Evaluation decision-making method after spacecraft impact event

Abstract

The invention discloses an evaluation decision method after a spacecraft impact event. The method comprises the following steps: receiving sensing data of a spacecraft sensor; ontology reasoning is carried out on the induction data to obtain a first decision result; inputting the sensing data into the trained large language model, and outputting to obtain a second decision result; and finally, fusing the first decision result and the second decision result to obtain a final decision result. By effectively fusing the two decision results, the optimal decision result is automatically output, and the flight safety of the spacecraft and the life of the astronaut are effectively guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to a spacecraft post-impact event evaluation decision method, belonging to the technical field of spacecraft health management. BACKGROUND

[0002] Since the first artificial satellite Sputnik-1 was sent into space to the end of 2020, humans have carried out about 6000 space launch activities, a total of about 10680 artificial satellites have been sent into the Earth's orbit, among which there have been 560 on-orbit spacecraft explosion or disintegration events, resulting in space debris, i.e. space debris. Space debris is mainly lost satellites, rocket body debris and rocket ejecta, etc. The size of space debris is very large, from microns to meters. By the end of 2021, there were about 23,000 space debris with a diameter of more than 10 cm, about 750,000 space debris with a diameter of 1 cm to 10 cm, about 100 million space debris with a diameter of 1 mm to 1 cm, and hundreds of billions of space debris with a diameter of less than 1 mm. They are captured by the Earth's gravity like "nebula", suspended around the Earth, and in the next 50 years with the large-scale low-orbit constellation plan and the explosive growth of the number of launches, it is estimated that the number will increase by 10% per year, and more and more space debris will be left in the Earth's orbit, threatening the safety of spacecraft.

[0003] By 2009, according to the detection results of the U.S. "Space Surveillance Network" (SSN), there were about 16300 objects on the surface of the planet recorded in the existing network, and a large number of small and micro debris were lacking of monitoring. Impact can damage spacecraft, and currently there are 800 active satellites that have lost contact due to debris collision, and spacecraft safety is severely threatened. Although in the past decade, about 6800 on-orbit debris have been recorded, by 2020 the space situational awareness system (SSA) still cannot track an estimated 300,000-500,000 1-10 cm on-orbit debris, NASA predicts that space debris will reach a critical condition in 200 years, triggering near-earth orbit debris cascade collision reaction.

[0004] Space debris mainly operates in low-orbit (LEO), medium-orbit (MEO) and high-orbit (GEO) regions. Among them, the number of space debris in the LEO region is the largest, because human space activities in the LEO region are the most frequent, and most artificial satellites are concentrated in the LEO region. Since the main structure of artificial satellites is mostly aluminum / aluminum alloy and composite materials, the main component of space debris is also aluminum / aluminum alloy. When space debris hits the surface of a spacecraft, it will produce very high pressure, which may cause the spacecraft surface to have pits, produce debris clouds, or even penetrate the spacecraft surface and cause the spacecraft to disintegrate.

[0005] The debris cloud generated by the impact will damage the normal operation of the internal components of the spacecraft, posing a great threat to the flight safety of the spacecraft and the life of the astronauts. For China's manned space engineering "three-step strategy", spacecraft such as space laboratories and large space stations with longer running time are more likely to be hit by space debris, such as the US space station mechanical arm and the Hubble telescope shown in the impact. Figure 1 Therefore, the monitoring and protection research on space debris impact is crucial to the safety of the space field. SUMMARY

[0006] The technical problem solved by the present application is that there is a lack of means to effectively monitor and make targeted decisions for space debris impact in the prior art. A spacecraft impact event evaluation decision method is proposed.

[0007] The present application solves the above technical problems by the following technical solutions:

[0008] A spacecraft impact event evaluation decision method, comprising:

[0009] Receiving the sensing data returned by the spacecraft sensor after the impact event;

[0010] Performing ontology reasoning on the obtained sensing data to obtain a first decision result;

[0011] Inputting the sensing data into a trained spacecraft evaluation decision model to obtain an output second decision result;

[0012] Fusing the first decision result and the second decision result to obtain a final decision result.

[0013] The method for performing ontology reasoning on the sensing data and obtaining the first decision result is:

[0014] Constructing an on-orbit impact event ontology knowledge base and an on-orbit impact event ontology rule base;

[0015] Constructing an ontology reasoning model according to the on-orbit impact event ontology knowledge base and the on-orbit impact event ontology rule base;

[0016] Performing rule matching and logical deduction on the sensing data according to the ontology reasoning model to obtain the first decision result.

[0017] In the ontology reasoning model, the on-orbit impact event ontology knowledge base is used for logical deduction on the sensing data, and the method for logical deduction is:

[0018] After the inductive data loading is completed, the space structure ontology related to the current task scene is obtained, the influence of the impact damage on the function of the space structure ontology is analyzed, the corresponding emergency measures are determined according to the influence, the emergency measures are used to formulate an emergency scheme, and the execution emergency operation flow of the spacecraft under a specific impact event is deduced;

[0019] The information obtained after the inductive data loading is completed is used as the logical deduction result.

[0020] In the ontology reasoning model, the on-orbit impact event ontology rule base is used for rule matching of the inductive data, and the rule matching method is as follows:

[0021] After the inductive data loading is completed, the object attribute and the data attribute of the impact event are obtained; the object attribute is used to describe the on-orbit impact event category of the spacecraft, and the data attribute is used to define the association between the on-orbit impact event category of the spacecraft and the inductive data;

[0022] According to the object attribute and the data attribute as the rule matching result, an instance is created, and the instance creation result and the logical deduction result are used as the first decision result.

