ARMA model-based liquid oxygen methane rocket engine fault prediction method and system

Abstract

The invention discloses a liquid oxygen methane rocket engine fault prediction method and system based on an ARMA model, and the method comprises the steps: constructing a modularized simulation model of a liquid oxygen methane rocket engine, injecting fault factors to simulate typical faults (such as methane pump cavitation, turbine blade ablation and the like), and obtaining normal and fault data containing Gaussian noise; carrying out time sequence modeling on the six key parameters by adopting an ARMA (Autoregressive Moving Average Model), and optimizing model parameters through ACF / PACF order determination and AIC criteria; establishing a self-adaptive residual threshold formula (threshold = residual mean + / -2 times of standard deviation), and realizing early warning of the fault by combining a judgment criterion that the threshold is continuously exceeded for three times and three parameters cooperatively give an alarm; experiments show that the alarm time is 0.1-0.13 seconds earlier than that of a traditional red line threshold method, the false alarm rate is reduced from 18.7% to 0.3%, and the spaceflight-level real-time performance and reliability requirements are met; the method does not need a large number of fault samples, is high in anti-interference capability, and is suitable for online health monitoring of reusable rocket engines.

Description

Technical Field

[0001] The present invention relates to the technical field of rocket engine health monitoring, and in particular to a liquid oxygen-methane rocket engine fault prediction method and system based on an ARMA model. Background Art

[0002] Liquid oxygen-methane rocket engines, thanks to their high specific impulse (vacuum specific impulse ≥ 360s), low cost (fuel cost is only one-fifth that of hydrogen-oxygen engines), and environmental friendliness, have become the core power unit for reusable launch vehicles. However, their closed-cycle system, consisting of a high-speed turbopump (speed > 30,000 rpm) and a high-temperature combustion chamber (temperature > 3,000K), is prone to failures such as methane pump cavitation, turbine blade erosion, and gas line leaks during operation.

[0003] Existing fault detection methods primarily include data-driven approaches, physical modeling, and threshold alarms. Data-driven approaches rely on a large number of fault samples, but rocket testing is expensive. For one type of liquid oxygen-kerosene engine, cumulative test data is only 187 hours, with fault data accounting for less than 0.3%, resulting in insufficient model generalization. While physical modeling can partially compensate for missing data, it struggles to accurately describe the degradation process of internal engine components. Furthermore, parameter identification cycles are long (typically exceeding six weeks), making it difficult to meet real-time requirements. Furthermore, traditional threshold alarm methods have a false alarm rate as high as 18.7%, making it impossible to distinguish between normal fluctuations and fault signals. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the prior art, it is desired to provide a liquid oxygen-methane rocket engine fault prediction method and system based on the ARMA model.

[0005] The present invention provides a liquid oxygen-methane rocket engine fault prediction method and system based on the ARMA model, comprising the following steps: 1) Build a modular simulation model of a liquid oxygen-methane rocket engine, including core components such as the methane pump, oxygen pump, turbine, gas generator, and thrust chamber; 2) generating normal operating data under normal working conditions through the simulation model, and performing error verification on the simulation results; 3) injecting step signals or fault factors into the simulation model to simulate five typical faults: methane pump cavitation, methane turbine blade ablation, methane gas line leakage, gas generator combustion efficiency reduction, and thrust chamber ablation; 4) generating fault data under a fault condition by injecting the step signal or the fault factor into the simulation model, and verifying whether the fault is actually occurring; 5) The simulation model is superimposed with Gaussian noise, and data of six key parameters including the flow rate of the methane pump, the flow rate of the oxygen pump, the rotational speed of the methane turbine, the rotational speed of the oxygen turbine, the chamber pressure of the gas generator, and the chamber pressure of the thrust chamber are collected; 6) ARMA models are respectively constructed for the six key parameters, namely the ARMA model of the methane pump flow rate, the ARMA model of the oxygen pump flow rate, the ARMA model of the methane turbine rotational speed, the ARMA model of the oxygen turbine rotational speed, the ARMA model of the gas generator chamber pressure, and the ARMA model of the thrust chamber chamber pressure; and the ARMA model parameters are optimized through ACF / PACF order determination and AIC criterion; 7) The normal data are respectively substituted into the six ARMA models to obtain predicted values, the differences between the predicted values and the normal operation data are calculated to obtain residual values, the mean μ and variance σ² of the residual value sequence are calculated, and the residual threshold is determined; 8) The fault data are respectively substituted into the six ARMA models for real-time prediction and application, and fault prediction is carried out through the determination criterion of exceeding the residual threshold continuously three times and the collaborative alarm of three parameters.

