A method and apparatus for fault editing in health management of liquid rocket engine test rigs

Abstract

This invention relates to the field of liquid rocket engine test bench health management systems, and discloses a method for editing fault signals on an engine test bench. The method includes studying the fault mechanism of a liquid rocket engine test bench and collecting historical state signals of the test bench; analyzing the engine test bench fault model, dissecting the fault manifestations and influencing mechanisms; establishing a mapping relationship between the normal state of the engine test bench and various failure modes; and verifying the editable fault signals of the engine test bench. The generated editable fault signals are received by the liquid rocket engine test bench health management system and used to complete a comprehensive functional test of the system. This invention can simulate fault signals such as liquid circuit leakage and valve blockage in liquid rocket engine test benches, providing an effective approach for studying liquid rocket engine test bench health management systems.

Description

Technical Field

[0001] This invention relates to the field of liquid rocket engine health management systems, and designs a fault editing method and system applicable to the verification of various technical indicators of rocket engine test bench health management systems. Background Technology

[0002] Rocket engine test rigs are characterized by long development cycles, high costs, and significant risks. Therefore, their health management systems must undergo rigorous validation before being formally applied to liquid rocket engine test rigs. In the health performance evaluation of rocket engine test rigs, some extreme operating conditions and complex faults are difficult or prohibitively expensive to replicate during testing. Therefore, it is necessary to research editable state signals for rocket engine test rigs, develop a fault editing generation device suitable for the test rig, and verify the fault detection capabilities of the rig's health management system.

[0003] The liquid rocket engine test health management system monitors various parameters throughout the entire test process, such as the dynamic performance of each sensor and measurement channel, equipment operation status, safety, and engine condition. It promptly takes corresponding measures for any test problems and faults detected, such as issuing reminders, warnings, or even suspending operations, to ensure the smooth progress of the test and the timely and high-quality completion of the test mission. The fault editing device edits faults during the engine test and converts the edited faults into corresponding electrical signals, which are then received by the test health management system. This verifies the accuracy, timeliness, and appropriateness of the health management system's fault feedback.

[0004] Chinese patent CN201510498261.5 discloses a design method for a fault response generator for a liquid rocket engine test bench. This method assesses the fault severity based on historical data to obtain fault diagnosis information; establishes a fault severity level database; maps fault information to the fault level database to obtain a fault response decision model; and implements fault intervention measures for different levels. This patent focuses on the liquid rocket engine itself, emphasizing how users can proactively develop maintenance plans and implement intervention decisions when a fault occurs. This invention, however, aims to determine whether the health management system in the test bench is operating normally.

[0005] Chinese patent CN201910396984.2 discloses a hardware-in-the-loop (HIL) simulation method and apparatus for a liquid rocket engine thrust adjustment system. This patent calculates the current state of the liquid rocket engine in real time based on a simulation program, and outputs corresponding signals to the physical thrust adjustment system and load simulation device via an interface board. The physical thrust adjustment system performs thrust adjustment actions based on sensor signals and load torque. The board measures the current operating state of the physical thrust adjustment system and inputs the value into the liquid rocket engine simulation program for iterative calculation. The main drawback of this invention is that it only considers the thrust signal in the liquid rocket engine and does not take into account parameters such as pipeline flow, pressure, and temperature in the engine test bench health management system.

[0006] Whether it is to study the fault manifestations of liquid engine test benches or to verify the protection capabilities of the test bench health management system and the sensitivity and reliability of fault detection, a simulation system capable of editing various internal faults of liquid engine test benches is needed to meet the needs of national defense science and technology. Summary of the Invention

[0007] This invention addresses the difficulty in establishing accurate dynamic mathematical models for fault detection and diagnosis on liquid rocket engine test rigs. It employs logical reasoning theory to determine the fault manifestations that cause the system based on the system's input and observed actual behavior. This invention provides a hardware-in-loop (HIL) simulation device for performance verification of the health management system of liquid rocket engine test rigs, thereby enabling the evaluation of the test rig's health management system's protection capabilities, fault monitoring sensitivity, and reliability.

