Semi-physical method and device for combining engine fault diagnosis and fault-tolerant control

Abstract

This invention discloses a semi-physical method and apparatus for fault diagnosis and fault-tolerant control of combined engines, belonging to the field of aerospace engineering technology. The method includes: a real-time simulator generating virtual data; the simulator injecting programmable faults and reconstructing fault channel data by weighted fusion according to a formula based on spatial neighborhood relationships; and a controller acquiring multi-source data and calculating based on the PCA algorithm. T2 Fault detection is performed using SPE statistics, the fault source is located through contribution values, and fault-tolerant control commands are generated within 20ms to achieve closed-loop control. The device includes a simulator, simulator, controller, test management system, and communication bus for performing the above steps. This invention achieves high-fidelity fault injection, hard real-time calculation and diagnosis, and multi-level fault-tolerant control closed-loop verification from sensor to system level, providing an efficient testing platform for combined engine health management.

Description

Technical Field

[0001] This invention belongs to the field of aerospace engineering technology, specifically relating to a semi-physical method and apparatus for fault diagnosis and fault-tolerant control of combined engines. Background Technology

[0002] Combined-cycle engines are the core power units of space launch vehicles, and their operation is characterized by high dynamics, strong coupling, and multiple variables. Especially in hydrogen peroxide combined-cycle engines, which employ staged catalytic decomposition cycles, the system structure is complex, including turbopumps, gas generators, dual thrust chambers, regulating valves, and intricate piping. Under transient conditions such as startup and variable thrust, parameters such as pressure, temperature, and flow rate change drastically, and component failures (such as leaks, blockages, and reduced pump efficiency) can easily lead to abnormal thrust or even catastrophic consequences. Therefore, developing a reliable fault diagnosis and fault-tolerant control (FTC) system and conducting thorough and credible verification of it is crucial to ensuring the safe operation of the engine.

[0003] Currently, the verification of fault diagnosis and fault-tolerant control systems for combined engines mainly relies on solid digital in-process (SIL) simulation and ground-based hot-fire testing. SIL simulation is low-cost, but it cannot be integrated with real controller hardware, making it difficult to simulate actual communication delays and I / O noise, resulting in insufficient engineering reliability of the verification results. While ground-based hot-fire testing is the ultimate verification method, it is extremely costly and risky, and it is difficult to safely and repeatedly inject multiple faults, limiting the research on fault mechanisms and the systematic verification of fault-tolerant strategies.

[0004] While existing hardware-in-the-loop (HIL) simulation technology can be integrated with real controllers, it still has the following shortcomings when dealing with highly complex systems such as combined engines: 1. The sensor signal has low analog fidelity, making it difficult to achieve channel-by-channel programmable physical-level fault injection; 2. The engine dynamics model is complex, and the traditional HIL platform can hardly meet the hard real-time requirements of real-time model solving and fast response of the control system at the same time. 3. The verification process is mostly limited to fault detection or simple control switching, lacking the ability to conduct full closed-loop, multi-level collaborative verification from sensor-level diagnosis to system-level fault-tolerant control.