[0006] Further, the simulation model is constructed through the MATLAB / Simulink platform, which is a component-level model containing thermodynamic equations and fluid dynamics equations, and includes the dynamic characteristics of all working conditions of startup, steady state, and shutdown; the verification error of the simulation results of the normal data is less than or equal to 2%.

[0007] Further, the six ARMA models in step 6) are all initially ordered through ACF / PACF to judge the truncation and tailing of the time series, and estimate the numerical range of the autoregressive model order p and the moving average model order q. Among them, the autoregressive model order p ≤ 4, and the moving average model order q ≤ 6; then the AIC criterion is used to determine its exact value to complete the model order determination, and the maximum likelihood estimation method is used to estimate the ARMA model parameters to complete the establishment of the ARMA model.

[0008] Further, the residual threshold in step 7) is obtained by statistically analyzing the residual distribution of normal data to ensure that more than 95% of the data fall within the threshold interval, the bandwidth coefficient a = 2, and the threshold calculation formula is: threshold = μ ± 2σ.

[0009] Further, the determination criterion in step 8) includes single key parameter alarm and system alarm; among them, For the single key parameter alarm, it is necessary to satisfy that the residuals of three consecutive sampling points exceed the threshold, and the time interval between every two adjacent sampling points is set to 0.002 s; For the system alarm, at least three key parameter alarms are required, and the alarm time of the third key parameter is taken as the final result.

[0010] Further, the range of the fault factor Ka in step 3) is set to 0.7 ≤ Ka ≤ 1.3, and its injection methods include: cavitation of the methane pump: Ka = 0.7 - 0.9, reducing the outlet pressure and flow rate of the methane pump; ablation of the methane turbine blades: Ka = 0.7 - 0.9, reducing the flow rate of the methane turbine; leakage of the methane gas path: Ka = 0.7 - 0.9, reducing the turbine drive flow rate; decline in the combustion efficiency of the gas generator: Ka = 0.7 - 0.9, insufficient combustion of methane and oxygen, reducing the generated gas and heat; ablation of the thrust chamber: Ka = 1.1 - 1.3, gradual degradation of the inner wall material of the thrust chamber, increasing heat loss.

[0011] The liquid oxygen-methane rocket engine fault prediction system based on the ARMA model includes a data acquisition module, an edge computing module, a threshold management module, and an alarm decision module; among them; The data acquisition module is a 24-bit ADC analog-to-digital conversion with a sampling rate of 100 Hz, covering 6 key parameters; The edge computing module is based on an ARMCortex-A72 processor, integrated with an ARMA prediction algorithm, and the calculation delay < 5 ms; The threshold management module stores the mean and variance of the residuals of each parameter and dynamically updates the threshold; The alarm decision module applies a continuous determination criterion and outputs the fault type and alarm time.

[0012] Further, the data acquisition module includes a pressure sensor with an accuracy of ±0.1% FS, a speed sensor with a resolution of 0.01 rpm, a flow sensor with a linearity of 0.2%, and an anti-aliasing filter with a cut-off frequency of 40 Hz.

[0013] Further, the fault prediction system further includes a human-machine interaction interface for real-time display of parameter curves, residual plots, and alarm information, and a historical database for storing the most recent 1000 sets of prediction data and supporting fault traceability analysis.