[0008] To achieve the above objectives, this invention proposes a fault editing method for health management of liquid rocket engine test rigs, comprising the following steps:

[0009] S1: Analyze the fluid circuit of the rocket engine test bench, identify the location and specific cause of the fault, and study the engine failure mechanism. Acquire historical normal state signal data of the test bench, analyze the signal patterns, and use this data as the foundation for generating editable fault signals.

[0010] S2: Analyze the failure mode model, including the top-down failure effects of each failure and the parameter manifestations of each failure, summarize the logical reasoning of the above failures, analyze the positive impact and catastrophicness, and study the impact of test bench measurement points.

[0011] S3: Establish state mode mapping relationships for linear and nonlinear mappings. Normal measured signals (including temperature, pressure, and flow rate) are used to generate editable fault signals through defined mapping relationships. Since fault editing does not alter the mechanical model of the dynamic test bench, when a fault occurs, the relevant fault signals are integrated into the normal signals through linear superposition, first-order and quadratic function design, etc., to simulate the state signals at the time of the fault.

[0012] S4: Verify the accuracy of fault editable signals using a support vector data description model based on mutated particle swarm optimization.

[0013] S5: The generated fault signals are converted into electrical signals using HIL and sent to the health management system of the liquid rocket engine to verify whether the system's feedback on various faults is timely and accurate.

[0014] On the device for implementing fault editing:

[0015] This invention provides a fault editing and monitoring platform. The platform consists of a user host computer and a PXI system. The PXI industrial platform comprises three basic parts: a PXI chassis, a PXI embedded real-time controller, and PXI peripheral expansion modules, meeting the system's real-time measurement and control requirements. The host computer interacts with the NI-PXI embedded controller via Ethernet, and the PXI platform communicates with the underlying fault simulation and transmission equipment. According to the overall design scheme of the fault editing system, the platform hardware architecture is as follows: Figure 8 As shown, the upper-level monitoring platform can monitor the fault occurrence process in real time.

[0016] This invention also provides a low-level fault generation device. In semi-physical fault simulation of engine testing, based on the fault model, including the fault mathematical model and the fault phenomenon model, fault settings and phenomenon simulation are realized, thereby controlling each fault simulation sub-module to inject corresponding faults at designated measurement points. The lower-level computer mainly responds to the fault simulation commands from the upper level, generates and sends fault electrical signals, and realizes synchronous simulation of multiple points and multiple faults based on the requirements of real-time dynamic fault simulation. The low-level fault device adopts a modular design concept, mainly divided into a main control module, a valve fault simulation module, a flow fault simulation module, a temperature fault simulation module, etc. Among them, the main control module not only needs to respond to the start and stop commands of the upper-level fault simulation device from the host computer, but also coordinates the work of other fault simulation modules. Other types of fault simulation modules are used to generate and send corresponding fault signals to realize the fault simulation of the liquid rocket engine test system. In addition, the control logic of the entire data transmission and the PXI interface control logic are integrated into the FPGA, which significantly improves the utilization of FPGA logic resources, simplifies hardware circuit design, and enhances the integration and reliability of the modules. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0018] Figure 1 Schematic diagram of fault editing design and health monitoring system verification for liquid rocket engine test bench.

[0019] Figure 2 Schematic diagram of the liquid circuit system of the liquid rocket engine test rig

[0020] Figure 3 Failure Mechanism and Failure Mode Analysis of Liquid Engine Test Bench

[0021] Figure 4 Liquid rocket engine test stand fault signal simulation

[0022] Figure 5 Historical data and fault editing signals of measurement point Pto2

[0023] Figure 6 Verification of ice blockage editing signal at measuring point Pto2

[0024] Figure 7 Hardware connection diagram of fault simulation device

[0025] Figure 8 Fault simulation device architecture schematic diagram

[0026] Figure 9 Appearance and structure of fault simulation device

[0027] Appendix Figure 2 The markings are as follows: (1) Pto1 represents the internal pressure of air tank 1; (2) Pto2 represents the internal pressure of air tank 2; (3) Ptf represents the internal pressure of kerosene tank 1; (4) Ptf1 represents the internal pressure of kerosene tank 2; (5) qmo represents the air flow rate of the pipeline of air tank 1; (6) qmo1 represents the air flow rate of the pipeline of air tank 2; (7) P44 represents the air pressure at the vacuum chamber inlet; (8) T42 represents the air temperature at the vacuum chamber inlet; (9) qmf1 represents the kerosene pipeline flow rate of kerosene tank 1; (10) qmf2 represents the kerosene pipeline flow rate of kerosene tank 2; (11) P47 represents the kerosene inlet pressure of the vacuum chamber. Detailed Implementation