[0023] The spacecraft evaluation decision model adopts a commercially available large language model that has been trained, and after the inductive data is input, the output result composed of the historical experience data output by the spacecraft evaluation decision model is used as the second decision result.

[0024] The method for fusion processing of the first decision result and the second decision result is as follows:

[0025] The semantic similarity between the first decision result and the second decision result is calculated;

[0026] If the semantic similarity is equal to or greater than a set threshold, the decision result with a larger weight is selected as the final decision result; if the semantic similarity is less than the set threshold, the decision result with a larger confidence is selected as the final decision result.

[0027] When the semantic similarity is equal to or greater than the set threshold, the decision result with a larger weight is selected as the final decision result, and the specific selection method is as follows:

[0028] If the calculated semantic similarity is equal to or greater than the set threshold, the similarity between the data corresponding to the set threshold and the inductive data is calculated, and if the obtained similarity is greater than a second set threshold, the historical success rates of the first decision result and the second decision result are queried from an existing preset database;

[0029] The historical success rates of the first decision result and the second decision result are multiplied by the initial weights of the first decision result and the second decision result respectively, the final weights of the first decision result and the second decision result are calculated, and the decision result with a larger final weight is used as the final decision result.

[0030] An evaluation decision system for implementing an evaluation decision method, comprising a sensing data receiving module, an ontology reasoning module, a model decision module and a decision result fusion module, wherein:

[0031] The sensing data receiving module is configured to receive sensing data in response to a spacecraft sensor after a collision event;

[0032] The ontology reasoning module is configured to perform ontology reasoning on the sensing data to obtain a first decision result;

[0033] The model decision module is configured to input the sensing data into a large language model that has been trained to output a second decision result;

[0034] The decision result fusion module is configured to fuse the first decision result and the second decision result to obtain a final decision result.

[0035] The ontology reasoning module is established according to an ontology reasoning model, comprising a collision event ontology knowledge base construction unit, a collision event ontology rule base construction unit, an ontology reasoning model construction unit and an ontology reasoning unit, wherein:

[0036] The collision event ontology knowledge base construction unit is configured to construct an on-orbit collision event ontology knowledge base of a spacecraft;

[0037] The collision event ontology rule base construction unit is configured to construct an on-orbit collision event ontology rule base according to a spacecraft collision event response;

[0038] The ontology reasoning model construction unit is configured to construct an ontology reasoning model based on the on-orbit collision event ontology knowledge base and the on-orbit collision event ontology rule base;

[0039] The ontology reasoning unit is configured to input the sensing data into the ontology reasoning model for rule matching and logical deduction to generate the first decision result.

[0040] The decision result fusion module comprises a semantic similarity calculation unit, a first decision result fusion output unit and a second decision result fusion output unit, wherein:

[0041] The semantic similarity calculation unit is configured to calculate the semantic similarity between the first decision result and the second decision result;

[0042] The first decision result fusion output unit is configured to select a decision result with a larger weight as the final decision result if the semantic similarity is equal to or greater than a set threshold;

[0043] The second decision result fusion output unit is configured to select a decision result with a larger confidence as the final decision result if the semantic similarity is less than the set threshold.

[0044] The advantages of the present application compared with the prior art are:

[0045] (1) The spacecraft impact event evaluation decision method provided by the present application receives the sensing data of the spacecraft sensor first; performs ontology reasoning on the sensing data to obtain a first decision result; inputs the sensing data into a trained large language model to output a second decision result; and finally fuses the first decision result and the second decision result to obtain a final decision result. By effectively fusing the two decision results, the best decision result is automatically output, effectively ensuring the flight safety of the spacecraft and the life of the astronauts;

[0046] (2) The present application provides a fault model library constructed in the historical database, which contains multiple potential fault scenarios and performs multi-layer test framework verification to ensure that the system can perform as expected under different conditions. Based on the test results, the priority and weight parameters of the ontology rules are adjusted to optimize the performance of the reasoning engine and improve the reliability and response speed of the system in real environment. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The image schematic diagram of the US space station mechanical arm and the Hubble telescope impacted by space debris provided by the present application;

[0048] Figure 2 The flowchart of the spacecraft impact event evaluation decision method provided by the present application;

[0049] Figure 3 The module diagram of the spacecraft impact event evaluation decision system provided by the present application;

[0050] Figure 4 The actual case flowchart of the impact protection structure provided by the present application in the case of penetrating a certain cabin section. DETAILED DESCRIPTION

[0051] A spacecraft impact event evaluation decision method receives the sensing data of the spacecraft sensor first; performs ontology reasoning on the sensing data to obtain a first decision result; inputs the sensing data into a trained large language model to output a second decision result; and finally fuses the first decision result and the second decision result to obtain a final decision result. By effectively fusing the two decision results, the best decision result is automatically output, effectively ensuring the flight safety of the spacecraft and the life of the astronauts.

[0052] The spacecraft impact event evaluation decision method, the specific steps include:

[0053] Receiving the sensing data returned by the spacecraft sensor after the impact event;

[0054] perform ontology reasoning on the obtained sensing data to obtain a first decision result;

[0055] a pre-trained large language model is used to obtain a second decision result of the large language model after inputting the sensing data;

[0056] the first decision result and the second decision result are fused to obtain a final decision result.

[0057] The method for performing ontology reasoning on the sensing data and obtaining the first decision result is as follows:

[0058] An on-orbit impact event ontology knowledge base and an on-orbit impact event ontology rule base are constructed.

[0059] An ontology reasoning model is constructed according to the on-orbit impact event ontology knowledge base and the on-orbit impact event ontology rule base.