[0014] Compared with the prior art, the beneficial effects of the present invention are: The liquid oxygen-methane rocket engine fault prediction method and system based on the ARMA model of the present invention focuses on constructing a modular simulation model of the rocket engine and using this model to generate data sets for normal and fault conditions; modeling the key operating parameters of the rocket engine (such as methane pump flow rate, oxygen pump flow rate, turbine speed, gas generator pressure, thrust chamber pressure, etc.) through the ARMA model, and performing fault prediction based on the adaptive residual threshold; the characteristics of this method include: I. Data generation based on a modular simulation model 1) Build a component-level simulation model of a rocket engine containing thermodynamic equations and fluid dynamics equations using the MATLAB / Simulink platform; 2) Simulate multiple typical faults by injecting different fault factors (such as cavitation of the methane pump, ablation of the turbine blade, decrease in the combustion efficiency of the gas generator, ablation of the thrust chamber, etc.); 3) Collect key parameter data when a fault occurs and superimpose Gaussian noise to simulate actual sensor measurement data; II. ARMA Modeling and Parameter Optimization 1) Use the ARMA model to perform time series prediction on key parameters and use the autocorrelation function (ACF) and partial autocorrelation function (PACF) to estimate the model order; 2) Optimize the model parameters using the Akaike information criterion (AIC) to determine their exact values to complete the model order determination; 3) Estimate the model parameters using the maximum likelihood estimation method; III. Adaptive Residual Threshold Judgment 1) In actual tests, it is found that when the bandwidth coefficient a = 2, the false alarm rate is reduced to 0.3%, and the fault can be detected 0.1 - 0.13 s in advance; if a takes a larger value, the sensitivity of fault detection decreases; if a takes a smaller value, the false alarm rate increases. Therefore, set the residual threshold formula: Threshold = Residual mean ± 2 times the standard deviation; 2) Combine the judgment criterion of exceeding the threshold continuously three times and the collaborative alarm of three parameters to achieve high-reliability fault detection; IV. Efficient Real-time Fault Prediction 1) Use the ARM Cortex-A72 processor as an edge computing device to achieve efficient real-time computing, with a data acquisition rate of 100 Hz and a computing delay of less than 5 ms; 2) The false alarm rate is reduced to 0.3%, and the alarm is advanced by 0.1 - 0.13 seconds compared with the traditional red line threshold method, meeting the real-time health monitoring requirements of spacecraft; In summary, the technical solution of this application combines time series prediction and statistical threshold to achieve early fault warning; this method does not require a large number of fault samples, the prediction step is only 0.002 seconds, and the false alarm rate is reduced from 18.7% to 0.3%. Compared with traditional methods, the ARMA model is not only applicable to stationary time series prediction but also can effectively reduce the dependence on large-scale fault data, improve the real-time performance and accuracy of detection, and provide a new solution for the health monitoring of liquid oxygen methane rocket engines.

[0015] It should be understood that the content described in the Summary of the Invention section is not intended to limit the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. Brief Description of the Drawings

[0016] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non - limiting embodiments read in conjunction with the accompanying drawings: Figure 1 is the overall flowchart of the fault prediction method for the liquid oxygen - methane rocket engine based on the ARMA model; Figure 2 is the simulation model diagram of the liquid oxygen - methane rocket engine; Figure 3 is the fault prediction flowchart of the ARMA model; Figure 4 is the schematic diagram of the ACF and PACF order - determination method; Figure 5 is the ARMA fault detection flowchart; Figure 6 are the residual threshold diagrams of six key parameters respectively when the methane pump cavitation fault occurs; Figure 7 are the residual threshold diagrams of six key parameters respectively when the methane gas path leakage fault occurs; Figure 8 are the residual threshold diagrams of six key parameters respectively when the methane turbine blade ablation fault occurs; Figure 9 are the residual threshold diagrams of six key parameters respectively when the combustion efficiency of the gas generator decreases; Figure 10 are the residual threshold diagrams of six key parameters respectively when the thrust chamber ablation fault occurs; Figure 11 is the comparison table of the red - line threshold and the ARMA alarm time when the combustion efficiency of the gas generator decreases. Specific Embodiments

[0017] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention and are not intended to limit the invention. Additionally, it should be noted that for the sake of description, only parts related to the invention are shown in the drawings.

[0018] It should be noted that, without conflict, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the drawings and embodiments.

