interval, and the combined selection threshold interval includes any combination of the quartile threshold interval, percentile threshold interval, sigma threshold interval, and self-clustering threshold interval. Step 211 specifically includes:

[0214] Step 5111: According to the combined selection threshold interval, perform anomaly detection on the telemetry parameter data respectively, and statistically obtain multiple independent anomaly counts.

[0215] In the embodiments of the present invention, the detection designs of the quartile method, the percentile method, the sigma criterion, and the Bootstrap algorithm are the same. The four algorithms respectively perform anomaly detection on the telemetry parameter data. The detection design of any algorithm is to determine whether the telemetry value exceeds all the fitted threshold intervals. If it exceeds the threshold interval n times, the anomaly count is incremented by n, and finally the independent anomaly count is statistically obtained.

[0216] Step 5112: According to the preset independent anomaly count threshold, respectively determine the anomaly results for multiple independent anomaly counts, and generate multiple independent detection results.

[0217] In the embodiments of the present invention, the independent anomaly count threshold can be set according to actual requirements, and the embodiments of the present invention do not limit this. It should be noted that the independent anomaly count thresholds corresponding to the quartile method, the percentile method, the sigma criterion, and the Bootstrap algorithm are the same.

[0218] In the embodiments of the present invention, if the independent anomaly count is greater than the preset independent anomaly count threshold, it indicates that there are too many abnormal telemetry values, and step 4113 is continued; if the independent anomaly count is less than or equal to the preset independent anomaly count threshold, it indicates that the number of abnormal telemetry values is within the normal range, and step 4114 is continued.

[0219] Step 5113: Determine whether multiple independent detection results are consistent. If they are consistent, execute step 5114; if they are inconsistent, execute step 5115.

[0220] In the embodiments of the present invention, if multiple independent detection results are consistent, it indicates that there is no controversy in the detection results of each algorithm, and step 5114 is continued; if multiple independent detection results are inconsistent, it indicates that there are differences in the detection results of each algorithm, and step 5115 is continued.

[0221] Step 5114: Determine the independent detection result as the parameter detection result, and this step ends.

[0222] In the embodiments of the present invention, if there is no controversy in the detection results of each algorithm, the independent detection result is determined as the parameter detection result.

[0223] Step 5115: Perform anomaly voting according to the normal dictionary weight structure, abnormal dictionary weight structure, algorithm threshold weight in the combination selection algorithm, and multiple independent detection results to generate the parameter detection result, and this step ends.

[0224] In the embodiments of the present invention, if there are differences in the detection results of each algorithm, it is necessary to perform anomaly voting on each result. Specifically, the weight structures of the normal dictionary and the abnormal dictionary are converted, and the conversion structure is as follows:

[0225] {0: {Weighted average: [Algorithm 1 label, Algorithm 2 label]}, 1: {Weighted average: [Algorithm 3 label, Algorithm 4 label]}}

[0226] Among them, 0 represents normal and 1 represents abnormal.

[0227] Before the abnormal vote, count the number N of detected normal fitting algorithms, the average value NA of normal weights, the number M of detected abnormal fitting algorithms, and the average value MA of abnormal weights respectively. And the abnormal voting process is divided into four types, and the process is as follows:

[0228] (a) When NA is equal to MA, if N is greater than M, the voting result is normal; if N is not greater than M, the voting result is abnormal.

[0229] (b) When NA is not equal to MA and N is greater than M, if NA is not less than the threshold pa, the voting result is normal; if NA is less than the threshold pa and MA is not less than the threshold pa, the voting result is abnormal; if both NA and MA are less than the threshold pa, the voting result is normal.

[0230] (c) When NA is not equal to MA and N is less than M, if MA is not less than the threshold pa, the voting result is abnormal; if MA is less than the threshold pa and NA is not less than the threshold pa, the voting result is normal; if both NA and MA are less than the threshold pa, the voting result is abnormal.

[0231] (d) When NA is not equal to MA and N is equal to M, if NA is greater than MA, the voting result is normal; if NA is not greater than MA, the voting result is abnormal.

[0232] It should be noted that the value of pa can be set in advance according to actual needs, and the embodiments of the present invention do not limit this.