[0060] The ontology reasoning model is used to perform rule matching and logical deduction on the sensing data to obtain the first decision result.

[0061] In the ontology reasoning model, the on-orbit impact event ontology knowledge base is used for logical deduction on the sensing data, and the method for logical deduction is as follows:

[0062] After the sensing data is loaded, a space structure ontology related to the current task scenario is obtained, the impact of the impact damage on the function of the space structure ontology is analyzed, the corresponding emergency measures are determined according to the impact, an emergency plan is formulated using the emergency measures, and the execution emergency operation process of the spacecraft under a specific impact event is deduced.

[0063] The information obtained after the sensing data is loaded is used as the logical deduction result.

[0064] In the ontology reasoning model, the on-orbit impact event ontology rule base is used for rule matching on the sensing data, and the method for rule matching is as follows:

[0065] After the sensing data is loaded, object attributes and data attributes of the impact event are obtained; the object attributes are used to describe the on-orbit impact event category of the spacecraft, and the data attributes are used to define the association between the on-orbit impact event category of the spacecraft and the sensing data.

[0066] The object attributes and the data attributes are used as the rule matching result, an instance is created, and the instance creation result and the logical deduction result are used as the first decision result.

[0067] The trained model is used as a spacecraft evaluation decision model.

[0068] The method for fusing the first decision result and the second decision result is as follows:

[0069] The semantic similarity between the first decision result and the second decision result is calculated.

[0070] If the semantic similarity is equal to or greater than the set threshold, the decision result with a larger weight is selected as the final decision result; if the semantic similarity is less than the set threshold, the decision result with a larger confidence is selected as the final decision result.

[0071] When the semantic similarity is equal to or greater than the set threshold, the decision result with a larger weight is selected as the final decision result, and the specific selection method is:

[0072] If the calculated semantic similarity is equal to or greater than the set threshold, the similarity between the data corresponding to the set threshold and the sensing data is calculated, and if the obtained similarity is greater than a second set threshold, the historical success rates of the first decision result and the second decision result are queried from the existing preset database.

[0073] The historical success rates of the first decision result and the second decision result are multiplied by the initial weights of the first decision result and the second decision result respectively, and the final weights of the first decision result and the second decision result are calculated and obtained, and the decision result with a larger final weight is selected as the final decision result.

[0074] An evaluation decision system for implementing the evaluation decision method is designed as:

[0075] It includes a sensing data receiving module, an ontology reasoning module, a model decision module and a decision result fusion module, wherein:

[0076] The sensing data receiving module is used to receive the sensing data responded by the spacecraft sensor after the impact event;

[0077] The ontology reasoning module is used to perform ontology reasoning on the sensing data to obtain a first decision result;

[0078] The model decision module is used to input the sensing data into a large language model that has been trained to output a second decision result;

[0079] The decision result fusion module is used to fuse the first decision result and the second decision result to obtain a final decision result.

[0080] The ontology reasoning module is established according to an ontology reasoning model and includes a collision event ontology knowledge base construction unit, a collision event ontology rule base construction unit, an ontology reasoning model construction unit and an ontology reasoning unit, wherein:

[0081] The collision event ontology knowledge base construction unit is used to construct an on-orbit collision event ontology knowledge base of the spacecraft;

[0082] The collision event ontology rule base construction unit is used to construct an on-orbit collision event ontology rule base according to the response of the spacecraft to the impact event;

[0083] The ontology reasoning model construction unit is configured to construct an ontology reasoning model based on the on-orbit impact event ontology knowledge base and the on-orbit impact event ontology rule base.

[0084] The ontology reasoning unit is configured to input the sensing data into the ontology reasoning model to perform rule matching and logical deduction, and generate a first decision result.

[0085] The decision result fusion module includes a semantic similarity calculation unit, a first decision result fusion output unit, and a second decision result fusion output unit.

[0086] The semantic similarity calculation unit is configured to calculate the semantic similarity between the first decision result and the second decision result.

[0087] The first decision result fusion output unit is configured to select the decision result with a larger weight as the final decision result if the semantic similarity is equal to or greater than a set threshold.

[0088] The second decision result fusion output unit is configured to select the decision result with a larger confidence as the final decision result if the semantic similarity is less than the set threshold.

[0089] The following will be further described in conjunction with the accompanying drawings and preferred embodiments:

[0090] In the current embodiment, the ontology reasoning module can be further expanded:

[0091] The spacecraft on-orbit impact event ontology knowledge base construction unit is configured to construct a spacecraft on-orbit impact event ontology knowledge base; the spacecraft on-orbit impact event ontology knowledge base includes four modules of a spacecraft structure ontology, a damage mode ontology, a system function ontology, and an emergency operation ontology.

[0092] The impact event ontology rule base construction unit is configured to construct an impact event ontology rule base according to the spacecraft impact event response; the impact event ontology rule base includes two parts of attribute definition and instance creation; the attribute definition includes object attributes and data attributes; the object attributes are used to describe the relationship between classes in the spacecraft on-orbit impact event ontology knowledge base; the data attributes refer to the attributes associated with classes and data; the instance creation refers to the impact event types that may occur.

[0093] The ontology reasoning model construction unit is configured to construct an ontology reasoning model based on the spacecraft on-orbit impact event ontology knowledge base and the impact event ontology rule base.

[0094] The ontology reasoning unit is configured to input the sensing data into the ontology reasoning model to perform rule matching and logical deduction, and generate the first decision result.