[0019] Please refer to Figures 1 to 11 , the embodiments of the present invention provide a fault prediction method for a liquid oxygen - methane rocket engine based on the ARMA model, including the following steps: 1) Construct a modular simulation model of the liquid oxygen - methane rocket engine, including core components such as a methane pump, an oxygen pump, a turbine, a gas generator, and a thrust chamber; 2) Generate normal operation data under normal conditions through the simulation model, and conduct a stationarity test (ADF test) on the normal operation data to ensure that the data mean and variance do not change significantly over time to meet the stationarity requirements; and verify the error of the simulation results. Among them, the simulation model is constructed through the MATLAB / Simulink platform. It is a component-level model containing thermodynamic equations and fluid dynamics equations, and includes the dynamic characteristics of all working conditions such as startup, steady state, and shutdown; the verification error of the simulation results of the normal data is less than or equal to 2%.

[0020] 3) Inject a step signal or a fault factor into the simulation model to simulate 5 types of typical faults, namely cavitation of the methane pump, ablation of the methane turbine blade, leakage of the methane gas path, decrease in the combustion efficiency of the gas generator, and ablation of the thrust chamber. Among them, the range of the fault factor Ka is set to 0.7 ≤ Ka ≤ 1.3, and its injection methods include: cavitation of the methane pump: Ka = 0.7 - 0.9, reducing the outlet pressure and outlet flow of the methane pump; ablation of the methane turbine blade: Ka = 0.7 - 0.9, reducing the flow rate of the methane turbine; leakage of the methane gas path: Ka = 0.7 - 0.9, reducing the turbine drive flow rate; decrease in the combustion efficiency of the gas generator: Ka = 0.7 - 0.9, insufficient combustion of methane and oxygen, reducing the generated gas and heat; ablation of the thrust chamber: Ka = 1.1 - 1.3, gradual degradation of the inner wall material of the thrust chamber, increasing heat loss.

[0021] 4) Generate fault data under fault conditions through the simulation model after injecting the step signal or the fault factor, and verify whether it conforms to the actual situation when the fault occurs. 5) The simulation model is superimposed with Gaussian noise, and data of 6 key parameters, namely methane pump flow rate, oxygen pump flow rate, methane turbine speed, oxygen turbine speed, gas generator chamber pressure, and thrust chamber chamber pressure, are collected. 6) Construct ARMA models for the 6 key parameters respectively, namely the methane pump flow rate ARMA model, the oxygen pump flow rate ARMA model, the methane turbine speed ARMA model, the oxygen turbine speed ARMA model, the gas generator chamber pressure ARMA model, and the thrust chamber chamber pressure ARMA model; and optimize the ARMA(p,q) model parameters through ACF / PACF order determination and AIC criterion. In some embodiments of the present application, the autoregressive moving average model ARMA(p,q) can be expressed as: is the autoregressive coefficient, is the moving average coefficient, p is the order of the autoregressive model, and q is the order of the moving average model. The method for determining the model order is as follows: First, use ACF and PACF to judge the truncation and trailing properties of the time series, and roughly estimate the numerical ranges of p and q, where p ≤ 4 and q ≤ 6; then determine their exact values through the AIC criterion to complete the model order determination (such as ARMA(4,3), ARMA(1,1)), and use the maximum likelihood estimation method to estimate the model parameters, thus completing the establishment of the ARMA model.

[0022] 7) Substitute the normal data into the 6 ARMA models respectively to obtain the predicted values, subtract the predicted values from the normal operation data to obtain the residual values, calculate the mean μ and variance σ² of the residual value sequence, and determine the residual threshold; Among them, the residual threshold is obtained by statistically analyzing the residual distribution of the normal data to ensure that more than 95% of the data falls within the threshold interval, the bandwidth coefficient a = 2, and the threshold calculation formula is: threshold = μ ± 2σ.

[0023] 8) Substitute the fault data into the 6 ARMA models respectively, perform real-time prediction and application, and conduct fault prediction through the determination criterion of exceeding the residual threshold continuously for 3 times and the collaborative alarm of three parameters.

