Aerospace power fault diagnosis and health assessment method and system

Abstract

The application relates to a spaceflight power fault diagnosis method and system, which comprises the following steps: establishing the relationship between a spaceflight power system and existing measuring points of the system based on a multi-signal flow diagram, dividing power system parameters into power system parameters capable of directly measuring system faults and the rest of the power system parameters; obtaining power system parameters in real time according to power system state monitoring information sent by a rocket-earth communication link and power system historical test and flight data; performing power fault diagnosis according to the power system parameters capable of directly measuring system faults; and establishing a fault diagnosis model based on a distributed TEAMS-RT algorithm for the rest of the power parameters, and performing power fault diagnosis in combination with a statistical and feature map mapping fault diagnosis mode.

Description

Technical Field

[0001] This invention relates to an aerospace propulsion fault diagnosis and health assessment technology based on multi-signal flow graphs and the distributed TEAMS-RT algorithm, belonging to the field of prediction and health management technology. Background Technology

[0002] Reusable launch vehicles (RUVs) are a new type of spacecraft capable of rapidly traversing the atmosphere to and from Earth, returning to Earth after completing their missions, and being reused. They represent a crucial direction in the development of space launch vehicles worldwide and are a vital means of achieving rapid access to space. Compared to traditional disposable launch vehicles, reusable RUVs require fault diagnosis and health management functions, which are essential for enabling reusability, improving reliability, safety, and ground-based autonomous support capabilities.

[0003] To fulfill the functions of air-ground collaborative health management, routine inspection and maintenance, pre-launch test decision-making, and health management system maintenance, a reusable launch vehicle power fault diagnosis and health management system was constructed.

[0004] Many basic research works have been carried out in this field both at home and abroad, but the relevant research is still in its initial stage. Further improvements are needed in the design of power fault diagnosis and health management systems, research on fault diagnosis and health management methods, and process design before the conditions for engineering implementation can be met. Summary of the Invention

[0005] The technical problem solved by this invention is to develop a method and system for fault diagnosis and health assessment of aerospace propulsion based on multi-signal flow graphs and distributed TEAMS-RT algorithm, so as to complete the functions of joint health management of propulsion system between the aircraft and the ground, daily inspection and maintenance, pre-launch test decision-making, and health management system maintenance.

[0006] The solution of this invention is: a method for diagnosing aerospace power system faults, comprising:

[0007] Based on FME(C)A analysis, a set of power system failure modes and a set of measurement parameters are obtained; the set of failure modes includes all failure modes obtained through FME(C)A analysis, and the set of measurement parameters includes the set of existing measurable parameters for any power system.

[0008] Based on the multi-signal flow graph analysis method, the relationship between fault modes of aerospace propulsion system and system measurement parameters is established. The detectable propulsion system parameters are divided into a set of single fault parameters and a set of parameters with fault ambiguity. A mapping table between the fault mode set and the parameter set is obtained. A single fault parameter set refers to a parameter set in which each parameter corresponds to only one fault mode. A parameter set with fault ambiguity refers to a parameter set in which each parameter corresponds to two or more fault modes. The mapping table between the fault mode set and the parameter set refers to the correspondence between any fault mode in the fault mode set and the parameter in the parameter set that represents this fault.

[0009] For a single fault parameter set in a power system, a traversal query method is used for fault diagnosis. For a parameter set in a power system with fault ambiguity, a fault diagnosis D-matrix model is established based on a multi-signal flow graph. Based on the TEAMS-RT diagnostic algorithm, a D-matrix merging method is used for power system fault diagnosis.

[0010] Preferably, the method of traversing and querying for fault diagnosis of a single fault parameter set in the power system mainly includes:

[0011] (1) Determine the envelope range of each parameter in a single parameter set;

[0012] (2) Normalize the parameters in the single fault parameter set, compare the measured parameter values ​​with the corresponding envelope range, and use the state variable "1" to indicate normal when the parameter is within the envelope range; use the state variable "0" to indicate abnormal when the measured parameter value is outside the parameter change envelope range; transform the single fault parameter set into a power system health state characteristic set.

[0013] (3) Based on the mapping relationship table between the fault mode set and the single parameter set, map any parameter with a characteristic state of "1" in the parameter set to the single fault parameter set. When all the parameter characteristic states of a certain fault mapping in the single fault mode set are "1", it indicates that the fault has occurred. When any parameter with a characteristic state of "0" in all the parameters of a certain fault mapping, it indicates that the fault has not occurred.

[0014] (4) In the next sampling period, proceed to the next fault diagnosis and repeat steps (2) to (4) until the fault diagnosis is completed.