[0095] The ontology reasoning unit comprises:

[0096] The data loading subunit is configured to input the sensing data into the ontology reasoning model, and dynamically load the sensing data and the space structure ontology related to the current task scenario;

[0097] The ontology association subunit is configured to analyze the potential impact of the damage on the spacecraft function based on the association relationship between the damage mode ontology and the system function ontology, and determine specific emergency measures to be taken;

[0098] The decision result output subunit is configured to deduce an executable emergency operation process of the spacecraft under a specific impact event according to an emergency scheme in the emergency operation ontology and in combination with the input sensing data, and output a decision result with semantic interpretation.

[0099] The first decision result fusion output unit comprises:

[0100] The similarity calculation subunit is configured to calculate the similarity between the sensing data and the preset data if the semantic similarity is equal to or greater than the set threshold value.

[0101] The historical success rate query subunit is configured to query the historical success rates of the first decision result and the second decision result from a preset database if the similarity is equal to or greater than the preset threshold value.

[0102] The weight adjustment subunit is configured to multiply the historical success rates of the first decision result and the second decision result by the initial weights of the first decision result and the second decision result respectively to obtain the final weights of the first decision result and the second decision result.

[0103] The first decision result fusion output subunit is configured to select the decision result with the larger final weight as the final decision result.

[0104] Embodiment one:

[0105] As shown in Figure 2 The spacecraft impact event post-evaluation decision method provided by the embodiment of the application comprises:

[0106] Step S110: receiving sensing data of a spacecraft sensor; the sensing data is used to reflect state information of a spacecraft structure part;

[0107] In this embodiment, the spacecraft sensor includes but is not limited to an acoustic emission sensor, a fiber Bragg grating (FBG) sensor, a current sensor, a thermal imaging sensor, and a barometric pressure sensor.

[0108] Step S120: performing ontology reasoning on the sensing data to obtain a first decision result;

[0109] Specifically, the step of performing ontology reasoning on the sensing data to obtain a first decision result includes:

[0110] An on-orbit spacecraft impact event ontology knowledge base is constructed, which is organized according to the structure, function and damage mode of the spacecraft, to ensure a quick response in different situations. Specifically, the on-orbit spacecraft impact event ontology knowledge base includes four modules: a spacecraft structure ontology, a damage mode ontology, a system function ontology and an emergency operation ontology. The spacecraft structure ontology is used to describe the structural information of the spacecraft, including the cabin section, auxiliary equipment and their connection relationship, and specific examples include the propulsion cabin, solar panel, communication module and their interaction, as well as the functional performance of each component in the normal and damaged states. The damage mode ontology is used to define the damage types and characteristics that may occur during the on-orbit operation of the spacecraft, covering various damage conditions such as cracks, leaks, fractures and deformations, and describing the formation mechanism of each damage and its impact on the overall performance of the spacecraft. The system function ontology is used to describe the system functions and states of the spacecraft, including power supply, attitude control, thermal management and communication functions, and clearly defines the expected performance and switching mechanism of each function in normal operation and emergency situations. The emergency operation ontology is used to define the operation strategies that need to be executed by the spacecraft in emergency situations, including isolating the damaged cabin section, starting the backup system, adjusting the attitude, switching the communication scheme, etc., to ensure that emergency measures can be quickly and effectively executed in the event of an impact event.

[0111] An impact event ontology rule base is constructed according to the response of the spacecraft impact event. The impact event ontology rule base includes attribute definition and instance creation. The attribute definition includes object attributes and data attributes. The object attributes are used to describe the relationship between classes in the on-orbit spacecraft impact event ontology knowledge base, including time, space, damage degree and other multi-dimensional relationship descriptions. These attributes can help the system understand the relative timing relationship and spatial relationship of the on-orbit impact event or other emergency response events. The data attributes refer to the attributes associated with classes and data, such as cabin number, impact event number, emergency measure number, etc., to ensure that each event and measure can be accurately tracked and managed. Instance creation refers to constructing specific instances through SWRL (Semantic Web Rule Language) rules according to the types of impact events that may occur. For example, when the FBG sensor detects a 10 J / cm 2 energy impact, it outputs a suggestion to start the cabin sealing within 300 ms. The specific SWRL statement is:

[0112] FBG_True(?x) ^ Safety schedule(?x, "In_progress") ^ with(?x,?y)

[0113] Energy_shock_J_cm(c) swrlb:greaterThan(?c, 10) -> HasMeasure(?x, Seal_cabin_section_300ms)

[0114] An ontology reasoning model is constructed based on the spacecraft in-orbit impact event ontology knowledge base and the impact event ontology rule base;

[0115] The sensing data is input into the ontology reasoning model for rule matching and logical deduction to generate a first decision result.

[0116] Specifically, the sensing data is input into the ontology reasoning model for rule matching and logical deduction to generate a first decision result, including:

[0117] The sensing data is input into the ontology reasoning model, and the sensing data and the spacecraft structure ontology related to the current task scene are dynamically loaded to ensure that decisions are made based on the latest task and environmental information;

[0118] Based on the association relationship between the damage mode ontology and the system function ontology, the potential impact of damage on the spacecraft function is analyzed, and specific emergency measures to be taken are determined;

[0119] According to the emergency scheme in the emergency operation ontology, combined with the input sensing data, the executable emergency operation process of the spacecraft under a specific impact event is deduced, and a decision result with semantic explanation is output to ensure the logicality and understandability of the decision.

[0120] Step S130: input the sensing data into the trained large language model to output a second decision result;

[0121] It should be noted that the step of obtaining a first decision result by ontology reasoning on the sensing data and the step of outputting a second decision result by inputting the sensing data into the trained large language model do not distinguish the order.