[0233] In the embodiments of the present invention, the voting result obtained from the abnormal vote is determined as the parameter detection result.

[0234] Since the learning algorithm summarizes the variation law of parameter historical data, misjudgment may occur during actual operation. Therefore, the present invention sets an abnormal count threshold for the data in a certain detection window. When the abnormal count in this window exceeds the threshold, an abnormality is reported, and this threshold can effectively reduce false alarms.

[0235] Step 212, determine whether the parameter detection result is abnormal. If so, execute step 213; if not, the process ends.

[0236] In the embodiments of the present invention, if the parameter detection result is abnormal, it is necessary to determine whether to trigger a risk level warning, and continue to execute step 213; if the parameter detection result is normal, it indicates that there is no need to perform a risk level warning, and the process ends.

[0237] Step 213: Determine whether to trigger a risk level warning based on the target fitting algorithm and telemetry parameter data. If it is determined to trigger a risk level warning, execute Step 214; if it is determined not to trigger a risk level warning, the process ends.

[0238] In the embodiments of the present invention, the risk level warning trigger conditions include:

[0239] (a) There are expert threshold parameters in the telemetry parameters.

[0240] (b) The threshold intervals for any anomaly when using the quartile method, percentile method, sigma criterion, and Bootstrap algorithm to detect anomalies.

[0241] (c) The telemetry value exceeds the anomaly threshold interval and does not exceed the expert threshold parameter.

[0242] (d) Accumulate the warning analysis of the last 10 telemetry data.

[0243] It should be noted that only when all four conditions (a)-(d) are met can the risk level warning be triggered.

[0244] In the embodiments of the present invention, based on the expert threshold parameters of the telemetry parameters, the algorithm fitting threshold intervals for detecting anomalies, and the accumulated warning analysis of the last n telemetry data, since there are one or more segments of threshold parameters for the expert threshold parameters and one or more segments of intervals for the threshold intervals, first obtain the minimum and maximum values of the expert threshold parameters where the most recent telemetry data is located, and determine whether the telemetry value is within any threshold interval; if the telemetry value is within any threshold interval, stop the risk level warning; if it is not within all threshold intervals, trigger the risk level warning, and it is necessary to predict the duration when the telemetry value of the telemetry parameter reaches the expert threshold parameter.

[0245] Step 214: Generate the threshold achievement duration based on the telemetry parameter data and the obtained expert threshold parameters through a pre-trained regression model.

[0246] In the embodiments of the present invention, the regression model is trained using the linear regression equation y1 = kx1 + b with the last n telemetry parameter data, where x1 is the time list [T1, T2, T3,..., T10] of n moments of the telemetry parameter data, y1 is the telemetry value list [V1, V2, V3,..., V10] of n moments of the telemetry parameter data, k is the slope of the regression model, and b is the intercept of the regression model; and then according to the method of predicting the duration when the telemetry value reaches the expert threshold parameter x2 = (y2 - b) / k, where x2 is the threshold achievement duration for predicting the telemetry value to reach the expert threshold parameter, y2 is the telemetry value reaching the expert threshold, k is the slope of the regression model, and b is the intercept of the regression model.

[0247] Further, according to the telemetry parameter data and the threshold interval of the algorithm for detecting anomalies, the change amplitude is calculated. Here, there are two cases: the telemetry value is less than the lower threshold of the threshold interval and the telemetry value is greater than the upper threshold of the threshold interval:

[0248] When the telemetry value is less than the lower threshold of the threshold interval, it is calculated by y1 = k / (a1 - b1) to obtain the change amplitude, where k is the regression model slope, a1 is the lower limit of the expert threshold parameter, b1 is the lower threshold, and y1 is the change amplitude.

[0249] When the telemetry value is greater than the upper threshold of the threshold interval, it is calculated by y2 = k / (a2 - b2) to obtain the change amplitude, where k is the regression model slope, a2 is the upper limit of the expert threshold parameter, b2 is the upper threshold, and y2 is the change amplitude.