[0015] Preferably, for a parameter set of a power system with fault ambiguity, a fault diagnosis D-matrix model is established based on a multi-signal flow graph. The model establishment method includes:

[0016] (1) Establishing a two-dimensional D m×n A matrix diagnostic model framework; a set of fault modes am Fault modes that do not contain a single set of fault parameters are used as D. m×n The headers of the rows in the matrix represent fault mode 1, fault mode 2, and so on. m ; The set of fault ambiguity parameters b n Each parameter is used as D m×n The column headers of the matrix are measurement parameter 1, measurement parameter 2, ..., measurement parameter n, respectively; the elements of the rows and columns together form D. m×n A fault diagnosis matrix framework;

[0017] (2) D m×n Any element D in ij The elements are normalized to obtain D. m×n Fault diagnosis matrix; based on the mapping table between the fault mode set and the parameter set, when fault mode j occurs, it can be detected by measurement point i, that is, when fault mode j can be characterized by measurement parameter i, then D m×n The corresponding D in ij The element is then set to "1"; for example, when fault mode j occurs, it cannot be detected by measurement point i, that is, when fault mode j cannot be characterized by measurement parameter i, then D m×n The corresponding D in ij The element is then set to "0"; in this way, the representational relationship between all fault modes and all measurement points is normalized to obtain D, which consists of state characteristic values ​​0 and 1. m×n Fault diagnosis matrix, also known as fault diagnosis D-matrix model.

[0018] Preferably, the step of establishing a fault diagnosis D-matrix model based on multi-signal flow graphs, and using a D-matrix merging method for power system fault diagnosis based on the TEAMS-RT diagnostic algorithm, includes:

[0019] (1) Within one sampling period, obtain the characteristic vectors of all measured parameters; using the real-time telemetry data of the power system transmitted by the airborne PHM data center of the power system through the rocket-to-ground communication link, as well as the historical test and flight data of the power system, merge all data into a unified set of collected parameters c. m Determine the envelope range of all acquired parameters, normalize all acquired parameters, and compare the acquired parameter values ​​with the parameter envelope. If the parameter is within the envelope range, use the state variable "1" to indicate normal operation; if the measured parameter value is outside the parameter variation envelope range, use the state variable "0" to indicate abnormal operation; calculate c within one sampling period. m The parameter set becomes an eigenvector d consisting of 0s and 1s. p (p = m);

[0020] (2) The feature vector d pTransform the rank to the vertical dimension and record the vector d. p Find the x-coordinate k (k∈[1, 2, 3...p]) of all rows containing "0" elements; delete D. m×n The fault diagnosis matrix in row K, D Kj (j∈[1, 2, 3...n]) is the column vector containing elements with eigenvalues ​​of "0". Assume that in D... m×n In a matrix, if q columns and K rows containing "0" elements are deleted, then D will be... m×n The fault diagnosis matrix is ​​merged into D m×(n-q) 1-order matrix;

[0021] (3) Within one sampling period, the feature vector d of the collected parameters p With D m×(n-q) All column vectors D in the fault diagnosis matrix Kj (K∈[1, 2, 3…m]; j∈[1, 2, 3…(nq)]) are compared one by one, when d p =D Kj When (K∈[1, 2, 3…m]; j∈[1, 2, 3…(nq)]), it indicates that the ambiguity of the fault diagnosis is 1, D m×(n-q) The fault diagnosis matrix header contains a unique corresponding fault mode j, which occurs when d p ≠D Kj When d, the corresponding fault mode j in the header of the Dm×(nq) fault diagnosis matrix has not occurred; when d p =D Kj (K∈[1, 2, 3…m]; j∈[1, 2, 3…(nq)])=D Kl When (K∈[1, 2, 3……m];l∈[1, 2, 3……(nq)]), it indicates that the ambiguity of the fault diagnosis is 2, which means that the fault may be fault mode j, fault mode l, or fault mode j and fault mode l may occur simultaneously.

[0022] (4) For the fault mode set a m In all fault modes, when the fault diagnosis ambiguity is ≤2, the fault diagnosis will proceed to the next fault diagnosis in the next sampling period, repeating steps (1) to (3) until the fault diagnosis ends; when the fault diagnosis ambiguity is ≥3, it indicates that the setting of the measurement points used for the measurement parameters is unreasonable or the number of measurement points is insufficient, so it is necessary to readjust or add measurement points, and regenerate the parameter set b with fault ambiguity. n A mapping table between the fault mode set and the parameter set, and the construction of a new D m×n After the fault diagnosis matrix is ​​completed, repeat steps (1) to (3) until the fault diagnosis is finished.

[0023] Preferably, after completing the fault diagnosis of the power system, a benchmark assessment model is constructed, and the health status of the power system is comprehensively assessed by combining real-time telemetry data, historical test data and flight data of the power system. Ground maintenance and support personnel adopt different maintenance and support strategies according to different health assessment levels of the power system, and after completing the maintenance and support of the power system, the health status of the power system is comprehensively assessed again.

[0024] Preferably, the comprehensive assessment of the health status of the power system includes:

[0025] (1) Construct a benchmark evaluation model P based on the characteristic parameters of the dynamic system under normal conditions. m ;

[0026] (2) Within one sampling period, based on the latest real-time telemetry data of the power system, obtain the feature matrix Q of the power system in its current state. n (m = n);

[0027] (3) Calculate the benchmark evaluation model matrix P obtained in (1). m The feature matrix Q obtained in (2) under the current state n The distance between them is KL, and the magnitude of KL represents P. m (i) and Q n The degree of closeness between (j);

[0028] P m (i) and Q n (j) are matrices P m With Q n Elements in;

[0029] (4) Using the distance KL obtained in (3) and combined with the fault mode level and number, the health status of the vehicle's power system is comprehensively judged and evaluated using the power system health assessment criteria.