[0005] Therefore, there is an urgent need to build a hardware-in-the-loop simulation platform that is hard real-time, high-fidelity, and supports multi-level fault injection and closed-loop verification, so as to systematically verify the fault diagnosis and fault-tolerant control algorithms of the combined engine. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a semi-physical method and apparatus for combined engine fault diagnosis and fault-tolerant control, which can achieve high-fidelity sensor fault injection, hard real-time model solving and diagnostic decision-making, as well as multi-level collaborative fault-tolerant control closed-loop verification.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a semi-physical method for combined engine fault diagnosis and fault-tolerant control, comprising the following steps: S1. Real-time model solving and virtual data generation: By running the combined engine dynamics model and fault model through a real-time simulator, virtual real-time data containing at least 32 pressure data points is generated with a cycle of less than 5ms. S2. Synchronous Fault Injection and Signal Simulation: The test management system issues a fault configuration command, and the 32-channel pressure scanning valve real-time acquisition simulator receives the pressure data and injects programmable sensor faults into the specified channels according to the command to generate a pressure simulation signal containing fault characteristics. S3. Distributed sensor diagnosis and data reconstruction: The simulator performs multi-source data fusion fault diagnosis and data reconstruction processing on the pressure simulation signal to generate sensor diagnosis results; The processing includes: S31. Establish the spatial neighborhood matrix of pressure measurement points based on the engine's physical layout. If the channel i With channel j Adjacent, then the correlation coefficient ,otherwise ; S32, for the channel i exist k Sample value at time Based on sliding window judgment if If so, the data in that channel is determined to be abnormal. For the most recent The mean of each sampling point This represents the historical standard deviation of the channel; S33, when determining When a channel fails, it is reconstructed using weighted fusion data from normal channels within its spatial neighborhood. The reconstructed value is... The calculation formula is: ; in, for The neighborhood set of the channel, For neighborhood channels The measured value, for The confidence level of the channel is 1 for normal operation and 0 for fault. S4. Multi-source data acquisition and system-level diagnosis: The EEC-3 fault diagnosis integrated controller acquires the pressure simulation signal, the virtual real-time data, the sensor diagnostic results and the actuator feedback data, and runs a fault diagnosis algorithm based on principal component analysis; The algorithm includes: S41. Acquired multidimensional observation vectors Standardization process: ; in, μ and σ These are the mean vector and standard deviation vector of the normal operating condition data obtained through offline training. S42. Load matrix obtained by training using normal operating condition data P Projecting standardized data and calculating T 2 Statistic: , of which Principal score vector, For the reason before k A diagonal matrix composed of eigenvalues; and SPE Statistic: ,in It is the residual vector; S43, if > and SPE > , and If a preset control limit is set, a fault is determined to have occurred, and the contribution values ​​of each variable are calculated. Locate the source of the fault; S5. Health-enhanced fault-tolerant control closed loop: Based on the system-level fault diagnosis results, the controller generates health-enhanced collaborative control instructions within a decision cycle of less than 20ms, and feeds the instructions back to the real-time simulator or external actuator through the communication bus to realize closed-loop verification of fault-tolerant control.

[0008] In S43 Statistics and SPE Control limits of statistics and It is obtained in advance by kernel density estimation at a set confidence level α.

[0009] In a second aspect, the present invention provides a semi-physical apparatus for implementing the above-described method of combined engine fault diagnosis and fault-tolerant control, comprising: A high-performance real-time simulator, equipped with a real-time operating system, is used to run the combined engine dynamics model and key component fault models, and generate virtual real-time engine data containing at least 32 channels of pressure data in a simulation cycle of less than 5ms. The 32-channel pressure scanning valve real-time acquisition simulator receives the 32-channel pressure data from the real-time simulator via an RS422 interface. It supports programmable sensor fault injection for any or multiple channels and outputs a pressure simulation signal containing fault characteristics. The simulator has a built-in multi-source data fusion fault diagnosis and data reconstruction algorithm, which is used to perform the distributed sensor diagnosis and data reconstruction function described in step S3 of claim 1. The EEC-3 fault diagnosis integrated controller is equipped with an ARM Cortex-M7 core and has multiple RS422 interfaces for real-time acquisition of the pressure simulation signal, the virtual real-time data, sensor diagnostic results and actuator feedback data. The controller incorporates a fault diagnosis algorithm and a health-enhanced fault-tolerant control algorithm based on principal component analysis, which are used to execute the system-level diagnosis and fault-tolerant control functions described in steps S4 and S5 of claim 1. The test management system communicates with the real-time simulator and the simulator via Ethernet for model deployment, online configuration of fault modes and parameter injection, and real-time data display and data storage. The communication bus system includes an RS422 serial bus, which connects the real-time emulator, the simulator, and the controller for data transmission.

[0010] The high-performance real-time simulator automatically generates real-time code from the simulation model using a model-based systems engineering approach.

[0011] The simulator supports fault injection modes including constant deviation, ramp drift, signal interference, and channel blockage.

[0012] The communication rate of the RS422 bus in the communication bus system is 460800bps.

