Online fault detection method for reusable liquid rocket engine

Abstract

The invention provides an online fault detection method for a reusable liquid rocket engine, and aims to realize real-time online adjustment of a fault threshold according to the change of the working condition of the engine under the condition of large-range variable thrust of the reusable engine so as to form a red line shutdown strategy based on an online dynamic threshold. Faults occurring in the working process of the engine are pre-warned and judged, so that effective measures are taken in time, and the safety of the carrier rocket and the load is protected.

Description

Technical Field

[0001] The invention belongs to the technical field of aerospace reliability, and in particular relates to an online fault detection method for a reusable liquid rocket engine. Background Art

[0002] Reusable launch vehicles are an important development direction for future aerospace. Liquid rocket engines are the core products of launch vehicles. Whether the engine works normally directly determines the success or failure of the launch vehicle's flight mission.

[0003] With the development of reusable launch technology, reusable launch vehicles will have a certain power redundancy capability, that is, when one or several engines fail, the failed engines can be shut down, and the thrust and swing angle of other normally working engines can be adjusted to ensure the success of the flight mission. For example, the first stage of SpaceX's Starship is composed of 33 Raptor engines. According to the Starship's carrying capacity and attitude adjustment capability, the success of the flight mission can still be guaranteed when 5 engines fail. Through power redundancy, although a certain load transportation capacity is sacrificed, the requirements for engine mission reliability are reduced, which greatly improves the mission reliability level of the launch vehicle.

[0004] Reusable liquid rocket engines will have power redundancy capabilities, and timely and effective measures will be taken after engine failure to protect the safety of the launch vehicle and payload. The significance of online engine fault diagnosis is prominent, and its measurement and control system will participate in flight decision-making. The occurrence and development of engine failures are extremely rapid and extremely destructive. If the failure is not handled in time, explosions will occur, affecting the safety of the launch vehicle and payload. Therefore, engine fault diagnosis needs to be timely.

[0005] Fault diagnosis of liquid rocket engines usually uses the size of signals such as pressure, flow, speed, vibration, and temperature measured by sensors at key parts of the engine to determine whether the engine is abnormal. Under normal operating conditions, the parameters of the engine are in a relatively stable range. Once a fault occurs, some signals may increase or decrease significantly. Once a certain range is exceeded, the engine is judged to be faulty. For single-use engines, the thrust adjustment range during flight is very small, generally within 5%. Therefore, a fixed threshold red line shutdown strategy is usually adopted, that is, under rated thrust, the key parameter signals such as pressure, flow, speed, vibration, and temperature of various parts of the engine are all in a relatively stable range. Within this range, a certain margin is added upward and downward, generally 10%-30%, as the criterion for online fault diagnosis of the engine. For reusable engines, a large range of thrust changes are required during operation, and the adjustment range will exceed 40%. Large adjustments to the engine thrust will cause large changes in the key parameters of various parts of the engine. During the change process, the parameters are in an unstable state, and the fixed threshold judgment method cannot meet the needs of online fault diagnosis of the engine. Summary of the invention

[0006] In view of this, an object of the present invention is to provide an online fault detection method for a reusable liquid rocket engine, which can realize online fault detection of the engine.

[0007] An online fault detection method for a reusable liquid rocket engine, including an online fault detection method in a steady-state working phase, specifically:

[0008] For the collected fault characteristic parameter sequence, the autoregressive moving average model ARMA is used to fit and predict the future fault characteristic parameter values:

[0009]

[0010] Where: x t represents the predicted value of a certain fault characteristic parameter at time t; x t-1 、x t-2 , …, x t-n represents the measured value of a certain fault characteristic parameter at time t-1, t-2, ..., tn; a t-1 , a t-2 , …, a t-m They respectively represent the difference between the measured value of the fault characteristic parameter at time t-1 and the measured value at time t-2, the difference between the measured values ​​at time t-2 and t-3, and the difference between the measured values ​​at time t-m+1 and tm; represents the autoregressive parameter; N represents the autoregressive order; θ j represents the sliding average parameter; M represents the sliding average order; represents white noise interference;

[0011] The predicted fault characteristic parameter value at each moment is multiplied by a set coefficient to obtain a red line shutdown threshold, and the measured value of the fault characteristic parameter is compared with the red line shutdown threshold to determine whether a fault occurs.