[0030] Preferably, the benchmark evaluation model matrix P m include:

[0031] (1) Conduct FME(C)A analysis on the launch vehicle power system to determine the categories of health characterization parameters of key components of the launch vehicle power system; extract the health characterization parameters A of key components from the historical test and flight data of the power system according to the categories of health characterization parameters. m ;

[0032] (2) For the health characterization parameter set A m The monitoring data was used to extract features using wavelet / wavelet packet decomposition methods to form a feature set B that reflects the health status. m ;

[0033] (3) Use the ISOMAP method to map the feature set B m A high-dimensional matrix with physical quantities is decomposed into a dimensionless low-dimensional characteristic matrix to obtain a set of characteristic parameters. This set of characteristic parameters is then used as a benchmark to evaluate the model matrix P. m .

[0034] Preferably, the characteristic matrix Q of the system in the current state of the power system is obtained. n (m = n) includes:

[0035] (1) Within a sampling period, the set of measurable parameters C of the key components of the launch vehicle's power system is obtained by receiving the latest real-time telemetry data of the power system. n ;

[0036] (2) For the set of monitorable parameters C n The monitoring parameters are used to extract features using the EMD method, forming a feature set E that reflects the current health status of key components of the power system. n ;

[0037] (3) Using the Local Multidimensional Scaling (LMDS) method to analyze the feature set E n A high-dimensional matrix with physical quantities is decomposed into a dimensionless low-dimensional characteristic matrix, yielding a set of characteristic parameters. This set of characteristic parameters is then used as the characteristic matrix Q. n .

[0038] The preferred formula for distance KL is as follows:

[0039]

[0040] Preferably, the health assessment criteria for propulsion systems include: for cryogenic liquid rocket propulsion systems:

[0041] When KL=0 and no faults are diagnosed, the power system is in good health and meets the requirements for performing the task.

[0042] When 0 < KL ≤ 2, and the number of non-Class I and Class II faults diagnosed is ≤ 3, the power system health status is abnormal, but it can still continue to perform tasks; the fault mode level is obtained by FME(C)A analysis.

[0043] When KL≥3 and the number of non-Class I and Class II faults diagnosed is >3, the power system health status is faulty, ground maintenance and support work is required, and the mission cannot continue.

[0044] When the number of diagnosed Class I and Class II faults is ≥1, regardless of the value of KL, the health status of the power system is faulty, and ground maintenance and support work needs to be carried out immediately. The mission cannot continue.

[0045] An aerospace propulsion fault diagnosis system includes an information processing unit and an air-ground collaborative health management unit;

[0046] Information Processing Unit: Based on FME(C)A analysis, it obtains the set of power system fault modes and the set of measurement parameters; based on the multi-signal flow graph analysis method, it establishes the relationship between the aerospace power system fault modes and the system measurement parameters, and generates a mapping table between the fault mode set and the parameter set; in the parameter set, the detectable power system parameters are divided into two categories: a set of single fault parameters and a set of parameters with fault ambiguity; a set of single fault parameters means that any parameter in the parameter set corresponds to only one fault mode in the mapping table, while a set of parameters with fault ambiguity means that any parameter in the parameter set corresponds to two or more fault modes in the mapping table; Ground-Air Collaborative Health Management Unit: For the set of single fault parameters of the power system, a traversal query method is used for fault diagnosis; for the set of parameters with fault ambiguity of the power system, a fault diagnosis D-matrix model is established based on the multi-signal flow graph, and based on the TEAMS-RT diagnostic algorithm, a D-matrix merging method is used for power system fault diagnosis.

[0047] Preferably, the system further includes a health assessment unit; receiving fault diagnosis results sent by the ground-to-ground collaborative health management unit, and combining real-time telemetry data, historical test data, and flight data of the power system to comprehensively assess the health status of the power system; ground maintenance personnel adopt different maintenance and support strategies according to different health assessment levels of the power system, and after completing the maintenance and support of the power system, they reassess the health status of the power system using the same method.

[0048] The advantages of this invention compared to the prior art are:

[0049] (1) This invention proposes a system for aerospace power fault diagnosis and health assessment based on multi-signal flow graph and distributed TEAMS-RT algorithm, which realizes functions such as information processing, air-ground collaborative health management and health assessment;

[0050] (2) This invention proposes a fault diagnosis method based on data-driven and model-based approaches for the power system of a launch vehicle, thereby improving the accuracy of fault diagnosis;

[0051] (3) This invention proposes a data-driven health status assessment method for the power system of a launch vehicle, which improves the accuracy of the assessment. Detailed Implementation

[0052] The present invention will be further described below with reference to the embodiments.

[0053] A method for diagnosing aerospace propulsion faults includes:

[0054] Based on FME(C)A analysis, a set of failure modes and a set of measurement parameters for the propulsion system are obtained. The failure mode set includes all failure modes obtained through FME(C)A analysis, and the measurement parameter set includes the set of measurable parameters for any specific propulsion system. Using multi-signal flow graph analysis and commercial testability modeling software, the relationship between aerospace propulsion system failure modes and system measurement parameters is established. The detectable propulsion system parameters are divided into a set of single-failure parameters and a set of parameters with fault ambiguity. A mapping table between the failure mode set and the parameter set is obtained. A single-failure parameter set refers to a parameter set where each parameter corresponds to only one failure mode, while a parameter set with fault ambiguity indicates that each parameter in the set corresponds to two or more failure modes. The mapping table between the failure mode set and the parameter set refers to the correspondence between any failure mode in the failure mode set and the parameter in the parameter set representing that failure.