[0122] Step S140: fuse the first decision result and the second decision result to obtain a final decision result. The final decision result is used to guide the emergency response of the spacecraft after the impact event.

[0123] The first decision result and the second decision result are fused to obtain a final decision result, including:

[0124] The semantic similarity between the first decision result and the second decision result is calculated; specifically, assuming that the result of ontology reasoning suggests performing a certain operation, while the LLM output suggests a different operation, conflict resolution should be performed at this time. Conflict recognition is based on the similarity recognition of text content, which is as follows:

[0125] Preprocessing: Preprocessing is done by word segmentation, removing stop words, stemming, lemmatization, and lowercasing.

[0126] At the same time, the text is numerically represented as a feature vector, processed by the bag-of-words model and TF-IDF model.

[0127] The bag-of-words model represents the text as a vocabulary, and constructs a feature vector by counting the number of occurrences of each word in the text.

[0128] For example, for two texts "T1" and "T2", if the vocabulary contains "spacecraft", "impact", and "emergency", the following feature vectors can be obtained:

[0129] T1: [1, 1, 0] ("spacecraft" appears once, "impact" appears once, "emergency" appears zero times)

[0130] T2: [0, 1, 1] ("spacecraft" appears zero times, "impact" appears once, "emergency" appears once)

[0131] The TF-IDF model is based on the bag-of-words model, considering the frequency of words in the document (TF) and the rarity of words in all documents (IDF), through the formula:

[0132]

[0133] Where N is the total number of documents, DF(t) is the number of documents containing the word t, used to reduce the influence of common words.

[0134] When the text is converted into a feature vector, the similarity between the vectors is used to determine the similarity between the texts.

[0135] Specifically, the Jaccard similarity coefficient and Euclidean distance can be used to calculate:

[0136] The Jaccard similarity coefficient is an index used to measure the similarity between two sets, defined as the ratio of the size of the intersection of the two sets to the size of the union:

[0137]

[0138] Where |A∩B| is the intersection size of set A and set B, |A∪B| is the union size of set A and set B.

[0139] Suppose the threshold is set to 0.5, and there are two texts "T1" and "T2":

[0140] T1: "Start the five cabin section sealed cabin"

[0141] T2: "Adjust the pressure of the sealed cabin to compensate for the leakage"

[0142] The word segmentation result of T1 is: [“start”, “number five”, “compartment”, “seal”, “compartment”]

[0143] The eigenvector is: [1,1,1,1,1,0,0,0,0,0]

[0144] The word segmentation results of T2 are: [“through”, “seal”, “cabin”, “pressure”, “regulation”, “compensation”, “leakage”]

[0145] The eigenvector is: [0,0,0,1,1,1,1,1,1,1]

[0146] Intersection |A∩B|=2

[0147] Union |A∪B|=10

[0148]

[0149] The Jaccard similarity coefficient is 0.2, which means the similarity is low and there is no conflict.

[0150] The Euclidean distance is calculated by representing the features and then calculating the distance between two texts:

[0151]

[0152] Convert distance to similarity:

[0153] Therefore, the similarity between texts T1 and T2 is low and there is no conflict.

[0154] If the semantic similarity is equal to or greater than the set threshold, it means that there is no conflict between the first decision result and the second decision result, and the weight management is included, and the decision result with the larger weight is selected as the final decision result;

[0155] Specifically, if the semantic similarity is equal to or greater than the set threshold, the decision result with a larger weight is selected as the final decision result, including:

[0156] If the semantic similarity is equal to or greater than the set threshold, the similarity between the sensed data and the preset data is calculated. Specifically, the similarity is calculated by calculating the Euclidean distance between the sensor data and the damage pattern to confirm whether the data is similar. For example, if the feature vector of the damage pattern is D and the feature vector of the sensor data is S, the similarity can be calculated using the following formula:

[0157]

[0158] If the similarity is equal to or greater than the preset threshold, query the historical success rate of the first decision result and the second decision result from the preset database;

[0159] Multiply the historical success rate of the first decision result and the second decision result by the initial weight of the first decision result and the second decision result respectively to obtain the final weight of the first decision result and the second decision result; for example, if the success rate of the ontology reasoning result based on historical data is 0.8, and the success rate of the LLM output is 0.9, the initial weight of the ontology reasoning result is 0.6, and the initial weight of the LLM output is 0.4, then the adjusted weights are:

[0160] Ontology reasoning result weight: 0.6 x 0.8 = 0.48

[0161] LLM output weight: 0.4 x 0.9 = 0.36

[0162] The normalized final weight after adjustment:

[0163] The final weight of the ontology reasoning result:

[0164] The final weight of the LLM output:

[0165] Select the decision result with the larger final weight as the final decision result.

[0166] If the semantic similarity is less than the set threshold, it means that the first decision result and the second decision result conflict, and the decision result with the higher confidence is selected as the final decision result.

[0167] Specifically, priority selection is performed by the analytic hierarchy process.

[0168] Score is obtained by comparing the effectiveness, feasibility, and risk of the three dimensions, and a matrix is formed. For example, the score of effectiveness T1 is 3, and the score of T2 is 1 / 3, and the matrix is Assuming the weights are 0.5, 0.3, and 0.2 respectively, the priority of the three matrices is calculated: Score(T1) = (0.5 x T1 effectiveness result) + … = 0.6

[0169] Score(T2) = 0.3

[0170] Therefore, the priority of T1 is higher than that of T2.

[0171] In this embodiment, the final decision includes a comprehensive decision suggestion, a decision output, and an implementation feedback mechanism.