[0024] Among them, the determination criterion includes single key parameter alarm and system alarm; among them, For the single key parameter alarm, it is necessary to satisfy that the residuals of 3 consecutive sampling points exceed the threshold, and the time interval between every two adjacent sampling points is set to 0.002s; For the system alarm, at least 3 key parameter alarms are required, and the alarm time of the third key parameter is taken as the final result.

[0025] The liquid oxygen methane rocket engine fault prediction system based on the ARMA model includes a data acquisition module, an edge computing module, a threshold management module, and an alarm decision module; among them; The data acquisition module is a 24-bit ADC analog-to-digital conversion with a sampling rate of 100Hz, covering 6 key parameters; The edge computing module is based on an ARMCortex-A72 processor, integrates the ARMA prediction algorithm, and the calculation delay < 5ms; The threshold management module stores the mean and variance of the residuals of each parameter and dynamically updates the threshold; The alarm decision module applies the continuous determination criterion and outputs the fault type and alarm time.

[0026] In a preferred embodiment, the data acquisition module includes a pressure sensor with an accuracy of ±0.1% FS, a rotational speed sensor with a resolution of 0.01rpm, a flow sensor with a linearity of 0.2%, and an anti-aliasing filter with a cut-off frequency of 40Hz.

[0027] Among them, in order to reduce noise interference, anti-aliasing filtering is adopted in the data preprocessing process, the filtering cut-off frequency is set to 40 Hz, and the standardization method is combined to eliminate data deviation to ensure data quality.

[0028] In a preferred embodiment, the fault prediction system further includes a human-machine interaction interface for real-time displaying parameter curves, residual plots and alarm information, and a historical database for storing the most recent 1000 sets of prediction data and supporting fault traceability analysis.

[0029] Embodiment 1

[0030] The present invention uses the MATLAB / Simulink platform to construct a modular simulation model of a liquid oxygen-methane rocket engine, covering the main components of the engine, including a methane pump, an oxygen pump, a methane turbine, an oxygen turbine, a gas generator, and a thrust chamber, etc.; the simulation model is calculated based on thermodynamic equations and fluid dynamics equations, and can simulate the changes of internal parameters of the engine under different working conditions; experimental data shows that under fault-free conditions, the error between the model calculation results and the actual engine test data does not exceed 2%, which can be referred to Figure 2 。

[0031] The data acquisition system uses a 24-bit ADC analog-to-digital converter, and the sampling rate is set to 100 Hz, covering 6 key parameters including methane pump flow rate, oxygen pump flow rate, methane turbine speed, oxygen turbine speed, gas generator chamber pressure, and thrust chamber chamber pressure; in order to reduce noise interference, anti-aliasing filtering is adopted in the data preprocessing process, the filtering cut-off frequency is set to 40 Hz, and the standardization method is combined to eliminate data deviation to ensure data quality.

[0032] The present invention uses the ARMA model to predict engine faults (such as Figure 3 ), first, the ADF (Augmented Dickey-Fuller) test is performed on the normal working condition data to ensure that the data mean and variance do not change significantly over time to meet the stationarity requirements.

[0033] The auto-correlation function (ACF) and partial auto-correlation function (PACF) are used for order determination (such as Figure 4 ), and the Akaike information criterion (AIC) is combined to optimize the order of the ARMA model. Finally, the optimal ARMA models for methane pump flow rate, oxygen pump flow rate, methane turbine speed, oxygen turbine speed, gas generator chamber pressure, and thrust chamber chamber pressure are respectively:

[0034] The maximum likelihood estimation (MLE) method is used to solve the ARMA model parameters, including autoregressive coefficients and moving average coefficients, and the root mean square error (RMSE) is used to evaluate the model fitting accuracy to ensure that the prediction error is less than 3%. For the normal operation data, the residuals between the predicted values and the actual values are calculated, and statistical analysis is performed on the residuals. The mean and standard deviation of the residuals are calculated, and an adaptive residual threshold is set. The calculation formula for each model threshold is: Threshold = Residual mean ± 2 times standard deviation.