[0055] For a single set of fault parameters, a traversal query method is used for power system fault diagnosis.

[0056] For a fault diagnosis model based on a multi-signal flow graph with a fault ambiguity parameter set, a D-matrix merging method is used to diagnose power system faults based on the TEAMS-RT diagnostic algorithm.

[0057] After completing the fault diagnosis of the power system, a benchmark assessment model is constructed, and the health status of the power system is comprehensively assessed by combining real-time telemetry data, historical test data and flight data. Ground maintenance personnel adopt different maintenance and support strategies according to different health assessment levels of the power system. After completing the maintenance and support of the power system, the same method is used to reassess the health status of the power system.

[0058] A fault diagnosis system for aerospace propulsion system includes an information processing unit, a ground-based collaborative health management unit, and, if maintenance support is required, a health assessment unit.

[0059] (I) Information Processing Unit

[0060] Based on FME(C)A analysis, a set of power system fault modes and a set of measurement parameters are obtained. Based on the multi-signal flow graph analysis method, commercial testability modeling software is used to establish the relationship between aerospace power system fault modes and system measurement parameters, generating a mapping table between the fault mode set and the parameter set. In the parameter set, the detectable power system parameters are divided into two categories: a set of single fault parameters and a set of parameters with fault ambiguity. A single fault parameter set refers to a parameter set in which any parameter corresponds to only one fault mode in the mapping table, while a set of parameters with fault ambiguity indicates that any parameter in the parameter set corresponds to two or more fault modes in the mapping table.

[0061] (II) Collaborative Health Management Unit between Machine and Ground

[0062] For a single fault parameter set in the power system, the machine-ground collaborative health management unit adopts a traversal query method for fault diagnosis. For a parameter set with fault ambiguity in the power system, a fault diagnosis D-matrix model is established based on a multi-signal flow graph. Based on the TEAMS-RT diagnostic algorithm, the D-matrix merging method is used for power system fault diagnosis.

[0063] (1) Within a sampling period, the ground-based collaborative health management unit uses real-time telemetry data of the power system sent by the power system PHM data center, as well as historical test and flight data of the power system, to perform fault diagnosis for a single fault parameter set of the power system by adopting a traversal query method. The specific steps are as follows:

[0064] 1) For a single fault parameter set generated by the information processing unit, determine the envelope range of each parameter in the parameter set. The determination of the parameter envelope range mainly includes three methods: ① Determining based on the variation range of the actual physical quantity represented by the parameter. For example, the fuel loading amount of a liquid engine is determined by the design volume of the tank. The deviation range of the tank design capacity can be used to determine the deviation range of the fuel loading amount; ② Determining the variation envelope of the parameter based on historical test and flight data through data statistics. For example, the impulse of a liquid engine can be determined by statistically analyzing historical test data and flight test data of the engine test. ③ Determining based on existing knowledge reasoning and analysis. For example, the specific impulse of a liquid engine can be determined by the propellant specific impulse based on the propellant mass using the a priori formula. The variation range of the specific impulse can be determined by the variation range of the propellant impulse and the variation range of the propellant loading.

[0065] 2) Normalize the parameters in the single fault parameter set, compare the measured parameter values ​​with the parameter envelope, and use the state variable "1" to indicate normal when the parameter is within the envelope range; use the state variable "0" to indicate abnormal when the measured parameter value is outside the parameter change envelope range; transform the single fault parameter set into a power system health state characteristic set.

[0066] 3) Based on the mapping table between the fault mode set and the parameter set generated by the information processing unit, any parameter with a feature state of "1" in the parameter set is mapped to a single fault parameter set. When all the parameter feature states of a certain fault mapping in the single fault mode set are "1", it indicates that the fault has occurred. When all the parameter feature states of a certain fault mapping are "0", it indicates that the fault has not occurred.

[0067] 4) In the next sampling period, proceed to the next fault diagnosis and repeat steps (1) to (3) until the fault diagnosis is completed.

[0068] For a set of fault-ambiguous parameters, the machine-ground collaborative health management unit establishes a D-matrix fault diagnosis model based on a multi-signal flow graph. Building upon the TEAMS-RT diagnostic algorithm, it employs a D-matrix merging method for power system fault diagnosis. The specific diagnostic steps are as follows:

[0069] (1) First, establish a two-dimensional D m×n A matrix diagnostic model framework; a set of fault modes a m The fault modes (excluding fault modes mapped to a single fault parameter set) are represented by the header of the rows of a Dm×n matrix, starting from fault mode 1, fault mode 2, ... fault mode. m ; The set of fault ambiguity parameters b n Each parameter is used as D m×n The column headers of the matrix are named from measurement parameter 1, measurement parameter 2, ..., measurement parameter n; the elements of the rows and columns together form D. m×n A fault diagnosis matrix framework;