[0012] Furthermore, it also includes an online fault detection method in the variable operating condition stage, specifically:

[0013] 1) Construct the functional relationship between the steady-state operating condition and the engine's fault characteristic parameters:

[0014] K=a·F+b (2)

[0015] In the formula, K represents the fault characteristic parameter of the engine; F represents the working condition before the engine adjustment;

[0016] 2) Aiming at the delayed response and non-steady-state change process of variable working conditions, the changing law of various parameters in the process of engine working condition change is simulated, and the parameter t d , t, F0, F1 as input, fault characteristic parameter K as output, and use machine learning method for training to form the engine change law of the working condition change process, that is, the fault characteristic parameter K and parameter t d , t, F0, F1 mapping relationship:

[0017] K=f(t d ,t,F0,F1) (3)

[0018] In the formula, t d Indicates the delay time of the working condition adjustment; t indicates the start time of the engine working condition adjustment; F0 indicates the working condition before the engine adjustment; F1 indicates the working condition after the engine adjustment;

[0019] 3) According to the working condition adjustment values ​​before and after the working condition change, the fault characteristic parameter value K0 before the working condition adjustment and the fault characteristic parameter value K1 after the working condition adjustment are calculated according to formula (2), that is, the upper and lower limit values ​​of the fault characteristic parameters before and after the working condition adjustment are obtained; and then the curve of the fault characteristic parameter K is obtained according to formula (3);

[0020] 4) Based on the curve of the fault characteristic parameter K obtained above, after multiplying it by the set coefficient, the red line shutdown threshold at each moment under the variable working condition is obtained, and the measured value of the fault characteristic parameter at each moment is compared with the red line shutdown threshold to determine whether there is a fault.

[0021] Furthermore, in the online fault detection method in the steady-state working stage, before predicting the future fault characteristic parameter values, the traditional fixed red line method is used for the extracted fault characteristic parameter data to determine whether the fault characteristic parameters are within the red line range. If they are outside the red line, an abnormal signal is issued and it is determined to be a fault.

[0022] Furthermore, in the online fault detection method in the steady-state working stage, before predicting the future fault characteristic parameter values, it is judged whether the data is stable; if the data is unstable, the data is stabilized by using methods such as differential calculation; if the data is stable, the ARMA model is used for fitting and the future fault characteristic parameter values ​​are predicted.

[0023] The present invention has the following beneficial effects:

[0024] The purpose of the present invention is to provide an online fault detection method for a reusable liquid rocket engine. Aiming at the situation of a reusable engine with a large range of variable thrust, the method follows the changes in the engine operating conditions, realizes real-time online adjustment of the fault threshold, forms a red line shutdown strategy based on the online dynamic threshold, and gives early warning and judgment on faults occurring during the operation of the engine, so as to take effective measures in time to protect the safety of the launch vehicle and the payload. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a certain type of engine operating condition adjustment curve in an embodiment of the present invention;

[0026] Figure 2 is the measured value of the pressure at a key position during steady-state operation of a certain type of engine in an embodiment of the present invention;

[0027] Figure 3 It is a flow chart of the implementation of the adaptive threshold and alarm strategy of the present invention;

[0028] Figure 4 A comparison diagram between the actual measured value and the predicted result of a parameter of a certain engine in an embodiment of the present invention;

[0029] Figure 5 The adaptive threshold calculation result of a parameter of a certain engine in a steady-state working state in an embodiment of the present invention;

[0030] Figure 6 It is a curve diagram of changes in key parameters of the engine during the operating condition change process in an embodiment of the present invention;

[0031] Figure 7 is a functional relationship between a certain type of engine operating condition and key parameters in an embodiment of the present invention;

[0032] Figure 8A comparison diagram of actual curves of changes in key engine parameters and calculation results during operating condition changes in an embodiment of the present invention;

[0033] Fig. 9 This is the online threshold warning area during the entire working process of a certain type of engine in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0035] Liquid rocket engines need to adjust their operating conditions according to the flight mission requirements. The engine mainly adjusts the engine thrust to achieve the purpose of operating condition adjustment. The operating condition adjustment and key parameter change curves of a certain type of engine are shown in the figure below. Figure 1 As shown, the sampling frequency is 100 Hz, the horizontal axis is the sampling sequence, the nth sampling point in actual work, that is, the working condition value at 10*n milliseconds; the vertical axis is the percentage of the rated thrust, and the 100% working condition is the rated thrust.