[0172] Specifically, the integrated decision recommendation refers to the weighted fusion of ontology reasoning results and LLM output results after weight adjustment and conflict resolution. The weight is the dynamic weight, and the final decision recommendation is generated. The fusion algorithm is as follows:

[0173] Final Decision=α·Ontology Result+(1-α)·LLM Output

[0174] The decision output is output in a clear and understandable manner, with the logical path of the decision basis.

[0175] For example:

[0176] Decision recommendation: output suggestion to start cabin sealing within 300ms

[0177] Other solutions can be adopted: compensate for leakage by adjusting the pressure of the sealed cabin

[0178] Supporting data:

[0179] Current cabin pressure data: 0.8atm (standard value 1.0atm)

[0180] Temperature: 22℃

[0181] Historical case analysis:

[0182] Case A: Start sealing under similar conditions on the ground simulation, success rate 90%.

[0183] Case B: Start sealing, only start pressure compensation, resulting in cabin damage.

[0184] Logical path:

[0185] 1. Sensor data shows that cabin pressure is lower than standard value

[0186] 2. Historical data supports starting sealing under similar conditions

[0187] 3. After weight adjustment, the recommendation to start sealing is the highest priority

[0188] The implementation feedback mechanism is to collect feedback information after the decision is implemented, observe the actual results of the decision effect and emergency response, improve the subsequent decision-making process, and store it in the historical database to form a closed-loop feedback mechanism after decision-making. The feedback mechanism is shown below:

[0189] Feedback collection: collect real-time data feedback such as cabin pressure changes, sealing status, etc. through sensor monitoring of the implementation effect. At the same time, perform effect evaluation such as analyzing the success rate after the decision is implemented, failure cases and their reasons.

[0190] Improvement mechanism: based on the feedback analysis results, identify the problems and potential improvements in the decision-making process; feedback the improvement suggestions to the decision model, adjust the parameters and weights, and optimize the subsequent decision-making process.

[0191] Historical database: store the process, execution and feedback results of each decision into the historical database to provide data support for future decision-making.

[0192] Feedback record:

[0193] Decision implementation time: 2025.3.20 21:15

[0194] Post-implementation cabin segment sensor data: pressure sensor 1.0 atm (returns to normal)

[0195] Implementation effect: success

[0196] Key indicators: response time: 250ms

[0197] Feedback analysis: the decision meets expectations, no significant problems are found

[0198] It is suggested that the decision be continued in the future.

[0199] It should be noted that in order to verify the accuracy and robustness of the reasoning, the embodiment of the present application also includes a fault injection test step. Specifically, a fault model library is constructed in the historical database, including a variety of potential fault scenarios, such as hardware failure (e.g. sensor failure), software failure (e.g. decision algorithm error) and environmental interference (e.g. sensor signal noise) etc.; a multi-layer test framework is verified, including but not limited to input data integrity test, noise data processing test and emergency response adaptability test, to ensure that the system performs as expected under different conditions. Based on the test results, the priority and weight parameters of the ontology rules are adjusted to optimize the performance of the reasoning engine and improve the reliability and response speed of the system in real environment.

[0200] As shown in Figure 3 , the spacecraft impact event evaluation decision system provided by the embodiment of the present application comprises:

[0201] The sensing data receiving module 100 is used for receiving sensing data of a spacecraft sensor; the sensing data is used for reflecting state information of a spacecraft structure part;

[0202] In this embodiment, the spacecraft sensor includes but is not limited to an acoustic emission sensor, a fiber bragg grating (FBG) sensor, a current sensor, a thermal imaging sensor and a barometric pressure sensor.

[0203] The ontology reasoning module 200 is used for performing ontology reasoning on the sensing data to obtain a first decision result;

[0204] Specifically, the ontology reasoning module 200 comprises:

[0205] The spacecraft on-orbit impact event ontology knowledge base construction unit is configured to construct a spacecraft on-orbit impact event ontology knowledge base, which is organized according to the structure, function and damage mode of the spacecraft, and ensures that the spacecraft can quickly respond in different situations. Specifically, the spacecraft on-orbit impact event ontology knowledge base comprises four modules, namely, a spacecraft structure ontology, a damage mode ontology, a system function ontology and an emergency operation ontology. The spacecraft structure ontology is configured to describe the structural information of the spacecraft, including the cabin section, the auxiliary equipment and the connection relationship therebetween. Specific examples include a propulsion cabin, a solar panel, a communication module and the interaction therebetween, and the function performance of each component in the normal and damaged states. The damage mode ontology is configured to define the damage types and characteristics that can occur during the on-orbit operation of the spacecraft, covering various damage conditions such as cracks, leaks, fractures and deformations, and describing the formation mechanism of each damage and the influence on the overall performance of the spacecraft. The system function ontology is configured to describe the system function and state of the spacecraft, including power supply, attitude control, thermal management and communication function, and to clearly define the expected performance and switching mechanism of each function in the normal operation and emergency situation. The emergency operation ontology is configured to define the operation strategy that needs to be executed by the spacecraft in the emergency situation, specifically including isolating the damaged cabin section, starting the backup system, adjusting the attitude and switching the communication scheme, so as to ensure that the emergency measures can be quickly and effectively executed when the impact event occurs.