[0070] (2) Secondly, D m×n Any element D in ij The elements are normalized to obtain D. m×n Fault diagnosis matrix; based on the mapping table between the fault mode set and parameter set generated by the information processing unit, when fault mode j occurs, it can be detected by measurement point i, that is, when fault mode j can be characterized by measurement parameter i, then D m×n The corresponding D in ij The element is then set to "1"; for example, when fault mode j occurs, it cannot be detected by measurement point i, that is, when fault mode j cannot be characterized by measurement parameter i, then D m×nThe corresponding D in ij The element is then set to "0"; based on the mapping table between the fault mode set and parameter set generated by the information processing unit, all fault modes are... (1……n) With all measuring points (1……m) Normalizing the representational relationship between them yields D, which consists of state eigenvalues ​​0 and 1. m×n Fault diagnosis matrix;

[0071] (3) During the sampling period, the characteristic vectors of all measured parameters are obtained; the airborne health management unit utilizes the real-time telemetry data of the power system sent by the airborne PHM data center of the power system through the rocket-to-ground communication link, as well as the historical test and flight data of the power system, to merge all data into a non-unified set of collected parameters c. m Determine all acquisition parameters (1……m) The envelope range, the three methods for determining the parameter envelope are the same as the three parameter envelope determination methods in the above-mentioned single fault parameter set traversal query fault diagnosis method 1); for all collected parameters (1……m) Normalization is performed, and the collected parameter values ​​are compared with the parameter envelope. If the parameter is within the envelope range, a state variable "1" indicates normal operation; if the measured parameter value is outside the parameter variation envelope range, a state variable "0" indicates abnormal operation. The c value within one sampling period is then used to represent the parameter variation envelope. m The parameter set becomes an eigenvector d consisting of 0s and 1s. p (p = m);

[0072] (4) The feature vector d p Transform the rank to the vertical dimension and record the vector d. p Find the x-coordinate k (k∈[1, 2, 3...p]) of all rows containing "0" elements; delete D. m×n The fault diagnosis matrix in row K, D Kj (j∈[1, 2, 3...n]) The column vector containing elements with eigenvalues ​​of "0", for example, in D m×n In the matrix, if q columns and K rows containing "0" elements are deleted, the Dm×n fault diagnosis matrix is ​​merged into D. m×(n-q) 1-order matrix;

[0073] (5) Within one sampling period, the feature vector d of the collected parameters p With D m×(n-q) All column vectors D in the fault diagnosis matrix Kj (K∈[1, 2, 3…m]; j∈[1, 2, 3…(nq)]) are compared one by one, when d p =D Kj When (K∈[1, 2, 3…m]; j∈[1, 2, 3…(nq)]), it indicates that the ambiguity of the fault diagnosis is 1, Dm×(n-q) The fault diagnosis matrix header contains a unique corresponding fault mode. j Occurs when d p ≠D Kj When, then D m×(n-q) The fault mode corresponding to the fault diagnosis matrix header j It did not happen; when d p =D Kj (K∈[1, 2, 3…m]; j∈[1, 2, 3…(nq)])=D Kl When (K∈[1, 2, 3…m]; l∈[1, 2, 3…(nq)]), it indicates that the ambiguity of the fault diagnosis is 2, suggesting that the fault may be a fault mode. j Or it could be a fault mode. l Or, fault mode j and fault mode l occur simultaneously;

[0074] (6) For the fault mode set a m In all fault modes, when the fault diagnosis ambiguity is ≤2, the fault diagnosis will proceed to the next fault diagnosis in the next sampling period, repeating steps (3) to (6) until the fault diagnosis ends; when the fault diagnosis ambiguity is ≥3, it indicates that the system measurement points are set unreasonably or the set measurements are insufficient, and it is necessary to readjust or add detection points and regenerate the parameter set b with fault ambiguity. n A mapping table between the fault mode set and the parameter set, and the construction of a new D m×n Fault diagnosis matrix, repeat steps (3) to (6) until the fault diagnosis is completed.

[0075] (II) Health Assessment Unit

[0076] The health assessment unit receives the fault diagnosis results sent by the ground-to-ground collaborative health management unit, and combines them with real-time telemetry data, historical test data and flight data of the power system to conduct a comprehensive assessment of the health status of the power system. Ground maintenance and support personnel adopt different maintenance and support strategies according to different health assessment levels of the power system. After completing the maintenance and support of the power system, they reassess the health status of the power system using the same method.

[0077] The health assessment unit performs the assessment of the launch vehicle's health status in the following ways:

[0078] (1) Construct a benchmark evaluation model based on the characteristic parameters of the dynamic system under normal conditions. The specific steps are as follows:

[0079] 1) Conduct FME(C)A analysis on the launch vehicle's propulsion system to determine the categories of health characterization parameters for key components of the propulsion system; extract the health characterization parameters A for key components from historical test and flight data of the propulsion system, based on the categories of health characterization parameters. m ;

[0080] 2) For parameter set A m The monitoring data was used to extract features using wavelet / wavelet packet decomposition methods to form a feature set B that reflects the health status. m ;

[0081] 3) Use the isometric feature mapping (ISOMAP) method to map feature set B m A high-dimensional matrix with physical quantities is decomposed into a dimensionless low-dimensional characteristic matrix to obtain a set of characteristic parameters. This set of characteristic parameters is then used as a benchmark to evaluate the model matrix P. m .