[0035] To improve the accuracy of fault detection, the present invention uses residual analysis as the detection benchmark. Specifically, time series prediction is first performed on the normal operation data, and then the difference between the data in the normal operation state and the predicted value is calculated, and this difference is used as the residual threshold. If the difference between the predicted value and the normal value exceeds the residual threshold three times in a row, and more than half of the key parameters issue alarms, the engine failure is officially determined. The alarm time of the third key parameter is taken as the engine alarm time, which can be referred to Figure 5 . <~

[0036] Experiments verify the effectiveness of the proposed method under different fault types, including five typical fault modes: cavitation of methane pump, leakage of methane gas path, ablation of methane turbine blade, decrease of combustion efficiency of methane gas generator, and ablation of thrust chamber. Cavitation fault of methane pump (see Figure 6 ): The fault factor is set to 0.9, and the fault occurrence time is 5 seconds. When the cavitation fault occurs in the methane pump, it will directly lead to the decrease of the outlet pressure and outlet flow of the methane pump, and then the methane flow rate of the gas generator and thrust chamber will decrease, and the chamber pressure will decrease. Subsequently, the efficiency of the methane turbine and oxygen turbine will be affected, resulting in the decrease of the turbine pump speed, the reduction of the methane and oxygen flow rates in the system, the reduction of the thrust generated by combustion, and the reduction of the overall performance of the engine vehicle. It can be seen from the residual plot that each parameter can detect the engine fault. The rotational speed of the methane turbine pump is not easy to directly observe from the graph, but its fault can be identified through the program and an alarm can be issued. The other 5 parameters all significantly exceed the threshold range.

[0037] Leakage fault of methane gas path (see Figure 7 ): The fault factor is set to 0.9, and the fault occurrence time is 5 seconds. When there is a leakage in the methane gas path, this phenomenon will directly lead to a significant reduction in the gas volume entering the turbine pump, and then cause a decrease in the rotational speed of the turbine pump, a weakening of the pumping capacity, and a corresponding decrease in the gas flow rate at the outlet. This series of chain reactions result in insufficient gas supply to the gas generator and thrust chamber, reducing the working efficiency of these two key components, specifically manifested as a decrease in the internal pressure of the thrust chamber. Finally, this pressure reduction will directly affect the overall performance of the engine, resulting in a decrease in the overall performance of the engine. It can be seen from the residual plot that each parameter can quickly detect the engine fault through the residual adaptive threshold.

[0038] Ablation failure of the methane turbine blade (see Figure 8 ): The fault factor is set to 0.9, and the fault occurrence time is 5 seconds; the ablation of the methane turbine blade will directly lead to a decrease in the flow rate of the methane turbine, which in turn causes a decrease in the rotational speed of the methane pump, a reduction in the methane flow rate, a decrease in the total flow rate entering the gas generator and the thrust chamber, a decrease in the gas generated by combustion, a decrease in the chamber pressure, a decrease in the gas flow rate flowing from the gas generator to the methane turbine and the oxygen turbine, a decrease in the efficiency of the turbopump, and a decrease in the rotational speed, ultimately resulting in a decline in the overall performance of the engine; it can be seen from the residual plot that each parameter can accurately detect the state of the engine through the residual adaptive threshold.

[0039] Decrease in the combustion efficiency of the gas generator (see Figure 9 ): The fault factor is set to 0.9, and the fault occurrence time is 5 seconds; the decrease in the combustion efficiency of the gas generator will lead to incomplete combustion of methane and oxygen, generating less gas and heat, reducing its chamber pressure, causing a decrease in the flow rate and rotational speed of the turbopump, a decrease in the flow rate flowing into the thrust chamber, a decrease in the chamber pressure of the thrust chamber, and a decline in the engine performance; it can be seen from the residual plot that each parameter can accurately detect the state of the engine through the residual adaptive threshold.