[0028] To make the implementation process of this invention more detailed and understandable, the technical solution of this invention will be described in complete and detailed form below with reference to examples of this invention. Obviously, the embodiments described herein are only one embodiment of this invention, and not all embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention, such as... Figure 1 As shown, this invention discloses a performance verification scheme for a health management system for a liquid rocket engine test bench.

[0029] The specific technical implementation plan of this solution (see...) Figure 1 The process includes the following steps:

[0030] S1: Construct a liquid rocket engine test bench model and analyze the liquid circuit conditions of the test bench (see...). Figure 2 The failure mechanism of the test bench was studied through model research.

[0031] S2: Acquire historical test data from the rocket engine test bench, including parameters such as fluid flow rate, pressure, temperature, fuel storage tank pressure, and valve control current, i.e., test data at measurement points under normal conditions (see...). Figure 4 ).

[0032] S3: Analyze the failure mode model of the test bench, dissect its failure effects, parameter manifestations, and failure control, etc. (see...) Figure 3 The failure modes of the hydraulic circuit on the test bench are analyzed as follows:

[0033] S31: Filter ice blockage causes abnormal filter pressure and fluctuating pipeline flow. Detect air tank pressure Pto2, pipeline flow qmo1, and pipeline pressure P44. Pto2 pressure shows a sudden increase, while P44 pressure and qmo1 flow show a sudden decrease. Preventive measures include regular cleaning and temperature control.

[0034] S32: Kerosene blockage in the filter caused abnormal filter pressure and fluctuating pipeline flow. The pressure of the kerosene storage tank Ptf2, the pipeline flow qmf2, and the pipeline pressure P47 were checked. It was found that the pressure of Ptf2 increased, while the pressure of P47 and the flow of qmf2 decreased slowly. The parameter change time was about 3 seconds. The preventive measures are to clean regularly and control the purity of kerosene.

[0035] S33: Pipeline leakage caused abnormal pipeline pressure and temperature. The pipeline pressure P44 and temperature T42 were detected. It was found that during the pressure holding period, the pressure P44 and temperature T42 decreased slowly, with the parameter change time being about 18 seconds.

[0036] S34: The pipeline valve cannot be opened, resulting in no control current in the valve body. The current If2 is detected and no pulse is found in If2.

[0037] S4: Conduct impact analysis of test bench measurement points using inductive logical reasoning, bottom-up positive analysis, and impact and lethality analysis.

[0038] S5: Establish a state mode mapping library for linear or nonlinear mappings. Use measured signals such as temperature, pressure, and flow rates from the test bench under normal conditions to edit fault modes, generating editable fault signals for liquid system leaks, valve leaks, etc. Common liquid rocket engine test bench fault mapping relationships are as follows:

[0039] S51: According to the fault model, the filter ice blockage fault mode manifests as follows:

[0040]

[0041] In the formula n is the signal length; Pto'2(t) and P' 44 (t) represents the pressure fault performance parameter at the two measuring points; qmo1'(t) is the flow fault performance parameter; This represents the average pressure of the air storage tank. This represents the average pipeline pressure. It is the average airflow rate in the pipeline.

[0042] S52: Based on the analysis of the kerosene blockage fault symptoms in the filter, the mapping relationship is as follows:

[0043]

[0044] In the formula Ptf′2(t) and qmf'(t) represent the pressure and flow rate fault parameters, respectively; and These represent the average pressure and flow rate of the kerosene storage tank, respectively.