[0082] Explanation: (1) The measurement parameters above are real-time telemetry data of the power system collected in real time within a sampling period; (2) The feature parameters here are feature extractions of historical test and flight data of the power system after wavelet / wavelet packet decomposition, forming feature values ​​that can reflect the health status; which types of historical test and flight data of the power system can be used as health characterization parameters of key components are determined by FME(C)A analysis of the power system.

[0083] (2) Within a sampling period, the health assessment unit receives real-time telemetry data of the power system sent by the ground-to-ground collaborative health management unit and obtains the feature matrix Q of the power system under the current state. n (m=n), the specific steps are as follows:

[0084] 1) Within a sampling period, the latest real-time telemetry data of the power system and historical test and flight data of the power system sent by the receiver-ground collaborative health management unit are used to obtain the set of monitorable parameters C of the key components of the launch vehicle power system. n ;

[0085] 2) For the set of monitorable parameters C n The monitoring parameters are used to extract features using the EMD method, forming a feature set E that reflects the health status. n ;

[0086] 3) Use the Local Multidimensional Scaling Analysis (LMDS) method to analyze the feature set E n A high-dimensional matrix with physical quantities is decomposed into a dimensionless low-dimensional characteristic matrix, yielding a set of characteristic parameters. This set of characteristic parameters is then used as the characteristic matrix Q. n .

[0087] (3) Calculate the benchmark evaluation model matrix P obtained in (1). m The eigenma matrix Q obtained in (2) n The distance between them is KL, and the magnitude of KL represents p. m (i) and q n The degree of closeness between (j) is given by the following formula.

[0088] (4) Using the distance KL obtained in (3), and combined with the fault mode level and number diagnosed by the ground-aircraft collaborative health management unit (the fault mode level is determined by FME(C)A analysis), and based on comprehensive engineering experience, a comprehensive judgment and assessment of the health status of the launch vehicle's propulsion system is made; for cryogenic liquid rocket propulsion systems, when KL = 0 and the ground-aircraft collaborative health management unit does not diagnose any faults, the propulsion system is in good health and meets the requirements for mission execution; when 0 < KL ≤ 2, and the number of non-Class I and Class II faults diagnosed by the ground-aircraft collaborative health management unit is ≤ 3 (the fault mode level is determined by FME(C)A analysis), the propulsion system is in abnormal health, but can still continue to perform the mission; when KL ≥ 3, the ground-aircraft collaborative health management unit... When the number of non-Class I and Class II faults diagnosed by the same health management unit is greater than 3, the power system health status is faulty, ground maintenance and support work is required, and the mission cannot continue. When the number of Class I and Class II faults diagnosed by the same machine-coordinated health management unit is ≥ 1, regardless of the value of KL, the power system health status is faulty, ground maintenance and support work is required immediately, and the mission cannot continue. The above are health assessment methods for cryogenic liquid power systems. In actual engineering applications, it is necessary to combine the technical characteristics, usage characteristics, and maintenance and support data of the actual power system to comprehensively adjust and correct the KL value, fault mode level, and quantity under different health states in order to adapt to the health assessment needs of different power models.

[0089] (5) Proceed to the next health assessment and repeat steps (1) to (4) until the fault diagnosis is completed.

[0090] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.

[0091] The parts of this invention not described in detail are common knowledge to those skilled in the art.