[0040] Ablation failure of the thrust chamber (see Figure 10 ): The fault factor is set to 1.1, and the fault occurrence time is 5 seconds; once the ablation failure of the thrust chamber occurs, it will directly cause the gradual degradation of the inner wall material of the thrust chamber, forming pits, cracks and irregular surfaces, and these changes significantly reduce the thermal efficiency and structural integrity of the thrust chamber; as the ablation intensifies, the amount of high-temperature gas that the thrust chamber can accommodate and effectively utilize gradually decreases, which in turn affects the working conditions of the turbopump. After receiving the reduced gas flow rate, the rotational speed of the turbopump will inevitably decrease, and the pumping efficiency will also decrease, further restricting the stable supply of methane and oxygen to the gas generator and the thrust chamber; this insufficient gas supply not only weakens the combustion intensity of the gas generator, but also causes a significant decrease in the internal pressure of the thrust chamber, unable to maintain the thrust output required by the design; ultimately, the ablation failure of the thrust chamber severely weakens the overall performance of the engine through this series of complex chain reactions; it can be seen from the residual plot that each parameter can accurately detect the state of the engine through the residual adaptive threshold.

[0041] The fault prediction system consists of a data acquisition module, an edge computing module, a threshold management module and an alarm decision module; among them, The data acquisition module uses high-precision sensors to cover key parameters such as pressure, rotational speed and flow rate, and is equipped with anti-aliasing filters to suppress high-frequency noise; The edge computing module adopts an ARM Cortex-A72 processor, integrates the ARMA prediction algorithm, realizes efficient real-time computing, and the computing latency is less than 5 ms; The threshold management module stores the mean and variance of the residuals of each parameter, supports dynamic update to adapt to different working conditions; The alarm decision module combines residual analysis and continuous determination criteria to output the fault type and alarm time.

[0042] The present invention combines the ARMA model with an adaptive residual threshold, and the experimental verification of the fault prediction effect of the present invention has been carried out. Typical faults such as cavitation of methane pumps, ablation of turbine blades, decrease in combustion efficiency of gas generators, and ablation of thrust chambers have been tested, realizing an efficient, real-time, and low false alarm liquid oxygen methane rocket engine fault prediction system; Experiments show that (such as Figure 11 ), for example, when a fault of decreasing combustion efficiency of the gas generator occurs, the alarm time is 0.1 - 0.13 seconds earlier than the red line threshold method, and the false alarm rate is reduced to about 0.3%, providing a new technical means for the health monitoring of spacecraft.

[0043] In the description of this specification, terms such as "connection", "installation", "fixation", etc. should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific situations.

[0044] In the description of this specification, the description of terms such as "one embodiment", "some embodiments", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0045] The above is only the preferred embodiment of this application and is not used to limit this application. For those skilled in the art, this application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

1. A fault prediction method for a liquid oxygen-methane rocket engine based on the ARMA model, characterized in that, It includes the following steps: 1) Construct a modular simulation model of a liquid oxygen-methane rocket engine, including key components such as a methane pump, an oxygen pump, a turbine, a gas generator, and a thrust chamber; 2) Generate normal operation data under normal conditions through the simulation model, and verify the error of the simulation results; 3) Inject a step signal or a fault factor into the simulation model to simulate 5 types of typical faults, namely methane pump cavitation, methane turbine blade ablation, methane gas path leakage, gas generator combustion efficiency decline, and thrust chamber ablation; 4) Generate fault data under fault conditions through the simulation model after injecting the step signal or the fault factor, and verify whether it conforms to the actual situation when the fault occurs; 5) Add Gaussian noise to the simulation model, and collect data of 6 key parameters, namely methane pump flow rate, oxygen pump flow rate, methane turbine speed, oxygen turbine speed, gas generator chamber pressure, and thrust chamber chamber pressure; 6) Construct ARMA models for the 6 key parameters respectively, namely the methane pump flow rate ARMA model, the oxygen pump flow rate ARMA model, the methane turbine speed ARMA model, the oxygen turbine speed ARMA model, the gas generator chamber pressure ARMA model, and the thrust chamber chamber pressure ARMA model; and optimize the ARMA model parameters through ACF / PACF order determination and AIC criterion; 7) Substitute the normal data into the 6 ARMA models respectively to obtain predicted values, subtract the predicted values from the normal operation data to obtain residual values, calculate the mean μ and variance σ² of the residual value sequence, and determine the residual threshold; 8) Substitute the fault data into the 6 ARMA models respectively, conduct real-time prediction and application, and conduct fault prediction through the determination criterion of exceeding the residual threshold continuously for 3 times and the collaborative alarm of three parameters.