[0250] The present invention combines the data analysis model with expert knowledge. Based on the parameter value sequence that recently exceeds the data analysis model, prediction algorithms such as regression are used to achieve the anomaly warning that the spacecraft telemetry parameters reach the fault boundary set by experts. Without the data analysis model, there may be frequent warnings due to data fluctuations; without expert knowledge, the standard for predicting anomalies cannot be given.

[0251] Step 215, perform risk warning according to the threshold achievement duration and the parameter detection result.

[0252] In the embodiment of the present invention, according to the preset duration-risk mapping relationship, the risk level is determined according to the threshold achievement duration; risk warning is performed according to the risk level and the parameter detection result, and the risk warning message of the risk warning includes but is not limited to the change amplitude, the threshold achievement duration, the risk level, and the parameter detection result.

[0253] It should be noted that the duration-risk mapping relationship is the mapping relationship between the threshold achievement duration and the risk level, and the duration-risk mapping relationship can be set according to actual needs, and the embodiment of the present invention does not limit this. As an optional solution, when the threshold achievement duration is within 1 hour, the corresponding risk level is the high risk level; when the threshold achievement duration is within 1 day, the corresponding risk level is the medium risk level; and the corresponding risk level for the remaining threshold achievement durations is the low risk level.

[0254] Next, a specific embodiment is used to detail the practical application of the spacecraft telemetry parameter anomaly detection method provided by the present invention:

[0255] Next, taking a spacecraft as an example, the steps of this method are described:

[0256] Step 1: Select training settings such as the telemetry parameter DCNH1677, the start time of the telemetry data 2023-07-04T00:00:00, and the end time 2023-07-06T00:00:00, and initiate training.

[0257] Step 2: Read the telemetry data corresponding to the first step of DCNH1677 from the collected data, automatically input it into the training device of the algorithm, analyze that the value range of the telemetry data is [-19.79242979, 5.16483516]. Since there are floating-point data in the data and the standard deviation within the data is greater than 0.8, the parameter classification result is a pseudo-periodic quantity.

[0258] Step 3: Due to the parameter classification result in the second step, allocate four fitting algorithms: the quartile method, the percentile method, the sigma criterion, and the Bootstrap algorithm. After the threshold fitting is optimized through the threshold interval, the fitting threshold interval of the quartile method is [-20.51526250000002, 5.741147740000017] with a weight of 0.8, the fitting threshold interval of the percentile method is [-20.664224663000013, 6.236874241000011] with a weight of 0.8, the fitting threshold interval of the sigma criterion is [-20.854476553550686, 6.064195268560859] with a weight of 0.8, and the fitting threshold interval of the Bootstrap algorithm is [-19.85347985500001, 6.11477411900001] with a weight of 0.8. Train the model and store the model.

[0259] Step 4: Select the model of DCNH1677 to initiate detection, load the model, read the cached telemetry data, and automatically input it into the detection device of the algorithm. When the telemetry value of the telemetry data is -20.0, after detection, it is abnormal. The quartile method determines it as normal, the percentile method determines it as normal, the sigma criterion determines it as normal, and the Bootstrap algorithm determines it as abnormal.

[0260] Step 5: Since only the Bootstrap algorithm determines it as abnormal in Step 4, only the threshold interval of the Bootstrap algorithm is used for risk level warning. According to the expert threshold [-50, 50] of the DCNH1677 parameter, and the accumulated recent 10 telemetry data [(T1, V1), (T2, V2), ……, (T10, V10)], analyze the warning level and the duration when the telemetry value of the predicted telemetry parameter reaches the expert threshold.

[0261] Step 6: Since the determination results of the four algorithms in Step 4 are inconsistent, an abnormal vote is required. Since the average weight of the determined abnormal is the same as the average weight of the normal, and the number of fitting algorithms that determine normal is greater than the number of fitting algorithms that determine abnormal, the final abnormal vote determines it as normal.

[0262] Through the present invention, the abnormal analysis of spacecraft telemetry parameters can be realized, and each telemetry parameter can be analyzed in the way of data analysis. All telemetry parameters can be analyzed according to data characteristics, reliable categories can be classified, reliable thresholds and weights can be fitted, and reliable methods can be adopted to detect the abnormality of each telemetry parameter.