Claims

1. A method for diagnosing aerospace power system faults, characterized in that... include: Based on FME(C)A analysis, a set of power system failure modes and a set of measurement parameters are obtained; the set of failure modes includes all failure modes obtained through FME(C)A analysis, and the set of measurement parameters includes the set of existing measurable parameters for any power system. Based on the multi-signal flow graph analysis method, the relationship between fault modes of aerospace propulsion system and system measurement parameters is established. The detectable propulsion system parameters are divided into a set of single fault parameters and a set of parameters with fault ambiguity. A mapping table between the fault mode set and the parameter set is obtained. A single fault parameter set refers to a parameter set in which each parameter corresponds to only one fault mode. A parameter set with fault ambiguity refers to a parameter set in which each parameter corresponds to two or more fault modes. The mapping table between the fault mode set and the parameter set refers to the correspondence between any fault mode in the fault mode set and the parameter in the parameter set that represents this fault. For a single fault parameter set in a power system, a traversal query method is used for fault diagnosis; for a parameter set in a power system with fault ambiguity, a fault diagnosis D-matrix model is established based on a multi-signal flow graph, and a D-matrix merging method is used for power system fault diagnosis based on the TEAMS-RT diagnostic algorithm. The method of establishing a fault diagnosis D-matrix model based on multi-signal flow graphs, and using a D-matrix merging method for power system fault diagnosis based on the TEAMS-RT diagnostic algorithm, includes: (4.1) Within one sampling period, obtain the characteristic vectors of all measured parameters; using the real-time telemetry data of the power system transmitted by the airborne PHM data center of the power system through the rocket-to-ground communication link, as well as the historical test and flight data of the power system, merge all data into a unified set of acquired parameters c. m Determine the envelope range of all acquired parameters, normalize all acquired parameters, and compare the acquired parameter values ​​with the parameter envelope. If the parameter is within the envelope range, use the state variable "1" to indicate normal operation; if the measured parameter value is outside the parameter variation envelope range, use the state variable "0" to indicate abnormal operation; calculate c within one sampling period. m The parameter set becomes an eigenvector d consisting of 0s and 1s. p , p=m; (4.2) The eigenvector d p Transpose the vector to a vertical dimension and record the vector d. p The x-coordinate k of all rows with "0" elements (k∈[1, 2, 3) …… p]); Delete D m×n The fault diagnosis matrix in row K, D Kj (j∈[1,2,3) …… [n]) is the column vector containing elements with eigenvalues ​​of "0". Assume that in D... m×n In a matrix, if q columns and K rows containing "0" elements are deleted, then D will be... m×n The fault diagnosis matrix is ​​merged into D m×（n-q） 1-order matrix; (4.3) Within one sampling period, the feature vector d of the collected parameters is... p With D m×（n-q） All column vectors D in the fault diagnosis matrix Kj (K∈[1,2,3) …… m];j∈[1, 2, 3 ... …… (nq)]) compare them one by one, when only d is satisfied p = D Kj (K∈[1,2,3) …… m];j∈[1, 2, 3 ... …… When (nq)]), it indicates that the ambiguity of the fault diagnosis is 1, D m×（n-q） The fault diagnosis matrix header contains a unique corresponding fault mode j, which occurs when d p ≠D Kj When, then D m×（n-q） The corresponding fault mode j in the fault diagnosis matrix header did not occur; when d p = D Kj (K∈[1,2,3) …… m];j∈[1, 2, 3 ... …… (nq)]) = D Kl (K∈[1,2,3) …… m];l∈[1, 2, 3 ... …… When (nq)]), it indicates that the ambiguity of the fault diagnosis is 2, which means that the fault may be fault mode j, fault mode l, or fault mode j and fault mode l may occur simultaneously. (4.4) For the set of failure modes a m For all fault modes, when the fault diagnosis ambiguity is ≤2, the fault diagnosis will proceed to the next fault diagnosis in the next sampling period, repeating steps (4.1) to (4.3) until the fault diagnosis ends; when the fault diagnosis ambiguity is ≥3, it indicates that the setting of the measurement points used for the measurement parameters is unreasonable or the number of measurement points is insufficient, so it is necessary to readjust or add measurement points, and regenerate the parameter set b with fault ambiguity. n A mapping table between the fault mode set and the parameter set, and the construction of a new D m×n After the fault diagnosis matrix is ​​completed, repeat steps (4.1) to (4.3) until the fault diagnosis is finished.

2. The method according to claim 1, characterized in that: The method of traversing and querying parameters to diagnose a single fault in a power system mainly includes: (2.1) Determine the envelope range of each parameter in a single parameter set; (2.2) Normalize the parameters in the single fault parameter set, compare the measured parameter values ​​with the corresponding envelope range, and use the state variable "1" to indicate normal when the parameter is within the envelope range; use the state variable "0" to indicate abnormal when the measured parameter value is outside the parameter change envelope range; transform the single fault parameter set into a power system health state characteristic set. (2.3) Based on the mapping relationship table between the fault mode set and the single parameter set, any parameter with a characteristic state of "1" in the parameter set is mapped to the single fault parameter set. When all the parameter characteristic states of a certain fault mapping in the single fault mode set are "1", it indicates that the fault has occurred. When any parameter with a characteristic state of "0" in all the parameters of a certain fault mapping, it indicates that the fault has not occurred. (2.4) In the next sampling period, proceed to the next fault diagnosis and repeat steps (2.2) to (2.4) until the fault diagnosis is completed.

3. The method according to claim 1, characterized in that: For a set of parameters in a dynamic system with fault ambiguity, a fault diagnosis D-matrix model is established based on a multi-signal flow graph. The model establishment methods include: (3.1) Establish a two-dimensional D m×n A matrix diagnostic model framework; a set of fault modes a m Fault modes that do not contain a single set of fault parameters are used as D. m×n The headers of the columns in the matrix represent fault mode 1, fault mode 2, ..., fault mode m, respectively; the set of fault ambiguity parameters b... n Each parameter is used as D m×n The column headers of the matrix are measurement parameter 1, measurement parameter 2, ..., measurement parameter n, respectively; the elements of the rows and columns together form D. m×n A fault diagnosis matrix framework; (3.2) D m×n Any element D in ij The elements are normalized to obtain D. m×n Fault diagnosis matrix; based on the mapping table between the fault mode set and the parameter set, when fault mode j occurs, it can be detected by measurement point i, that is, when fault mode j can be characterized by measurement parameter i, then D m×n The corresponding D in ij The element is then set to "1"; for example, when fault mode j occurs, it cannot be detected by measurement point i, that is, when fault mode j cannot be characterized by measurement parameter i, then D m×n The corresponding D in ij The element is then set to "0"; in this way, the representational relationship between all fault modes and all measurement points is normalized to obtain D, which consists of state characteristic values ​​0 and 1. m×n Fault diagnosis matrix, also known as fault diagnosis D-matrix model.