2. The method for predicting faults of a liquid oxygen-methane rocket engine based on an ARMA model according to claim 1, wherein, The simulation model is constructed through the MATLAB / Simulink platform, which is a component-level model containing thermodynamic equations and fluid dynamics equations, and includes the dynamic characteristics of the full operating conditions of startup, steady state, and shutdown; the verification error of the simulation results of the normal data is less than or equal to 2%.

3. The method for predicting faults of a liquid oxygen-methane rocket engine based on an ARMA model according to claim 1, characterized in that The 6 ARMA models in step 6) are initially ordered through ACF / PACF to judge the truncation and tailing of the time series, and estimate the numerical range of the autoregressive model order p and the moving average model order q. Among them, the autoregressive model order p ≤ 4, and the moving average model order q ≤ 6; then determine its exact value through the AIC criterion to complete the model order determination, and use the maximum likelihood estimation method to estimate the ARMA model parameters to complete the establishment of the ARMA model.

4. The method for predicting faults of a liquid oxygen-methane rocket engine based on the ARMA model according to claim 1, characterized in that, The residual threshold in step 7) is obtained by statistically analyzing the residual distribution of normal data to ensure that more than 95% of the data falls within the threshold interval, the bandwidth coefficient a = 2, and the threshold calculation formula is: threshold = μ ± 2σ.

5. The method for predicting the faults of a liquid oxygen-methane rocket engine based on the ARMA model according to claim 1, wherein The determination criterion in step 8) includes single key parameter alarm and system alarm; among them, For the single key parameter alarm, it is necessary to satisfy that the residuals of 3 consecutive sampling points exceed the threshold, and the time interval between every two adjacent sampling points is set to 0.002 s; For the system alarm, at least 3 key parameters need to alarm, and the alarm time of the third key parameter is taken as the final result.

6. The method for predicting the faults of a liquid oxygen-methane rocket engine based on the ARMA model according to claim 1, wherein, The range of the failure factor Ka in step 3) is set to 0.7 ≤ Ka ≤ 1.3, and its injection methods include: cavitation of the methane pump: Ka = 0.7 - 0.9, reducing the outlet pressure and outlet flow rate of the methane pump; ablation of the methane turbine blade: Ka = 0.7 - 0.9, reducing the flow rate of the methane turbine; leakage of the methane gas path: Ka = 0.7 - 0.9, reducing the turbine drive flow rate; decline in the combustion efficiency of the gas generator: Ka = 0.7 - 0.9, incomplete combustion of methane and oxygen, reducing the generated gas and heat; ablation of the thrust chamber: Ka = 1.1 - 1.3, gradual degradation of the inner wall material of the thrust chamber, increasing heat loss.

7. A liquid oxygen-methane rocket engine fault prediction system based on the ARMA model, characterized in that, It includes a data acquisition module, an edge computing module, a threshold management module, and an alarm decision-making module; Among them; The data acquisition module performs 24-bit ADC analog-to-digital conversion with a sampling rate of 100 Hz, covering 6 key parameters; The edge computing module is based on an ARM Cortex-A72 processor, integrated with an ARMA prediction algorithm, and the calculation delay < 5 ms; The threshold management module stores the mean and variance of the residuals of each parameter and dynamically updates the threshold; The alarm decision-making module applies a continuous determination criterion and outputs the fault type and alarm time.

8. The liquid oxygen methane rocket engine fault prediction system based on the ARMA model according to claim 7, characterized in that, The data acquisition module includes a pressure sensor with an accuracy of ±0.1% FS, a speed sensor with a resolution of 0.01 rpm, a flow sensor with a linearity of 0.2%, and an anti-aliasing filter with a cut-off frequency of 40 Hz.

9. The liquid oxygen-methane rocket engine fault prediction system based on the ARMA model according to claim 7, characterized in that, The fault prediction system further includes a human-machine interaction interface for real-time display of parameter curves, residual plots, and alarm information, and a historical database for storing the most recent 1000 sets of prediction data and supporting fault traceability analysis.

Images