[0045] S53: Analysis of pipeline leakage faults shows that during pressure holding, the amount of gas introduced is approximately 1.6 times the amount of gas introduced previously. The ideal state equations for the pipeline before and after pressurization are as follows:

[0046] P1V=n1RT1 (3)

[0047] P2V=n2RT2 (4)

[0048] In the formula, P1 and P2 represent the pressure (Pa) before and after pressure holding, respectively; V1 and V2 represent the volume (m³) before and after pressure holding, respectively. 3 ); n1 and n2 represent the amount of substance (mol) before and after the pressurized gas is introduced; R is the molar gas constant (J / mol / K); T1 and T2 represent the temperature (K) before and after pressurization.

[0049] Based on historical data, when P1 = 3.11 MPa, the pipeline temperature T1 = 33.42℃; after gas is introduced during pressure holding, P2 = 5 MPa. By combining the equations, it can be seen that during the gas introduction and pressure holding process, as the pressure increases, the temperature rises to 309.65 K, which means the pipeline gas temperature rises by 3.1℃. If there is a pipeline leak, the temperature T42 will slowly decrease, but will still be higher than the initial temperature, requiring periodic pressure holding checks.

[0050]

[0051] In the formula, Pio'(t) and Tio'(t) are the pressure and temperature fault parameters, respectively; t represents the time; Tio(t) and Pio(t) are the temperature and pressure of the pipeline under normal conditions, respectively.

[0052] S54: Analysis of the fault symptoms of the pipeline valve failing to open reveals...

[0053] If2(t)=0 (7)

[0054] S6: Verify the correctness of editable fault signals using a support vector data description model based on mutant particle swarm optimization. The specific verification steps are as follows:

[0055] S61: Divide the normal state time series into 32 samples, each containing 1024 points, as training samples. To address the issue of a small sample size, add white Gaussian noise to the data to increase the sample size.

[0056] S62: Process the noisy data using step S61, and merge it with the original samples to obtain 64 samples. Perform time-domain, frequency-domain, and other symptom features on each sample to obtain a training set containing 18 features for each sample.

[0057] S63: Fault data processing is similar to normal data processing. It is worth noting that fault data, as a test set, does not need to be increased by adding noise to expand the sample size. Finally, a test set of 32 samples is obtained.

[0058] S64: Develop a mutant particle swarm optimization algorithm to optimize the penalty factor C and kernel parameter δ (i.e., particle position x) in the support vector data description model. id The particle velocity and position are updated as shown in the equation. The optimal hyperparameter combination is obtained by training the Gaussian kernel SVDD model using different combinations of the penalty factor C and the kernel parameter δ.

[0059]

[0060] In the formula, i = 1, 2... N, where N represents the number of particles in the particle swarm; d = 1, 2... D, where D represents the dimension of each particle; r1 and r2 are random numbers distributed between 0 and 1; c1 is used to adjust the step size of the particle's adjustment towards its individual optimal position, and c2 is used to adjust the step size of the particle's adjustment towards the global optimal position; w is the inertia factor of the particle swarm (w ≥ 0).

[0061] S65: After normalizing the features obtained from the training and test sets, train and test the optimized SVDD model separately. The model training objective is as follows:

[0062]

[0063] In the formula ζ i represents the slack variable, used to control the influence of outliers on the decision boundary; C is used to balance the hypersphere volume and decision error.

[0064] The radius R of the hypersphere can be obtained using language multipliers and the KKT theorem. R can be derived from any support vector x. k The distance to the center is calculated using the formula, which yields the decision boundary for the normal state sample. For the sample to be tested, the distance between the sample and the center of the hypersphere is determined by R. 2 The size relationship between the two values ​​determines whether a sample belongs to the target sample. If the formula holds true, the sample to be tested is a target sample; otherwise, it is a non-target sample and is rejected.

[0065]

[0066] S66: Train the SVDD model using the training set, use the model to test the test set, obtain the results, and verify the accuracy of the fault signal.

[0067] S7: A fault editing signal based on HIL is generated and received and verified by the test health management system to complete a comprehensive functional test of the health monitoring system. The specific test steps are as follows:

[0068] S71: The host computer configures the simulation program and establishes the connection between the simulation program and the I / O channels of the interface board.

[0069] S72: According to the test plan, the simulated fault is converted into voltage, current or digital values ​​in the form of sensors and input into the health management system of the liquid rocket engine.