[0263] It should be noted that in the technical solution of this application, the acquisition, storage, use, processing, etc. of data all comply with the relevant regulations of laws and regulations. The user information in the embodiments of this application is obtained through legal and compliant channels, and the acquisition, storage, use, processing, etc. of user information have obtained the authorization and consent of the customer.

[0264] It should be noted that the information collected in this application is information and data authorized by the user or fully authorized by all parties. Moreover, the processing of relevant data, such as collection, storage, use, processing, transmission, provision, disclosure, and application, all comply with the relevant laws, regulations, and standards of relevant countries and regions, necessary confidentiality measures are taken, it does not violate public order and good customs, and a corresponding operation entry is provided for users to choose to authorize or refuse.

[0265] It should be noted that the technical solution provided in this application provides a corresponding operation entry for users to choose to agree or refuse the result of automated decision-making; if the user chooses to refuse, the expert decision-making process will be entered.

[0266] In the technical solution of the spacecraft telemetry parameter abnormality detection method provided by the embodiment of the present invention, the telemetry parameter data transmitted by the spacecraft is acquired; the telemetry parameter data is classified to generate a parameter type; through a preset fitting algorithm allocation table, according to the parameter type, the corresponding target fitting algorithm is determined; through the threshold fitting result and algorithm threshold weight of the target fitting algorithm, the telemetry parameter data is detected for abnormality to generate a parameter detection result. The threshold fitting result is obtained by pre-training the target fitting algorithm, realizing the full-process automation from threshold fitting to optimization, reducing the operation and maintenance cost, and improving the operation efficiency of the system; dynamically adjusting the threshold according to the characteristics of different categories of data, with strong adaptability, suitable for a variety of complex scenarios, reducing the dependence on manual experience, enhancing the flexibility of the system, and increasing the monitoring coverage area of telemetry parameters; by automatically optimizing the threshold parameters, the problems of false judgment or missed detection that may be brought by the traditional fixed threshold method are avoided, and the accuracy of abnormality detection is significantly improved.

[0267] Figure 3 It is a schematic structural diagram of a spacecraft telemetry parameter abnormality detection device provided by an embodiment of the present invention. This device is used to execute the above-mentioned spacecraft telemetry parameter abnormality detection method, as Figure 3As shown in the figure, the device includes: a telemetry parameter data acquisition unit 11, a parameter classification unit 12, an algorithm determination unit 13, and an anomaly detection unit 14.

[0268] The telemetry parameter data acquisition unit 11 is used to acquire the telemetry parameter data transmitted by the spacecraft.

[0269] The parameter classification unit 12 is used to classify the telemetry parameter data to generate parameter types.

[0270] The algorithm determination unit 13 is used to determine the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm distribution table.

[0271] The anomaly detection unit 14 is used to perform anomaly detection on the telemetry parameter data through the threshold fitting result of the target fitting algorithm to generate a parameter detection result. The threshold fitting result is obtained by pre-training the target fitting algorithm.

[0272] In an embodiment of the present invention, the parameter classification unit 12 is specifically used to determine whether the telemetry values in the telemetry parameter data are all integers; if they are all integers, determine whether the number of different telemetry values in the telemetry parameter data is less than or equal to a preset number threshold; if so, determine that the parameter type is a status quantity; if not, perform a trend analysis on the telemetry parameter data. If the telemetry parameter data is on an upward trend, determine that the parameter type is a counter; if the telemetry parameter data is not on an upward trend, determine that the parameter type is an enumeration; if they are not all integers, determine whether the standard deviation of the telemetry parameter data is less than a preset standard deviation threshold; if so, determine that the parameter type is a stable pseudo-period; if not, determine that the parameter type is a pseudo-period.

[0273] In an embodiment of the present invention, the target fitting algorithm is an independent selection algorithm. The independent selection algorithm includes one of the quartile method, the percentile method, the sigma criterion, and the self-organizing clustering algorithm. The threshold fitting result is an independent selection threshold interval. The independent selection threshold interval includes one of the quartile threshold interval, the percentile threshold interval, the sigma threshold interval, and the self-organizing clustering threshold interval. The anomaly detection unit 14 is specifically used to perform anomaly detection on the telemetry parameter data according to the independent selection threshold interval, and count the independent anomaly count; determine whether the independent anomaly count is greater than a preset independent anomaly count threshold; if so, determine that the parameter detection result is abnormal; if not, determine that the parameter detection result is normal.