[0036] The engine mainly monitors the working status in real time through sensor signals such as pressure, flow, speed, temperature, and vibration. The threshold of each sensor is set by the red line method, and whether the parameters collected by the sensor exceed the red line threshold is used as the basic basis for engine fault diagnosis. In order to minimize the possibility of false shutdown, the multi-sensor joint judgment strategy is used to determine whether the engine needs to execute the shutdown procedure. The sensor is the smallest detection unit for fault diagnosis. The determination of each sensor threshold affects the timeliness and accuracy of fault diagnosis. The fixed red line method is mainly used for fault detection in a single-use engine. If the key parameters actually measured exceed the threshold, the engine will be shut down urgently. During the flight, the engine needs to adjust the working condition according to the needs of the launch vehicle. In order to improve the accuracy of fault detection and make fault detection more timely, a red line shutdown method based on dynamic threshold is studied. It can be seen from the thrust adjustment curve in the figure above that the engine mainly includes steady-state working stage and variable working condition stage during operation. For these two situations, online fault detection methods are given based on adaptive threshold adjustment schemes. .

[0037] Embodiment 1: Online fault detection method in steady-state operation phase

[0038] Due to the harsh working conditions of the engine, violent combustion and high-speed rotating parts, even in the steady-state working stage, the key parameters of the engine still have certain fluctuations, as shown in the figure below. Figure 1 The working condition adjustment curve shows that in the 0s-4s period, the engine did not adjust the working condition. The measured pressure value of a key part of the engine is as follows: Figure 2 shown.

[0039] Under stable working conditions, the engine does not send out working condition adjustment signals, the working conditions do not change, and the fault characteristic parameters such as engine pressure, flow, and speed are in a relatively stable state. The data acquisition system collects relevant data at a certain sampling frequency, which is generally 100Hz. For the fault characteristic parameters under steady-state working conditions, they belong to time series signals with stationary characteristics, namely:

[0040] If any subset of the time series Satisfy: Expected μ t =E(y t )=μ y is a finite constant, that is, it does not change with time t; the covariance γ s,t = Cov(y s ,y t ) exists and only depends on the interval size |st| of s and t; then the sequence is called a weakly stationary (covariance stationary, wide stationary) sequence, usually referred to as a stationary sequence.

[0041] For stationary time series, the autoregressive moving average model (ARMA model) is used to fit and predict future time series. The ARMA model combines the autoregressive model with the moving average model to describe complex sequence dependencies in a very concise model form. The basic form of the ARMA model is shown below:

[0042]

[0043] Where: x t represents the predicted value of a certain fault characteristic parameter at time t; x t-1 、x t-2 , …, x t-n represents the measured value of a certain fault characteristic parameter at time t-1, t-2, ..., tn; a t-1 , a t-2 , …, a t-m They respectively represent the difference between the measured value of the fault characteristic parameter at time t-1 and the measured value at time t-2, the difference between the measured values ​​at time t-2 and t-3, and the difference between the measured values ​​at time t-m+1 and tm; represents the autoregressive parameter; N represents the autoregressive order; θ j represents the sliding average parameter; M represents the sliding average order; Represents white noise interference.

[0044] According to the characteristics of data collection of fault characteristic parameters such as engine pressure, speed, flow, etc., the collected test data is used to fit the data of steady-state conditions using the ARMA model, and then the fault characteristic parameters are predicted. The predicted value is compared with the actual value. If it is greater than the red line shutdown threshold, an alarm signal of abnormal engine operation is given. The implementation process of the steady-state condition adaptive threshold and alarm strategy is as follows: Figure 3 shown.

[0045] During the engine start-up process or after a period of time after the signal of changing the working condition is issued, the working condition of the engine is in a non-steady state. The threshold algorithm for the changing working condition process will be studied in the next section. After the engine working condition is stable, the judgment is made according to the following process:

[0046] 1) Extract the time period from t0 to t1 (assuming t s =t1-t0) fault characteristic parameter data;

[0047] 2) During this period, the fixed red line method is used to determine whether the fault characteristic parameters in the time period t0 to t1 are within the red line range. If they are outside the red line, an abnormal signal is issued;

[0048] 3) If the fault characteristic parameter data in the time period t0 to t1 are all within the red line range, the Augmented Dickey-Fuller method is used to test whether the data is stable;

[0049] 4) If the data is unstable, use methods such as differential calculation to stabilize the data;

[0050] 5) If the data is stationary, the ARMA model is ordered using the data from the time period t0 to t1. The order n of the AR model and the order m of the MA model are determined by the AIC criterion to ensure the fitting accuracy of the model.