[0206] The impact event ontology rule base construction unit is configured to construct an impact event ontology rule base according to the response of the spacecraft to the impact event. The impact event ontology rule base comprises two parts, namely, attribute definition and instance creation. The attribute definition comprises object attributes and data attributes. The object attributes are configured to describe the relationship between classes in the spacecraft on-orbit impact event ontology knowledge base, including time, space, damage degree and other multi-dimensional relationship descriptions. These attributes can help the system understand the relative time sequence relationship and spatial relationship of the on-orbit impact event or other emergency response events. The data attributes refer to the attributes associated with classes and data, such as cabin number, impact event number and emergency measure number, so as to ensure that each event and measure can be accurately tracked and managed. The instance creation refers to constructing specific instances through SWRL (Semantic Web Rule Language) rules according to the types of impact events that can occur.

[0207] The ontology reasoning model construction unit is configured to construct an ontology reasoning model based on the spacecraft on-orbit impact event ontology knowledge base and the impact event ontology rule base.

[0208] The ontology reasoning unit is configured to input the sensing data into the ontology reasoning model for rule matching and logical deduction, and generate a first decision result.

[0209] Further, the ontology reasoning unit comprises:

[0210] The data loading subunit is configured to input the sensing data into the ontology reasoning model, dynamically load the aerospace structure ontology related to the current task scene, and ensure that the decision is made according to the latest task and environmental information.

[0211] The ontology association subunit is configured to analyze the potential impact of the damage on the spacecraft function based on the association relationship between the damage mode ontology and the system function ontology, and determine the specific emergency measures to be taken.

[0212] The decision result output subunit is configured to deduce the executable emergency operation process of the spacecraft under a specific impact event according to the emergency scheme in the emergency operation ontology and in combination with the input sensing data, and output a decision result with semantic interpretation, so as to ensure the logicality and understandability of the decision.

[0213] The model decision module 300 is configured to input the sensing data into the trained large language model to output a second decision result.

[0214] It should be noted that the step of obtaining the first decision result by performing ontology reasoning on the sensing data and the step of obtaining the second decision result by inputting the sensing data into the trained large language model do not distinguish the order.

[0215] The decision result fusion module 400 is configured to fuse the first decision result and the second decision result to obtain a final decision result. The final decision result is used to guide the emergency response of the spacecraft after the impact event.

[0216] Specifically, the decision result fusion module 400 comprises:

[0217] The semantic similarity calculation unit is configured to calculate the semantic similarity between the first decision result and the second decision result.

[0218] The first decision result fusion output unit is configured to select the decision result with a larger weight as the final decision result if the semantic similarity is equal to or greater than a set threshold.

[0219] Further, the first decision result fusion output unit comprises:

[0220] The similarity calculation subunit is configured to calculate the similarity between the sensing data and the preset data if the semantic similarity is equal to or greater than a set threshold. Specifically, the similarity is calculated by calculating the Euclidean distance between the sensor received data and the damage mode to determine whether the data is similar. For example, assuming that the feature vector of the damage mode is D and the feature vector of the sensor data is S, the similarity can be calculated by the following formula:

[0221] a historical success rate query subunit configured to query, from a preset database, a historical success rate of the first decision result and the second decision result if the similarity is equal to or greater than a preset threshold;

[0222] a weight adjustment subunit configured to multiply the historical success rate of the first decision result and the second decision result by an initial weight of the first decision result and the second decision result, respectively, to obtain a final weight of the first decision result and the second decision result;

[0223] a first decision result fusion output subunit configured to select a decision result with a larger final weight as a final decision result.

[0224] a second decision result fusion output unit configured to select a decision result with a larger confidence as a final decision result if the semantic similarity is less than a set threshold, indicating that the first decision result and the second decision result conflict.

[0225] In the embodiment, the final decision includes a comprehensive decision suggestion, a decision output, and an implementation feedback mechanism.

[0226] Specifically, the comprehensive decision suggestion refers to weighting and fusing the ontology reasoning result and the LLM output result after weight adjustment and conflict resolution, and the weight is the weight of the dynamic weight, to generate a final decision suggestion. The fusion algorithm is as follows:

[0227] Final Decision = a Ontology Result + (1-a) LLM Output

[0228] The decision output is to output the decision result in a clear and understandable manner, with the logical path of the decision basis.

[0229] The implementation feedback mechanism collects feedback information after the decision is implemented, observes the actual results of the decision effect and emergency response, improves the subsequent decision-making process, and stores it in the historical database to form a closed-loop feedback mechanism after decision-making.

[0230] It should be noted that, in order to verify the accuracy and robustness of the reasoning, the embodiment of the present application further includes a fault injection test module for constructing a fault model library in the historical database, including various potential fault scenarios such as hardware failure (e.g. sensor failure), software failure (e.g. decision algorithm error) and environmental interference (e.g. sensor signal noise) etc. A multi-layer test framework is verified, including but not limited to integrity testing of input data, processing testing of noise data and emergency response adaptability testing, to ensure that the system performs as expected under different conditions. Based on the test results, the priority and weight parameters of the ontology rules are adjusted to optimize the performance of the reasoning engine and improve the reliability and response speed of the system in real environment.

[0231] In the embodiment, it can be seen that:

[0232] like Figure 4 As shown, when the FBG sensor detects 10J / cm 2 During energy impact:

[0233] Ontological reasoning: It is recommended to start the compartment sealing within 300ms;

[0234] The large language model suggests: gradually compensating for leakage by adjusting the cabin pressure;

[0235] Fusion decision: perform the sealing operation first and then start the pressure equalization procedure.

[0236] In summary, the embodiments of the present invention provide a cross-sensor intelligent decision-making method and system that integrates ontology reasoning and a large language model, which is suitable for emergency fault handling of spacecraft such as space stations and satellites.

[0237] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0238] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes 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 a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0239] These computer program instructions may 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, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0240] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0241] Although the present application has been disclosed in connection with the preferred embodiments shown, it should be understood that certain modifications would be permitted under the patent statutes and application

[0242] equivalent variations and modifications, all of which are intended to be within the scope of this application.