[0274] In an embodiment of the present invention, the target fitting algorithm is a combination selection algorithm, and the combination selection algorithm includes any combination of the quartile method, the percentile method, the sigma criterion, and the bootstrap clustering algorithm. The threshold fitting result is a combination selection threshold interval, and the combination selection threshold interval includes any combination of the quartile threshold interval, the percentile threshold interval, the sigma threshold interval, and the bootstrap clustering threshold interval. The anomaly detection unit 14 is specifically configured to perform anomaly detection on the telemetry parameter data respectively according to the combination selection threshold interval, and count multiple independent anomaly counts; perform anomaly result determination on the multiple independent anomaly counts respectively according to a preset independent anomaly count threshold, and generate multiple independent detection results; determine whether the multiple independent detection results are consistent; if they are consistent, determine the independent detection result as the parameter detection result; if they are inconsistent, perform anomaly voting according to the normal dictionary weight structure, the abnormal dictionary weight structure, the algorithm threshold weight, and the multiple independent detection results in the combination selection algorithm to generate the parameter detection result.

[0275] In an embodiment of the present invention, the method further includes: a warning trigger determination unit 15, a duration generation unit 16, and a risk warning unit 17.

[0276] The warning trigger determination unit 15 is configured to, if the parameter detection result is abnormal, perform risk level warning trigger determination according to the target fitting algorithm and the telemetry parameter data.

[0277] The duration generation unit 16 is configured to, if it is determined to trigger a risk level warning, generate a threshold achievement duration according to the telemetry parameter data and the obtained expert threshold parameters through a pre-trained regression model.

[0278] The risk warning unit 17 is configured to perform risk warning according to the threshold achievement duration and the parameter detection result.

[0279] In an embodiment of the present invention, the method further includes: a training parameter acquisition unit 18, a training parameter classification unit 19, a fitting algorithm determination unit 20, a first threshold fitting unit 21, a second threshold fitting unit 22, and a weight verification and optimization unit 23.

[0280] The training parameter acquisition unit 18 is configured to acquire a training telemetry value list and expert threshold parameters.

[0281] The training parameter classification unit 19 is configured to classify the training telemetry value list to generate a training parameter type.

[0282] The fitting algorithm determination unit 20 is configured to determine a corresponding pre-trained fitting algorithm according to the training parameter type through a fitting algorithm allocation table.

[0283] The first threshold fitting unit 21 is configured to, if the training parameter type is a status quantity or a counter, perform threshold fitting on the pre-trained fitting algorithm to generate a threshold fitting result.

[0284] The second threshold fitting unit 22 is configured to perform threshold fitting on the pre-training fitting algorithm according to the expert threshold parameter if the training parameter type is stable pseudo-period, pseudo-period, or enumeration, and generate a threshold fitting result.

[0285] The weight verification and optimization unit 23 is configured to perform threshold weight verification and threshold weight optimization according to the training telemetry value list, and generate an algorithm threshold weight.

[0286] In the solution of the embodiment of the present invention, telemetry parameter data transmitted by the spacecraft is acquired; the telemetry parameter data is classified to generate a parameter type; through a preset fitting algorithm allocation table, the corresponding target fitting algorithm is determined according to the parameter type; through the threshold fitting result of the target fitting algorithm and the algorithm threshold weight, the telemetry parameter data is subjected to anomaly detection to generate a parameter detection result. The threshold fitting result is obtained by pre-training the target fitting algorithm, realizing the full-process automation from threshold fitting to optimization, reducing the operation and maintenance cost, and improving the operation efficiency of the system; dynamically adjusting the threshold according to the characteristics of different categories of data, with strong adaptability, suitable for a variety of complex scenarios, reducing the dependence on manual experience, enhancing the system flexibility, and increasing the coverage area of telemetry parameter monitoring; by automatically optimizing the threshold parameter, avoiding the misjudgment or missed detection problems that may be brought by the traditional fixed threshold method, and significantly improving the accuracy of anomaly detection.