[0013] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a four-level closed-loop verification platform comprising a high-performance real-time simulator, a dedicated sensor simulator, an intelligent diagnostic controller, and a comprehensive test management system. This effectively overcomes the limitations of pure digital simulation, such as the lack of real hardware-in-the-loop capability and difficulty in simulating noise and delay, as well as the high cost and inability to safely and repeatedly inject faults during ground-based hot-fire testing. The platform enables high-fidelity sensor fault injection that is programmable channel by channel. By integrating online diagnostic and data reconstruction algorithms based on spatial neighborhood and weighted fusion at the simulator level, and employing a rapid statistical diagnostic and contribution rate localization method based on principal component analysis (PCA) at the controller level, a multi-level collaborative diagnostic capability is formed from the sensor layer to system components. Simultaneously, the system ensures hard real-time solution of complex engine models in less than 5ms and rapid diagnostic decision-making by the controller in less than 20ms, thereby enabling accurate capture of fault transient characteristics and complete closed-loop verification of health-enhanced fault-tolerant control strategies. This provides a safe, reliable, repeatable and efficient semi-physical verification environment for the development, testing and optimization of combined engine fault diagnosis and fault-tolerant control algorithms, significantly reducing the R&D risks and costs based on real test runs, and strongly supporting the engineering application and iterative upgrade of engine predictive health management (PHM) technology. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments are briefly introduced below.

[0015] Figure 1 The figure shows the experimental results without any faults.

[0016] Figure 2 The results show the experimental findings of a decrease in oxidant pump efficiency.

[0017] Figure 3 This is a comparison chart showing the results after using the fault-tolerant control algorithm. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is based on data from 12 key sensors collected by a ground test bench for a combined engine, but the invention is not limited to this specific application scenario.

[0019] Example 1 A semi-physical method for combined engine fault diagnosis and fault-tolerant control includes the following steps: S1. Real-time model solving and virtual data generation: Based on the MATLAB / Simulink model-based systems engineering (MBSE) method, the combined engine dynamics model and key component fault model are automatically generated into real-time code and run on a high-performance real-time simulator to solve the engine virtual real-time data, including at least 32 pressure data, in a simulation cycle of less than 5ms.

[0020] S2. Fault Synchronous Injection and Signal Simulation: The test management system issues fault configuration commands via Ethernet. The 32-channel pressure scanning valve real-time acquisition simulator synchronously receives the 32 channels of pressure data and injects programmable sensor faults into any one or more of the 32 channels of pressure data according to the fault configuration commands to generate a pressure simulation signal containing fault characteristics.

[0021] S3. Distributed Sensor Diagnosis and Data Reconstruction: The simulator performs low-level multi-source data fusion fault diagnosis and data reconstruction processing on the pressure simulation signal to assess the sensor's health status and data confidence, and generate sensor diagnostic results. Specifically, this includes the following steps: Spatial neighborhood matrix construction: Based on the physical layout of the combined engine thrust chamber and piping, a spatial neighborhood matrix of 32 pressure measurement points was established. If channel i and channel j are adjacent, then the correlation coefficient... ,otherwise .

[0022] Outlier detection (based on a sliding window): Sampled values ​​of channel i at time k. Based on sliding window judgment if If so, the data in that channel is determined to be abnormal. For the most recent The mean of each sampling point This represents the historical standard deviation of the channel.

[0023] Data reconstruction based on weighted fusion: when determining When a channel fails, it is reconstructed using weighted fusion data from normal channels within its spatial neighborhood. The reconstructed value is... The calculation formula is: ; in, for The neighborhood set of the channel, For neighborhood channels The measured value, for The confidence level of the channel is 1 for normal operation and 0 for fault. This reconstructed value will replace the original measurement value output to ensure that the upper-level control does not diverge due to a single-point sensor failure.

[0024] S4. Multi-source data acquisition and system-level diagnosis: The EEC-3 fault diagnosis integrated controller acquires the simulated pressure signal (or reconstructed data), the virtual real-time data, the sensor diagnostic results, and actuator feedback data in real time. The controller runs a system-level fault diagnosis algorithm based on principal component analysis (PCA) to achieve real-time identification, isolation, and location of engine component faults. The specific steps are as follows: Data preprocessing and standardization: Acquiring multi-dimensional observation vectors including multiple pressure, flow, speed, and temperature data. The mean vector of normal operating condition data obtained during the offline training phase is used. μ and standard deviation vector σ Perform Z-score normalization on the real-time data: ; Principal Component Model Projection and Statistics Calculation: Using the load matrix P obtained from training with normal operating condition data, the standardized data is projected, and T is calculated. 2 Statistic: , of which Principal score vector, Here is a diagonal matrix consisting of the first k eigenvalues; and the SPE statistic: ,in It is the residual vector; Fault identification and isolation: setting control limits at confidence level α and ,like > And SPE> If a fault occurs, it is determined that a fault has occurred, and the contribution values ​​of each variable are calculated. To locate the source of the fault, select the variable with the largest contribution value as the source of the fault.