4. The method according to claim 1, characterized in that: After completing the fault diagnosis of the power system, a benchmark assessment model is constructed, and the health status of the power system is comprehensively assessed by combining real-time telemetry data, historical test data and flight data. Ground maintenance and support personnel adopt different maintenance and support strategies according to different health assessment levels of the power system, and after completing the maintenance and support of the power system, the health status of the power system is comprehensively assessed again.

5. The method according to claim 4, characterized in that: The comprehensive assessment of the health status of the power system includes: (5.1) Construct a benchmark evaluation model P based on the characteristic parameters of the dynamic system under normal conditions. m ; (5.2) Within one sampling period, based on the latest real-time telemetry data of the power system, obtain the feature matrix Q of the power system in its current state. n m=n; (5.3) Calculate the benchmark evaluation model matrix P obtained in (5.1). m The feature matrix Q obtained in (5.2) under the current state n The distance between them is KL, and the magnitude of KL represents P. m (i) and Q n The degree of closeness between (j); P m (i) and Q n (j) represents matrix P m With Q n Elements in; (5.4) Using the distance KL obtained in (5.3) and combined with the fault mode level and number, the health status of the vehicle's power system is comprehensively judged and evaluated using the power system health assessment criteria.

6. The method according to claim 5, characterized in that: The benchmark evaluation model matrix P m include: (6.1) Conduct FME(C)A analysis on the launch vehicle power system to determine the categories of health characterization parameters of key components of the launch vehicle power system; extract the health characterization parameters A of key components from the historical test and flight data of the power system according to the categories of health characterization parameters. m ; (6.2) For the health characterization parameter set A m The monitoring data was used to extract features using wavelet / wavelet packet decomposition methods to form a feature set B that reflects the health status. m ; (6.3) Use the ISOMAP method to map the feature set B m A high-dimensional matrix with physical quantities is decomposed into a dimensionless low-dimensional characteristic matrix to obtain a set of characteristic parameters. This set of characteristic parameters is then used as a benchmark to evaluate the model matrix P. m .

7. The method according to claim 5, characterized in that: The feature matrix Q of the system in the current state of the power system is obtained. n m=n includes: (7.1) Within one sampling period, the set of monitorable parameters C of the key components of the launch vehicle's power system is obtained by receiving the latest real-time telemetry data of the power system. n ; (7.2) For the set of monitorable parameters C n The monitoring parameters are used to extract features using the EMD method, forming a feature set E that reflects the current health status of key components of the power system. n ; (7.3) The feature set E is analyzed using the Local Multidimensional Scaling (LMDS) method. n A high-dimensional matrix with physical quantities is decomposed into a dimensionless low-dimensional characteristic matrix, yielding a set of characteristic parameters. This set of characteristic parameters is then used as the characteristic matrix Q. n .

8. The method according to claim 5, characterized in that: The formula for distance KL is as follows: 。 9. The method according to claim 5, characterized in that: The health assessment criteria for propulsion systems include: for cryogenic liquid rocket propulsion systems: When KL=0 and no faults are diagnosed, the power system is in good health and meets the requirements for performing the task. When 0 < KL ≤ 2, and the number of non-Class I and Class II faults diagnosed is ≤ 3, the power system health status is abnormal, but it can still continue to perform tasks; the fault mode level is obtained by FME(C)A analysis. When KL≥3 and the number of non-Class I and Class II faults diagnosed is >3, the power system health status is faulty, ground maintenance and support work is required, and the mission cannot continue. When the number of diagnosed Class I and Class II faults is ≥1, regardless of the value of KL, the health status of the power system is faulty, and ground maintenance and support work needs to be carried out immediately. The mission cannot continue.

10. An aerospace power fault diagnosis system based on the aerospace power fault diagnosis method according to claim 1, characterized in that: Includes an information processing unit and a machine-ground collaborative health management unit; Information Processing Unit: Based on FME(C)A analysis, the set of power system fault modes and the set of measurement parameters are obtained; based on the multi-signal flow graph analysis method, the relationship between the fault modes of the aerospace power system and the system measurement parameters is established, and a mapping table between the fault mode set and the parameter set is generated; in the parameter set, the detectable power system parameters are divided into two categories: the power system single fault parameter set and the parameter set with fault ambiguity; the single fault parameter set refers to the parameter set in which any parameter corresponds to only one fault mode in the mapping table, and the parameter set with fault ambiguity refers to the parameter set in which any parameter corresponds to two or more fault modes in the mapping parameter table; Machine-Ground Collaborative Health Management Unit: For a single fault parameter set of the power system, a traversal query method is adopted for fault diagnosis. For a parameter set of the power system with fault ambiguity, a fault diagnosis D matrix model is established based on a multi-signal flow graph. Based on the TEAMS-RT diagnostic algorithm, a D matrix merging method is used for power system fault diagnosis.

11. The system according to claim 10, characterized in that... It also includes a health assessment unit; receiving fault diagnosis results sent by the ground-to-ground collaborative health management unit, and combining real-time telemetry data, historical test data and flight data of the power system to conduct a comprehensive assessment of the health status of the power system; ground maintenance and support personnel adopt different maintenance and support strategies according to different health assessment levels of the power system, and after completing the maintenance and support of the power system, they reassess the health status of the power system using the same method.

Images