[0287] The system, device, module, or unit illustrated in the above embodiments may be specifically implemented by a computer chip or an entity, or by a product with a certain function. A typical implementation device is a computer device. Specifically, the computer device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0288] The embodiment of the present invention provides a computer device, including a memory and a processor. The memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions. When the program instructions are loaded and executed by the processor, the steps of the above embodiment of the method for anomaly detection of spacecraft telemetry parameters are implemented. For specific descriptions, reference may be made to the above embodiment of the method for anomaly detection of spacecraft telemetry parameters.

[0289] Next, refer to Figure 4 , which shows a schematic structural diagram of a computer device 600 suitable for implementing the embodiments of the present application. As Figure 4As shown, computer device 600 includes a central processing unit (CPU) 601, which can perform various appropriate operations and processes according to programs stored in a read-only memory (ROM) 602 or programs loaded from a storage section 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the computer device 600 are also stored. The CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0290] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 610 as needed so that a computer program read from it can be installed in the storage section 608 as needed.

[0291] Specifically, according to an embodiment of the present invention, the processes described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product that includes a computer program tangibly embodied on a machine-readable medium, the computer program including program code for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 609 and / or installed from the removable medium 611.

[0292] Computer-readable media include both permanent and non-permanent, removable and non-removable media implemented by any method or technology for information storage. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transitory media that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0293] For convenience of description, the above-described apparatus is described by functionally dividing it into various units. Of course, when implementing the present application, the functions of each unit can be implemented in one or more software and / or hardware.

[0294] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0295] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0296] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one process or multiple processes and / or blocks. Figure 1 one process or multiple processes and / or blocks Figure 1 steps for implementing the functions specified in one block or multiple blocks.

[0297] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, commodity or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, commodity or device comprising the element.

[0298] In the technical solution of this application, the acquisition, storage, use, processing, etc. of data all comply with the relevant provisions of national laws and regulations. It should be noted that in the embodiments of this application, some industry-existing solutions such as certain software, components, models, etc. may be mentioned. They should be regarded as exemplary, and their purpose is only to illustrate the feasibility in the implementation of the technical solution of this application, but it does not mean that the applicant has already or necessarily used this solution.

[0299] Those skilled in the art should understand that the embodiments of this application can be provided as a method, a system or a computer program product. Therefore, this application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0300] This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0301] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, reference can be made to the description of the method embodiment.

[0302] The above description is only for the embodiments of the present application and is not intended to limit the present application. For those skilled in the art, various modifications and changes can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for detecting anomalies in spacecraft telemetry parameters, characterized in that, The method includes: Obtaining the telemetry parameter data transmitted by the spacecraft; Classifying the telemetry parameter data to generate parameter types; Determining the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm allocation table; Performing anomaly detection on the telemetry parameter data through the threshold fitting result and the algorithm threshold weight of the target fitting algorithm to generate a parameter detection result, where the threshold fitting result is obtained by pre-training the target fitting algorithm.

2. The method for detecting abnormal spacecraft telemetry parameters according to claim 1, wherein The classifying the telemetry parameter data to generate parameter types includes: Judging whether the telemetry values in the telemetry parameter data are all integers; If they are all integers, judging whether the number of different telemetry values in the telemetry parameter data is less than or equal to a preset number threshold; If so, determining that the parameter type is a status quantity; If not, performing trend analysis on the telemetry parameter data. If the telemetry parameter data is in an upward trend, determining that the parameter type is a counter; If the telemetry parameter data is not in an upward trend, determining that the parameter type is an enumeration; If they are not all integers, judging whether the standard deviation of the telemetry parameter data is less than a preset standard deviation threshold; If so, determining that the parameter type is a stable pseudo-period; If not, determining that the parameter type is a pseudo-period.