[0025] S5. Health-enhanced fault-tolerant control closed loop: Based on the system-level fault diagnosis results, the controller generates health-enhanced collaborative control instructions (such as correcting valve opening and adjusting speed) within a decision cycle of less than 20ms, and transmits the instructions to the real-time simulator or external controller via RS422 bus to realize online adjustment of the engine model and complete the closed-loop verification of the fault-tolerant control strategy.

[0026] A semi-physical device for implementing the above-described combined engine fault diagnosis and fault-tolerant control method is characterized by comprising the following components: High-performance real-time simulator: This simulator is equipped with a high-performance central processing unit and a real-time operating system (such as Real-Time Linux) for real-time execution of the combined engine's overall dynamics model and key component fault models. The model covers components such as the gas generator, turbopump, thrust chamber, valves, and pipelines, and integrates typical fault models such as pipeline leakage and pump efficiency degradation. The simulator employs a model-based systems engineering approach, automatically generating real-time running code from the simulation model to ensure that the model's solution cycle meets hard real-time requirements (typical cycle < 5ms). The simulator outputs virtual real-time engine data, including 32 channels of pressure, flow, temperature, and speed, via an RS422 interface.

[0027] 32-Channel Pressure Scanning Valve Real-Time Acquisition Simulator: This simulator is a dedicated hardware module that receives 32 channels of virtual pressure data from a real-time simulator via an RS422 interface. Its core function is to inject programmable sensor faults into any one or more channels. Fault modes are flexibly configurable, including constant deviation, ramp drift, signal transmission interference, and channel blockage. The simulator converts the 32 channels of pressure data affected by the fault into a physical simulation signal output containing fault characteristics. The simulator incorporates a multi-source data fusion fault diagnosis and data reconstruction algorithm to execute the sensor-level diagnosis and data reconstruction function described in step S3 of claim 1, and outputs the sensor diagnosis results.

[0028] EEC-3 Fault Diagnosis Integrated Controller: The controller is built on an ARM Cortex-M7 core with a main frequency > 200MHz and has at least 5 RS422 interfaces. It collects in real time the pressure simulation signal output by the simulator, the virtual real-time data output by the real-time simulator, the sensor diagnostic results, and the feedback data from external actuators. The controller incorporates a multi-level fault diagnosis algorithm and a health-enhanced fault-tolerant control algorithm to execute the system-level fault diagnosis function described in step S4 of claim 1 and to generate health-enhanced collaborative control instructions in step S5.

[0029] Test Management System: This system is implemented based on a high-performance host computer and interacts with the entire system via Ethernet to achieve global control. The system provides a graphical interface that supports engine model deployment, test procedure editing, online fault mode configuration, and parameter injection (e.g., setting pipeline blockage thresholds and sensor drift rates). The system features real-time data display (including 32-channel pressure curves), 3D engine visualization, and test data storage.

[0030] Communication bus system: includes an RS422 serial bus, connecting the real-time simulator, the simulator and the integrated controller, for high-speed real-time transmission of the virtual real-time data, the pressure simulation signal and the health enhancement collaborative control command, the RS422 communication rate is 460800bps.

[0031] like Figure 1 The data shown is generated during normal engine testing and is used for comparison with fault data.

[0032] Figure 2 The experimental results show that the efficiency of the oxidant pump decreased significantly after the failure occurred 13 seconds later. The parameters such as pressure and flow rate decreased significantly under the failure condition. Figure 3 The image shows a comparison of the results after using the fault-tolerant control algorithm in this embodiment. After about 2 seconds, the fault-tolerant control pulls the fault back to the normal value, proving the effectiveness of the algorithm in fault-tolerant control.

[0033] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not describe all details exhaustively, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification.