[0070] S73: Observe the system's response and handling methods, and test whether the health management system's feedback on faults is accurate, its response is timely, and its handling is appropriate.

[0071] Example 1

[0072] This embodiment provides a case study of filter ice blockage failure on a liquid rocket engine test bench:

[0073] S1: Analyze historical test data, such as normal test data for Pto2 measuring points. Figure 5 As shown in (a), under normal test conditions, the pressure fluctuation is between 2.96 and 3.17 MPa. At the moment of engine ignition, due to air consumption, the pressure in the air line drops significantly. Fault Mode and Effects Analysis (FMEA) is used to generate a fault signal for the pressure at the Pto2 upstream of the filter. The specific analysis is as follows:

[0074] S11: When the air pipeline operates under temperature conditions of -40℃ to 75℃ (temperature control unit capacity), moisture in the air condenses to form ice fragments. The accumulation of these ice fragments leads to pipeline blockage. According to Failure Mode and Effects Analysis (FMEA), the increased pressure at the filter inlet manifests as a sudden change in the influencing factor.

[0075] S2: Statistical analysis of historical normal test run data shows that each parameter always varies within a certain range. Based on the formula Pto2, fault simulation at the measuring point is performed as follows: Figure 5 As shown in (b), due to the low-temperature working conditions, trace amounts of water vapor in the air freeze to form ice blockage. At this time, the pressure of the air storage tank will rise instantly, and the pressure of the pipeline after the ice blockage will decrease. The pressure fluctuation at this time is between 3.05 and 3.45 MPa.

[0076] S3: Fault signal verification is performed using test data from a certain type of liquid engine. Following the steps in the specific implementation method, data augmentation, sample partitioning, feature extraction, and feature optimization are performed to obtain a sample set for training and testing the support vector data description model.

[0077] S4: A mutant particle swarm optimization (SVDD) approach is used to optimize key parameters, achieving the final optimization goal through collaborative teamwork. The resulting optimal key parameters are used to build an evaluation model. After 50 iterations, the penalty factor C is found to be 0.51, and the kernel parameter δ is 120.51.

[0078] S5: Train the SVDD model using the above optimal hyperparameters, feed the test set into the trained validation model, and obtain the results. Figure 6 As shown in the figure. The solid line in the figure represents the radius of the hypersphere, and the data within the radius represents the sample under normal conditions; the dashed line represents the test sample. The figure shows that the anomaly begins at 5.4 seconds, which is consistent with... Figure 5 If the timing of the failure is consistent, the accuracy of the valve ice blockage failure simulation can be verified.

Claims

1. A method for generating editable fault signals, characterized in that: Includes the following steps S1: Analyze the hydraulic circuit of the test bench and obtain data on the normal state of the test run; S2: Through failure mode and effects analysis, a failure mode model is obtained, including failure effects, parameter performance, and failure control. S3: Establish a state-mode mapping relationship using linear or nonlinear mappings. Through this mapping relationship, editable fault signals are obtained from normal measured data. The mapping library is as follows: S31: Filter ice blockage mapping relationship: In the formula n is the signal length; Pto'2(t) represents the pressure fault performance parameter; P′ 44 (t) represents the pressure fault performance parameter; qmo1'(t) represents the flow fault performance parameter; This represents the average pressure of the air storage tank. This represents the average pipeline pressure. It is the average airflow rate in the pipeline; S32: Filter kerosene clogging mapping relationship: In the formula Ptf′2(t) and qmf'(t) represent the pressure and flow rate fault parameters, respectively; and These represent the average pressure and flow rate of the kerosene storage tank, respectively. S33: Pipeline Leakage Fault Mapping Relationship: In the formula, Pio'(t) and Tio'(t) are the pressure and temperature fault parameters, respectively; Tio(t) and Pio(t) are the temperature and pressure of the pipeline under normal conditions, respectively. S34: Pipeline valve fault mapping relationship: If2(t)=0 S4: Verify the correctness of the fault-editable signal using a support vector data description model based on mutant particle swarm optimization; S5: The fault editing signal based on HIL is received and verified by the test health management system, realizing the evaluation of the protection capability, fault monitoring sensitivity, and reliability of the test bench health management system.

Images