[0051] 6) After the model is determined, the ARMA model is trained using the data from the time period t0 to t1 to determine the model parameters, including the autoregressive parameters and the moving average parameters;

[0052] 7) Use the trained ARMA model to predict the fault characteristic parameter value P at the future time f Make predictions;

[0053] 8) The actual measured value P m and the predicted value P f Calculate the difference. If the difference exceeds the set threshold range, an abnormal signal is issued;

[0054] 9) If it does not exceed the threshold range, the latest data of the ts time period is used to predict the fault characteristic parameters at the next moment until the instruction of changing the working condition is received, and then the interpretation of the steady state stage is ended;

[0055] 10) After the operating condition changes and the system enters the next steady-state stage, the above process is repeated to calculate the online adaptive threshold and determine whether the parameters are abnormal, thereby achieving online fault detection.

[0056] The online dynamic threshold analysis method based on ARMA time series mainly establishes the mapping relationship between the future signal size and the current signal size through fitting the stable time series, predicts the signal fluctuation, and thus predicts the future value through the current measurement value (based on the signal establishment of stability). When the deviation between the predicted future value and the actual measured value is large to a certain extent, it is determined that the system has a fault, such as Figure 4 As shown, it is a comparison diagram between the actual measured value and the predicted result of a certain parameter of a certain engine in an embodiment of the present invention.

[0057] According to the given upper and lower threshold ratios, such as ±10%, the adaptive threshold under steady-state working conditions can be obtained, such as Figure 5 If the actual measured value exceeds the upper and lower thresholds, the engine alarm will sound.

[0058] Embodiment 2:

[0059] Online fault detection method during variable operating conditions:

[0060] The purpose of changing the engine operating condition is generally to adjust the engine thrust by adjusting the valve opening and changing the propellant flow rate. Through the adjustment of the operating condition, the engine can achieve thrust regulation within a certain range to meet the flight mission requirements. During the engine operating condition adjustment process, the key parameters related to the red line shutdown, such as thrust, flow, pressure, and speed, will also change accordingly, and are positively correlated with the range of the operating condition adjustment. Therefore, during the operating condition adjustment process, the shutdown red line of the key parameters should also be adjusted accordingly to achieve adaptive adjustment of the threshold.

[0061] The operating condition change command is issued by the control computer. After the control computer issues the operating condition adjustment signal, the valve control motor begins to respond and drives the valve opening to adjust, thereby adjusting the engine operating condition. Since there is a certain delay effect between the motor response and the engine response after the valve opening adjustment, the engine parameters change nonlinearly during the operating condition change. Take the speed data of a certain engine as an example. Figure 6 As shown, the operating condition adjustment command was issued at 66 seconds, lasting nearly 1 second, and the steady-state operation of the new operating condition was completed at 67 seconds.

[0062] For the variable operating conditions stage, determine the adaptive threshold according to the following steps:

[0063] 1) Fitting the mapping relationship between engine steady-state operating conditions and key parameters

[0064] Through the historical engine test data or theoretical calculation model, the least squares method is used to obtain the mapping relationship between engine thrust and key parameters. The relationship between a certain engine sensor parameter and the operating condition is approximately linear. Through the least squares method fitting, the functional relationship between the steady-state operating condition and the engine's fault characteristic parameters is constructed:

[0065] K=a·F+b (2)

[0066] In the formula, K represents the fault characteristic parameter of the engine; F represents the working condition before the engine adjustment; the fitting effect is as follows Figure 6 shown.

[0067] 2) Construct the functional change relationship under the non-steady-state process:

[0068] In view of the delayed response and non-steady-state change process of variable working conditions, the changing law of various parameters in the process of engine working condition change is simulated, and the parameter t d , t, F0, F1 as input, fault characteristic parameter K as output, and use neural network and other machine learning methods for training to form the engine change law of the working condition change process, that is, the fault characteristic parameter K and parameter t d , t, F0, F1 mapping relationship:

[0069] K=f(t d ,t,F0,F1) (3)

[0070] Where, t d Indicates the delay time of the operating condition adjustment; t indicates the starting time of the engine operating condition adjustment; F0 indicates the operating condition before the engine adjustment; F1 indicates the operating condition after the engine adjustment.

[0071] 3) Description of key engine parameters during variable operating conditions

[0072] According to the working condition adjustment values ​​before and after the working condition change, the fault characteristic parameter value K0 before the working condition adjustment and the fault characteristic parameter value K1 after the working condition adjustment are calculated according to formula (2), that is, the upper and lower limit values ​​of the fault characteristic parameters before and after the working condition adjustment are obtained; then, according to formula (3), the working condition change process is fitted to form a description of the key parameters of the start from the steady-state working condition to the variable working condition process, and then to the next steady-state working condition, that is, the curve of the fault characteristic parameter K is obtained;

[0073] 4) Fault detection

[0074] Based on the curve of the fault characteristic parameter K obtained above, after multiplying it by the set coefficient, the red line threshold is obtained, and the measured value of the fault characteristic parameter at each moment is compared with the red line threshold to determine whether it is abnormal.