Claims

1. A semi-physical method for combining engine fault diagnosis and fault-tolerant control, characterized in that, Includes the following steps: S1. Real-time model solving and virtual data generation: By running the combined engine dynamics model and fault model through a real-time simulator, virtual real-time data containing at least 32 pressure data points is generated with a cycle of less than 5ms. S2. Synchronous Fault Injection and Signal Simulation: The test management system issues a fault configuration command, and the 32-channel pressure scanning valve real-time acquisition simulator receives the pressure data and injects programmable sensor faults into the specified channels according to the command to generate a pressure simulation signal containing fault characteristics. S3. Distributed sensor diagnosis and data reconstruction: The simulator performs multi-source data fusion fault diagnosis and data reconstruction processing on the pressure simulation signal to generate sensor diagnosis results; The processing includes: S31. Establish the spatial neighborhood matrix of pressure measurement points based on the engine's physical layout. If the channel i With channel j Adjacent, then the correlation coefficient ,otherwise ; S32, for the channel i exist k Sample value at time Based on sliding window judgment if If so, the data in that channel is determined to be abnormal. For the most recent The mean of each sampling point This represents the historical standard deviation of the channel; S33, when determining When a channel fails, it is reconstructed using weighted fusion data from normal channels within its spatial neighborhood. The reconstructed value is... The calculation formula is: ； in, for The neighborhood set of the channel, For neighborhood channels The measured value, for The confidence level of the channel is 1 for normal operation and 0 for fault. S4. Multi-source data acquisition and system-level diagnosis: The EEC-3 fault diagnosis integrated controller acquires the pressure simulation signal, the virtual real-time data, the sensor diagnostic results and the actuator feedback data, and runs a fault diagnosis algorithm based on principal component analysis; The algorithm includes: S41. For the acquired multidimensional observation vectors Standardization process: ； in, μ and σ These are the mean vector and standard deviation vector of the normal operating condition data obtained through offline training. S42. Load matrix obtained by training using normal operating condition data P Projecting standardized data and calculating T2 Statistic: , of which Principal score vector, For the reason before k A diagonal matrix composed of eigenvalues; and SPE Statistic: ,in It is the residual vector; S43, if > and SPE > , and If a preset control limit is set, a fault is determined to have occurred, and the contribution values ​​of each variable are calculated. Locate the source of the fault; S5. Health-enhanced fault-tolerant control closed loop: Based on the system-level fault diagnosis results, the controller generates health-enhanced collaborative control instructions within a decision cycle of less than 20ms, and feeds the instructions back to the real-time simulator or external actuator through the communication bus to realize closed-loop verification of fault-tolerant control.

2. The semi-physical method for combined engine fault diagnosis and fault-tolerant control according to claim 1, characterized in that, In S43 Statistics and SPE Control limits of statistics and It is obtained in advance by kernel density estimation at a set confidence level α.

3. A semi-physical device for implementing the combined engine fault diagnosis and fault-tolerant control method as described in claim 1 or 2, characterized in that, include: A high-performance real-time simulator, equipped with a real-time operating system, is used to run the combined engine dynamics model and key component fault models, and generate virtual real-time engine data containing at least 32 channels of pressure data in a simulation cycle of less than 5ms. The 32-channel pressure scanning valve real-time acquisition simulator receives the 32-channel pressure data from the real-time simulator via an RS422 interface. It supports programmable sensor fault injection for any or multiple channels and outputs a pressure simulation signal containing fault characteristics. The simulator has a built-in multi-source data fusion fault diagnosis and data reconstruction algorithm, which is used to perform the distributed sensor diagnosis and data reconstruction function described in step S3 of claim 1. The EEC-3 fault diagnosis integrated controller is equipped with an ARM Cortex-M7 core and has multiple RS422 interfaces for real-time acquisition of the pressure simulation signal, the virtual real-time data, sensor diagnostic results and actuator feedback data. The controller incorporates a fault diagnosis algorithm and a health-enhanced fault-tolerant control algorithm based on principal component analysis, which are used to execute the system-level diagnosis and fault-tolerant control functions described in steps S4 and S5 of claim 1. The test management system communicates with the real-time simulator and the simulator via Ethernet for model deployment, online configuration of fault modes and parameter injection, and real-time data display and data storage. The communication bus system includes an RS422 serial bus, which connects the real-time emulator, the simulator, and the controller for data transmission.

4. The semi-physical device for combined engine fault diagnosis and fault-tolerant control according to claim 3, characterized in that, The high-performance real-time simulator automatically generates real-time code from the simulation model using a model-based systems engineering approach.

5. The semi-physical device for combined engine fault diagnosis and fault-tolerant control according to claim 3, characterized in that, The simulator supports fault injection modes including constant deviation, ramp drift, signal interference, and channel blockage.

6. The semi-physical device for combined engine fault diagnosis and fault-tolerant control according to claim 3, characterized in that, The communication rate of the RS422 bus in the communication bus system is 460800bps.

Images