3. The spacecraft telemetry parameter anomaly detection method according to claim 1, characterized in that The target fitting algorithm is an independent selection algorithm, and the independent selection algorithm includes one of the quartile method, the percentile method, the sigma criterion, and the self-organizing map algorithm. The threshold fitting result is an independent selection threshold interval, and the independent selection threshold interval includes one of the quartile threshold interval, the percentile threshold interval, the sigma threshold interval, and the self-organizing map threshold interval; The performing anomaly detection on the telemetry parameter data through the threshold fitting result and the algorithm threshold weight of the target fitting algorithm to generate a parameter detection result includes: Performing anomaly detection on the telemetry parameter data according to the independent selection threshold interval and counting the independent anomaly count; Judging whether the independent anomaly count is greater than a preset independent anomaly count threshold; If so, determining that the parameter detection result is abnormal; If not, determining that the parameter detection result is normal.

4. The spacecraft telemetry parameter anomaly detection method according to claim 1, characterized in that The target fitting algorithm is a combined selection algorithm, and the combined selection algorithm includes any combination of the quartile method, the percentile method, the sigma criterion, and the self-organizing map algorithm. The threshold fitting result is a combined selection threshold interval, and the combined selection threshold interval includes any combination of the quartile threshold interval, the percentile threshold interval, the sigma threshold interval, and the self-organizing map threshold interval; The performing anomaly detection on the telemetry parameter data through the threshold fitting result and the algorithm threshold weight of the target fitting algorithm to generate a parameter detection result includes: Performing anomaly detection on the telemetry parameter data respectively according to the combined selection threshold interval and counting multiple independent anomaly counts; Judging the anomaly results of multiple independent anomaly counts respectively according to a preset independent anomaly count threshold to generate multiple independent detection results; Judging whether the multiple independent detection results are consistent; If they are consistent, determining the independent detection result as the parameter detection result; If they are inconsistent, perform an abnormal vote based on the normal dictionary weight structure, abnormal dictionary weight structure, algorithm threshold weight, and multiple independent detection results in the combination selection algorithm to generate the parameter detection result.

5. The spacecraft telemetry parameter anomaly detection method according to claim 1, characterized in that The method further includes: If the parameter detection result is abnormal, determine whether to trigger a risk level warning according to the target fitting algorithm and the telemetry parameter data; If it is determined to trigger a risk level warning, generate the threshold achievement duration according to the telemetry parameter data and the obtained expert threshold parameters through a pre-trained regression model; Perform a risk warning according to the threshold achievement duration and the parameter detection result.

6. The method for detecting abnormal spacecraft telemetry parameters according to claim 1, wherein The method further includes: Obtain a list of training telemetry values and expert threshold parameters; Classify the list of training telemetry values to generate training parameter types; Determine the corresponding pre-trained fitting algorithm according to the training parameter type through the fitting algorithm allocation table; If the training parameter type is a status quantity or a counter, perform threshold fitting on the pre-trained fitting algorithm to generate a threshold fitting result; If the training parameter type is a stable pseudo-period, pseudo-period, or enumeration, perform threshold fitting on the pre-trained fitting algorithm according to the expert threshold parameters to generate a threshold fitting result; Perform threshold weight verification and threshold weight optimization according to the list of training telemetry values to generate the algorithm threshold weight.

7. A spacecraft telemetry parameter anomaly detection device, characterized in that, The device includes: A telemetry parameter data acquisition unit, configured to acquire telemetry parameter data transmitted by a spacecraft; A parameter classification unit, configured to classify the telemetry parameter data to generate parameter types; An algorithm determination unit, configured to determine the corresponding target fitting algorithm according to the parameter type through a preset fitting algorithm allocation table; An abnormal detection unit, configured to perform abnormal detection on the telemetry parameter data through the threshold fitting result of the target fitting algorithm to generate a parameter detection result, where the threshold fitting result is obtained by pre-training the target fitting algorithm.

8. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the spacecraft telemetry parameter abnormal detection method according to any one of claims 1 to 6.

9. A computer device, comprising a memory and a processor, the memory being used for storing information including program instructions, and the processor being used for controlling the execution of the program instructions, characterized in that, When the program instructions are loaded and executed by a processor, it implements the spacecraft telemetry parameter abnormal detection method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by a processor, it implements the spacecraft telemetry parameter abnormal detection method according to any one of claims 1 to 6.

Images