[0075] During the adjustment process of an engine working condition, the speed parameter changes and the fitting results are as follows: Figure 8 As shown, the actual measured values ​​of engine parameters will fluctuate to a certain extent. The green fluctuation line is the real data, and the blue dotted line is the calculated data. The calculated value can well reflect the actual change process, and the determination coefficient reaches 0.9982.

[0076] The steady-state working condition and the marginal working condition process are described by the "adaptive red line shutdown threshold determination method in the steady-state working stage" and the "adaptive red line shutdown threshold determination method in the variable working condition stage". In order to prevent the wrong shutdown, the fixed multiple of the actual working condition is determined as the critical threshold of the engine red line. The online threshold alarm red line of a certain type of engine during the whole working process is as follows: Fig. 9 As shown, the yellow area is a safe area. Once the signal actually collected by the sensor exceeds the yellow area, it is judged as abnormal.

[0077] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for online fault detection of a reusable liquid rocket engine, characterized in that: Including online fault detection method in steady-state working stage, specifically: For the collected fault characteristic parameter sequence, the autoregressive moving average model ARMA is used to fit and predict the future fault characteristic parameter values: Where: x t represents the predicted value of a certain fault characteristic parameter at time t; x t-1 、x t-2 , …, x t-n represents the measured value of a certain fault characteristic parameter at time t-1, t-2, ..., tn; a t-1 , a t-2 , …, a t-m They respectively represent the difference between the measured value of the fault characteristic parameter at time t-1 and the measured value at time t-2, the difference between the measured values ​​at time t-2 and t-3, and the difference between the measured values ​​at time t-m+1 and tm; represents the autoregressive parameter; N represents the autoregressive order; θ j represents the sliding average parameter; M represents the sliding average order; represents white noise interference; The predicted fault characteristic parameter value at each moment is multiplied by a set coefficient to obtain a red line shutdown threshold, and the measured value of the fault characteristic parameter is compared with the red line shutdown threshold to determine whether a fault occurs.

2. A reusable liquid rocket engine online fault detection method as claimed in claim 1, characterized in that: It also includes an online fault detection method during the variable operating condition stage, specifically: 1) Construct the functional relationship between the steady-state operating condition and the engine's fault characteristic parameters: K=a·F+b (2) In the formula, K represents the fault characteristic parameter of the engine; F represents the working condition before the engine adjustment; 2) Aiming at the delayed response and non-steady-state change process of variable working conditions, the changing law of various parameters in the process of engine working condition change is simulated, and the parameter t d , t, F0, F1 as input, fault characteristic parameter K as output, and use machine learning method for training to form the engine change law of the working condition change process, that is, the fault characteristic parameter K and parameter t d , t, F0, F1 mapping relationship: K=f(t d ,t,F0,F1) (3) Where, t d Indicates the delay time of the working condition adjustment; t indicates the start time of the engine working condition adjustment; F0 indicates the working condition before the engine adjustment; F1 indicates the working condition after the engine adjustment; 3) According to the working condition adjustment values ​​before and after the working condition change, the fault characteristic parameter value K0 before the working condition adjustment and the fault characteristic parameter value K1 after the working condition adjustment are calculated according to formula (2), that is, the upper and lower limit values ​​of the fault characteristic parameters before and after the working condition adjustment are obtained; and then the curve of the fault characteristic parameter K is obtained according to formula (3); 4) Based on the curve of the fault characteristic parameter K obtained above, after multiplying it by the set coefficient, the red line shutdown threshold at each moment under the variable working condition is obtained, and the measured value of the fault characteristic parameter at each moment is compared with the red line shutdown threshold to determine whether there is a fault.

3. A reusable liquid rocket engine online fault detection method as claimed in claim 1 or 2, characterized in that: In the online fault detection method in the steady-state working stage, before predicting the future fault characteristic parameter values, the traditional fixed red line method is used for the extracted fault characteristic parameter data to determine whether the fault characteristic parameters are within the red line range. If they are outside the red line, an abnormal signal is issued and it is determined to be a fault.

4. A reusable liquid rocket engine online fault detection method as claimed in claim 1 or 2, characterized in that: In the online fault detection method in the steady-state working stage, before predicting the future fault characteristic parameter values, it is judged whether the data is stable; if the data is unstable, the data is stabilized by using methods such as differential calculation; if the data is stable, the ARMA model is used for fitting and the future fault characteristic parameter values ​​are predicted.

Images