[0243] The contents of the present application which are not described in detail in the specification are known to those skilled in the art.

Claims

1. A method for evaluating and making decisions after a spacecraft impact event, characterized in that include: Receive sensing data sent back by spacecraft sensors after the impact event; Performing ontological reasoning on the obtained sensing data to obtain a first decision result; Input the sensing data into the trained spacecraft evaluation decision model to obtain the output second decision result; The first decision result and the second decision result are integrated to obtain the final decision result.

2. The method for evaluating and making decisions after a spacecraft impact event according to claim 1, characterized in that: The method for performing ontological reasoning on the sensing data and obtaining the first decision result is: Construct a knowledge base of spacecraft on-orbit impact events and a rule base of on-orbit impact events; Construct an ontology reasoning model based on the on-orbit impact event ontology knowledge base and the on-orbit impact event ontology rule base; According to the ontology reasoning model, the sensing data is matched with rules and logically deduced to obtain the first decision result.

3. The method for evaluating and making decisions after a spacecraft impact event according to claim 2, characterized in that: In the ontology reasoning model, the on-orbit impact event ontology knowledge base is used to perform logical deduction on the sensing data. The logical deduction method is as follows: After the sensing data is loaded, the aerospace structure body related to the current mission scenario is obtained, the impact of the impact damage on the function of the aerospace structure body is analyzed, the corresponding emergency measures are determined according to the impact situation, and the emergency plan is formulated using the emergency measures to guide the spacecraft to perform emergency operation procedures under specific impact events; The information obtained after the sensing data loading is completed is used as the logical deduction result.

4. The method for evaluating and making decisions after a spacecraft impact event according to claim 3, characterized in that: In the ontology reasoning model, the on-orbit impact event ontology rule base is used to perform rule matching on the sensing data. The rule matching method is as follows: After the sensing data is loaded, the object attributes and data attributes of the impact event are obtained; the object attributes are used to describe the category of the spacecraft on-orbit impact event, and the data attributes are used to define the association between the category of the spacecraft on-orbit impact event and the sensing data; An instance is created based on the object attributes and data attributes as the rule matching result, and the instance creation result and the logical deduction result are used as the first decision result.

5. The method for evaluating and making decisions after a spacecraft impact event according to claim 4, characterized in that: The spacecraft evaluation decision model adopts a commercial large language model that has been trained. After the sensing data is input, the output result composed of the historical experience data output by the spacecraft evaluation decision model is used as the second decision result.

6. The method for evaluating and making decisions after a spacecraft impact event according to claim 5, characterized in that: The method for fusing the first decision result and the second decision result is: Calculating semantic similarity between the first decision result and the second decision result; If the semantic similarity is equal to or greater than the set threshold, the decision result with the larger weight is selected as the final decision result; If the semantic similarity is less than the set threshold, the decision result with greater confidence is selected as the final decision result.

7. The method for evaluating and making decisions after a spacecraft impact event according to claim 6, characterized in that: When the semantic similarity is equal to or greater than the set threshold, the decision result with the larger weight is selected as the final decision result. The specific selection method is as follows: If the calculated semantic similarity is equal to or greater than a set threshold, the similarity between the data corresponding to the set threshold and the sensed data is calculated. If the obtained similarity is greater than a second set threshold, the historical success rates of the first decision result and the second decision result are queried from an existing preset database. Multiply the historical success rates of the first decision result and the second decision result by the initial weights of the first decision result and the second decision result respectively, calculate the final weights of the first decision result and the second decision result, and use the decision result with the larger final weight as the final decision result.

8. An evaluation and decision-making system for implementing the evaluation and decision-making method according to claim 7, characterized in that: It includes a sensing data receiving module, an ontology reasoning module, a model decision module, and a decision result fusion module, among which: A sensing data receiving module is used to receive sensing data responded by the spacecraft sensor after the impact event; An ontology reasoning module, configured to perform ontology reasoning on the sensing data to obtain a first decision result; A model decision module is used to input the sensing data into the trained large language model and output a second decision result; The decision result fusion module is used to fuse the first decision result and the second decision result to obtain the final decision result.

9. The evaluation and decision-making system according to claim 8, characterized in that: The ontology reasoning module is established according to the ontology reasoning model, and includes a collision event ontology knowledge base construction unit, a collision event ontology rule base construction unit, an ontology reasoning model construction unit, and an ontology reasoning unit, wherein: Impact event ontology knowledge base construction unit, used to construct the spacecraft on-orbit impact event ontology knowledge base; An impact event ontology rule base construction unit is used to construct an on-orbit impact event ontology rule base based on the spacecraft impact event response situation; An ontology reasoning model construction unit, configured to construct an ontology reasoning model based on an on-orbit impact event ontology knowledge base and an on-orbit impact event ontology rule base; The ontology reasoning unit is used to input the sensing data into the ontology reasoning model for rule matching and logical deduction to generate a first decision result.

10. An evaluation and decision-making system according to claim 9, characterized in that: The decision result fusion module includes a semantic similarity calculation unit, a first decision result fusion output unit, and a second decision result fusion output unit, wherein: a semantic similarity calculation unit, configured to calculate the semantic similarity between the first decision result and the second decision result; The first decision result fusion output unit is used to select the decision result with a larger weight as the final decision result if the semantic similarity is equal to or greater than a set threshold; The second decision result fusion output unit is used to select the decision result with greater confidence as the final decision result if the semantic similarity is less than a set threshold.

Images